FOOD AND DRUG ADMINISTRATION

 

 

NINETY-FIFTH MEETING OF THE

CARDIOVASCULAR AND RENAL DRUGS ADVISORY COMMITTEE

 

 

 

 

 

 

 

 

 

 

 

 

 

8:30 a.m.

Friday, January 18, 2002

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Kennedy Ballroom

Holiday Inn

8777 Georgia Avenue, N.W.

Washington, D.C.

ATTENDEES

COMMITTEE MEMBERS:

JEFFREY BORER, M.D., Chairman

Director, Division of Pathophysiology

Weill Medical College

Cornell University

525 East 68th Street, Room F467

New York, New York 10021

JAIME HENRIQUEZ, Executive Secretary

Advisors and Consultants Staff

Center for Drug Evaluation and Research

Food and Drug Administration

5630 Fishers Lane

Rockville, Maryland 20857

MICHAEL F. ARTMAN, M.D.

Professor of Pediatrics

Pediatric Cardiology

New York University Medical Center

530 First Avenue, FPO Suite 9-V

New York, New York 10016

BLASE A. CARABELLO, M.D.

Chief, Medical Service

Veterans Affairs Medical Center

Medical Service (111)

2002 Holcombe Boulevard

Houston, Texas 77030

SUSANNA LEE CUNNINGHAM, PH.D.

Professor, Department of Biobehavioral

Nursing & Health Systems

School of Nursing, Box 357266

Seattle, Washington 98195-7286

THOMAS FLEMING, PH.D.

Professor and Chair

Department of Biostatistics

University of Washington

Box 357232

Seattle, Washington 98195-7232

ATTENDEES (Continued)

COMMITTEE MEMBERS: (Continued)

ALAN T. HIRSCH, M.D.

Associate Professor of Medicine and Radiology

Minnesota Vascular Diseases Center

University of Minnesota Medical School

420 Delaware Street, S.E., Box 508 FUMC

Minneapolis, Minnesota 55455

BEVERLY H. LORELL, M.D.

Professor of Medicine

Beth Israel Deaconess Medical Center

Cardiovascular Division

330 Brookline Avenue

Boston, Massachusetts 02215

STEVEN NISSEN, M.D., F.A.C.C.

Vice Chairman, Department of Cardiology

Professor of Medicine

Ohio State University

The Cleveland Clinic Foundation

9500 Euclid Avenue, F15

Cleveland, Ohio 44195

 

SPECIAL GOVERNMENT EMPLOYEES:

ROBERT KREISBERG, M.D.

Interim Dean, School of Medicine

University of South Alabama

307 University Boulevard, Room 170 CSAB

Mobile, Alabama 36688

ATTENDEES (Continued)

GUESTS:

PAUL THOMPSON, M.D.

Hartford Hospital

Division of Cardiology

80 Seymour Street

Hartford, Connecticut 06102

TERJE PEDERSEN, M.D.

Aker University Hospital

Cardiology Department

N 0514 Oslo, Norway

 

FOOD AND DRUG ADMINISTRATION STAFF:

RAYMOND LIPICKY, M.D.

ROBERT TEMPLE, M.D.

 

SPONSOR REPRESENTATIVES:

RENE BELDER, M.D.

DONALD A. BERRY, PH.D.

FRED FIEDOREK, M.D.

CHARLES H. HENNEKENS, M.D.

THOMAS A. PEARSON, M.D., PH.D.

FRANK SACKS, M.D.

ANDREW TONKIN, M.D.

C O N T E N T S

NDA 21-387, Pravastatin/Aspirin

Bristol-Myers Squibb, Co-package,

for Long-term Management to Reduce the Risk of Death,

Nonfatal Myocardial Infarction,

Myocardial Revascularization Procedures,

and Ischemic Stroke in Patients with

Clinically Evident Coronary Heart Disease

* * *

AGENDA ITEM PAGE

 

CONFLICT OF INTEREST STATEMENT

By Jaime Henriquez 6

 

SPONSOR PRESENTATION

Introduction

By Dr. Fred Fiedorek 8

Efficacy and General Considerations

By Dr. Rene Belder 17

Efficacy - Meta-analyses

By Dr. Donald A. Berry 41

The Medical Need

By Dr. Thomas A. Pearson 79

 

APPROPRIATE DOSE AND RELATIONSHIP

TO CHOLESTEROL LOWERING

By Dr. Terje Pedersen 114

 

COMMITTEE DISCUSSION OF QUESTIONS PRESENTED 127

P R O C E E D I N G S

(8:30 a.m.)

DR. BORER: Good morning. We'll begin now the second day of the 95th meeting of the Cardiovascular and Renal Drugs Advisory Committee.

This morning we'll be considering NDA 21-387 for the combination of pravastatin and aspirin. Before we begin, Jaime Henriquez will present the conflict of interest statement.

MR. HENRIQUEZ: Conflict of interest statement. The following announcement addresses the issue of conflict of interest with regards to this meeting, and is made part of the record to preclude even the appearance of such at this meeting.

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

In accordance with 18 U.S.C. 208(b)(3), a full waiver has been granted to Dr. Alan Hirsch for unrelated speaking for the sponsor. He received between $5,000 and $10,000 a year.

A copy of the waiver statement may be obtained by submitting a written request to the agency's Freedom of Information Office, Room 12A-30 of the Parklawn Building.

With respect to FDA's invited guests, there are reported interests which we believe should be made public to allow the participants to objectively evaluate their comments.

Dr. Terje Pedersen would like to disclose that he has lectured for and received speaking fees from Bristol-Myers Squibb.

Dr. Paul Thompson would like to disclose that one of his daughters, age 27, owns 200 shares of stock in Bristol-Myers Squibb. He co-manages the account with her. In addition, he has received grant research support from Bristol-Myers Squibb.

In the event that the discussions involve any other products or firms not already on the agenda for which an FDA participant has a financial interest, the participants are aware of the need to exclude themselves from such involvement, and their exclusion will be noted for the record. With respect to all other participants, we ask in the interest of fairness that they address any current or previous financial involvement with any firms whose products they wish to comment upon.

DR. BORER: Okay. There's no comment about that. We'll move on to the sponsor's presentation.

As I said, there's an application for approval of the combination of pravastatin and aspirin to be co-packaged, first in the same package, then in the same pill, for long-term management to reduce the risk of death, nonfatal myocardial infarction, myocardial revascularization procedures, and ischemic stroke in patients with clinically evident coronary heart disease.

The sponsor's presentation will be introduced by Dr. Fiedorek.

DR. FIEDOREK: Good morning, ladies and gentlemen, committee members, FDA, and everyone else who is here today in Silver Spring. My name is Fred Fiedorek, actually pronounced like the hat "fedora," with an EK instead of an A, and it's my pleasure to be here to talk to you about pravastatin/aspirin as an important product for secondary prevention of cardiovascular disease.

I should add that during my formative years I was also trained at Washington University in St. Louis and the University of North Carolina as an endocrinologist, and so doing research and treating patients primarily with type 2 diabetes makes me aware of the need for secondary prevention in diabetics and other patients with similar problems.

On behalf of my colleagues and our consultant panels here today, I am going to discuss and review with you an overview of what's going to be presented regarding data, meta-analysis on these data, and the public health and medical need for pravastatin/aspirin. I think we will show you today that there's a strong rationale, based on the best available evidence, to support such a combination product.

However, before I begin, I do want to spend a little bit of time on the scope of this problem in the United States. As you can see in this slide, which documents the top five causes for death in the United States for both men and women, cardiovascular disease and cancer certainly dominate. We're not going to be talking about accidents and why men seem to suffer from accidents and not women. We're focusing on the leading cause, cardiovascular disease for both and strikingly for women, and it's this condition that we're talking about in terms of offering pravastatin/aspirin as a preventative product for secondary prevention to prevent these deaths.

There has been progress in this area, and it's been well documented over the last two decades. The green line here shows a reduction in the age-adjusted mortality rate that really occurs for a variety of reasons, including improvements in acute coronary care, better diet and exercise recommendations, better medicines, and all of this has led to this reduction in age-related mortality.

However, you can see from the blue bars that the overall number of deaths for the whole population, admittedly a population that's increasing in size, has remained constant. So, if you really put these two together and sort of think implicitly about something that's not on this figure, you'll realize that as more and more patients survive acute events, have established clinically evident coronary artery disease, there is a need to prevent them from having recurrent events. In fact, in patients such as myocardial infarction patients, approximately 80 or 90 percent of those patients are the ones that ultimately die from a subsequent cardiovascular event.

So, to move on to the rationale of why we think pravastatin/aspirin will be quite useful in this area of secondary prevention, it's mainly three key points or features that we think will address both the clinical and the public health needs for secondary cardiac prevention.

The main features really in the first set of bullets refer to adherence and accuracy in dosing. Clearly pravastatin and aspirin are two of the core elements in the guidelines for preventing cardiovascular disease in the U.S. population, and this has been repeatedly encouraged over the last several years.

In addition, the availability now of a combination product, which is a prescription product, will allow for both patients and health care providers the assurance that they're getting correct doses for this secondary prevention problem as well as the correct product. In part of the talk you'll be hearing later, it's actually quite striking that aspirin is recommended for these patients but many times, given the availability of aspirin and OTC substitutes for aspirin, such as Tylenol, many patients actually end up on the incorrect product, not appropriate for secondary prevention in this cardiovascular disease state.

Finally, using primarily sort of a common sense argument, the availability of having one combination product with two core parts of the guidelines to prevent cardiovascular disease, pravastatin and aspirin, does offer a common sense way of reducing pill burden for patients and hopefully enhancing the convenience. Admittedly, when we all were in medical school, those of us who were physicians, this sort of idea runs counter to traditional teaching, where the importance of titrating and dosing individual components separately was emphasized. However, just recently I think it's been recognized that these sort of patients with diabetes, hypertension, cardiovascular disease, all require increasing medicines to manage their problems, and so this should be one way of helping.

Now, if we actually look at the labels for these two products, starting with aspirin approved by this committee in these many different indications over the years, aspirin is indicated for a set of both cardiovascular and cerebrovascular prevention indications in patients with clinically evident coronary heart disease. You'll see that this includes evident heart disease, including myocardial infarction, unstable angina, stable angina, and even patients who've undergone revascularization procedures.

For the prevention of cardiovascular disease, those three or four bullet points at the beginning refer to how aspirin can help in prevention. The fourth one is a much more sort of acute preventative that's been recognized as very important for aspirin.

And the final bullet point refers to how aspirin is very critical in preventing cerebrovascular disease.

If we move on to pravastatin, very similarly pravastatin also possesses an array of indications as secondary prevention for cardiovascular disease. It's indicated to reduce the risk of a variety of subsequent events in patients who have clinically evident coronary heart disease.

So, if we consider what overlap exists for these two labels -- and we offer you sort of for our combined pravastatin/aspirin product an inner section label, if you will, of both pravastatin and aspirin in a combination tablet -- we're looking to see and provide evidence to support how this co-tablet could be used in the long-term management to reduce the risk of the following cardiovascular events in patients with clinically evident coronary heart disease. These events are death, nonfatal myocardial infarction, revascularization procedures, and ischemic stroke.

I'll add that these four events, in the large pravastatin database that we'll be discussing today, represent both the primary and secondary endpoints that were actually a priori specified for these trials when they were conducted. They're part of the prespecified endpoints, and they're also the subject of the analysis of the data and the meta-analysis of all these studies that we will present today.

Moving on to the population that we want to discuss, again, I've described the indications for secondary cardiovascular disease prevention. And what population in the United States does this entail?

You'll see that the potentially eligible population is approximately 12.4 million subjects, and given the indications described previously for aspirin and pravastatin, it overlaps to a very large degree. Even if you consider possible contraindications for both aspirin and pravastatin due to the well recognized problems of GI bleeding or aspirin sensitivity for aspirin, and other contraindications for pravastatin, we're still left with a population of approximately 10.4 million patients.

So, when these two therapies are combined, what do we need in terms of the properties for a combination product, either recommended properties or required properties for a combination product?

This list is what we will discuss today, and we will cover how obviously, as is well known to everybody here, aspirin and pravastatin have different mechanisms of action, one through platelet aggregation and platelet effects, and one through lipid lowering and vessel wall effects.

We will also demonstrate data on PK, pharmacokinetics, and pharmacodynamics for these two products when administered concurrently.

We will also review, in the large pravastatin database of approximately 14,000 patients, how the safety and tolerability of these agents do not magnify any of the effects of the agents when given alone.

We will discuss in the recommended combination doses for this product how these are appropriate doses for pravastatin, given the clinical endpoint data that will be presented, as well as the appropriate doses for aspirin, given its cardiovascular and cerebrovascular prevention indications.

A large part of our presentation today will deal with efficacy, and really there are three core components of this which I'll get to later, but clearly we need to show for a combination product how pravastatin and aspirin contribute ideally in an additive fashion to efficacy, and you'll see this in the data we'll describe today.

Finally, from the point of view of preventing a leading cause of mortality in the United States for both men and women, we think this product addresses an important medical need, public health need, that is also impactful for our discussion.

Helping today in this presentation are our five-member consultant panel. The first two members, Dr. Donald Berry and Dr. Thomas Pearson, will be speakers along with me this morning. Dr. Berry is a biostatistician from the M.D. Anderson Cancer Center and he will be presenting data on the meta-analysis of our large pravastatin database. Dr. Pearson, a preventive cardiologist from the University of Rochester School of Medicine, will follow Dr. Berry and discuss the medical need, both clinical need and public health need, for this combination product.

Also here with us to answer any questions, should they arise, are Dr. Charles Hennekens from the University of Miami School of Medicine. Dr. Hennekens is founding collaborator of the Antithrombotic Trialists Collaboration and will certainly be well-placed to answer any questions on aspirin.

Additionally here are Dr. Andrew Tonkin and Dr. Frank Sacks. Dr. Tonkin and Sacks respectively were our principal investigators for the large LIPID and CARE pravastatin trials conducted over the last 10 years or so, and they will be able to take any questions specific to these trials or about medical practice for cardiovascular prevention in general.

The agenda this morning essentially mirrors the sort of recommended properties I described for a combination product. My colleague, Dr. Rene Belder, will lead off and talk about the first five bullet points. Dr. Belder has been at Bristol-Myers Squibb for 14 years, and has actually been, over the last several years, the main clinical coordinator for all of the pravastatin clinical trials. He's the glue, if you will, of the pravastatin programs.

When Rene is finished with these five topics, he will then hand over to Dr. Berry, again our biostatistician consultant, who will deal with the efficacy from the meta-analysis of all these pravastatin trials and the database that it represents, and then also discuss how the efficacy, as evidenced in these trials, particularly ones that last five years or more, really provide evidence of consistent and durable benefit for both pravastatin and aspirin when administered concurrently.

Finally, Dr. Berry will turn over to Dr. Thomas Pearson from the University of Rochester School of Medicine. He will discuss medical need, both in terms of the clinical need and public health need.

Our presentation is meant to last about an hour, assuming no interruptions. If there are interruptions -- you need to interrupt -- we'll certainly be glad to take any questions, and if you do let us go through, you can note that on the bottom of the slide in the lower right-hand corner are numbers and letters that can help you call us back up to, as needed, to answer any specific questions.

With this overview, I now want to turn over to Dr. Belder.

DR. BELDER: Good morning, ladies and gentlemen. It's a pleasure to be here today to share with you some of the results of the clinical development program with pravastatin that spans well over 15 years.

As Dr. Fiedorek already mentioned, I will address these five points with you, the mechanism of action of the components, the possibility of a pharmacokinetic interaction between pravastatin/aspirin, the safety and the tolerability of the combination, the doses that we plan to make available in this combination product, as well as the efficacy based on individual trials.

Starting with the easiest part, every one of you is aware, of course, that pravastatin and aspirin reduce cardiovascular events by different mechanisms of action. Aspirin is, of course, an inhibitor of platelet aggregation. Pravastatin reduces cholesterol levels. One would therefore expect that the benefits that these compounds have on clinical events would be independent from each other.

With respect to the pharmacokinetic interaction, we did a single dose, three-way crossover study in 30 healthy volunteers. I'll go over this slide with you so that you'll understand the data on this slide.

The left-hand panel on this slide indicates concentrations with respect to the Cmax. The right-hand panel of this slide indicates the AUC, area under the curve.

In the left two bars in each panel, you see the pravastatin concentrations. In the right two bars in each panel, you see the salicylate concentrations.

Every time you see a green bar, that means the pravastatin and aspirin were dosed at the same time. When you see a blue bar, only pravastatin was dosed. When you see an orange bar, only aspirin was dosed.

Important for the interpretation of the results for this study are the confidence intervals indicated here, here, here and here, and these are the relative concentrations. The confidence intervals indicate that the concentrations were all well within the limits set by regulatory guidelines to declare bioequivalence. So, the conclusion is that there's no pharmacokinetic interaction between pravastatin and aspirin.

With respect to the possibility of a pharmacodynamic interaction between the two products, we had some discussions with the agency before we submitted the NDA. In light of the absence of a pharmacokinetic interaction and in the light of the fact that the ultimate endpoint that we are after is clinical event reduction, we agreed that doing a pharmacodynamic interaction study would not contribute valuable information. However, we are able to show you the effect of pravastatin in the presence or absence of aspirin with respect to the effects on several lipid fractions.

You see the results from the CARE study in this slide. The green bar again means that pravastatin and aspirin were dosed. The blue bar indicates that only pravastatin was dosed. You see here the lipid-lowering efficacy with respect to total cholesterol, LDL cholesterol, triglycerides, and HDL cholesterol. And it's clear from this slide that aspirin does not influence the cholesterol-lowering efficacy of pravastatin. So, with respect to pharmacodynamic interaction, there is no pharmacodynamic interaction between pravastatin and aspirin with respect to the cholesterol-lowering efficacy of pravastatin.

Before I discuss with you the safety findings from the analysis that we did, I would like to briefly introduce to you the clinical program that we did with pravastatin.

The pravastatin atherosclerosis intervention program consisted of seven placebo-controlled trials, all randomized, 40 milligrams of pravastatin versus placebo. Highlighted here on this slide it shows you the three trials that contributed most of the data in this program. Highlighted are the two secondary prevention trials that are the topic of discussion for today. Those were the long-term intervention with pravastatin in ischemic disease study, the LIPID study, involving 9,000 subjects, the CARE study involving 4,200 subjects.

Also part of this program was the primary prevention study, the West of Scotland study, and again is not a topic of discussion for today.

Also part of this program were four regression of atherosclerosis trials. These trials had as the primary endpoint the evaluation of pravastatin with respect to the progression of atherosclerosis in coronary and carotid arteries. The three trials that are highlighted were in a secondary prevention population. These patients had all evidence of coronary artery disease. The trial that is not highlighted, the KAPS study, was a trial in patients who did not have evidence of carotid or coronary disease and was therefore a primary prevention trial. So, only these studies are being discussed today.

To put these trials in perspective and the contribution that they made to the database that we have, we developed this schematic. You can see here that the LIPID and the CARE study contributed 96 percent of the total patient-years of follow-up in these trials, and that the regression trials contributed about 4 percent of the total exposure. In total, it's a very impressive 74,000 patient-years of exposure, so it provides a very robust database to perform analysis on.

I should also emphasize here that the LIPID and the CARE study were designed as clinical event studies, and therefore complete follow-up of all subjects was attempted. And indeed, in the LIPID and CARE, there was near complete follow-up. Only one subject in the LIPID trial and one subject in the CARE trial escaped the investigators, so the final status of only two subjects was not known at the end of the studies.

This database of 74,000 patient-years of exposure forms the basis of the safety conclusions with respect to the pravastatin/aspirin combination that you see here on this slide. In the interest of time, I do not show you the data that led us to these conclusions, but you can see here the conclusions that we have with respect to some of the events that may be of interest for either a statin or for aspirin.

With respect to CK abnormalities, note that we did not have any case of rhabdomyolysis in any of the trials with pravastatin. So, we have looked at CK abnormalities, liver function test abnormalities, gastrointestinal bleeds, or hemorrhagic stroke. There was no signal with respect to the combination of pravastatin and aspirin, relative to pravastatin by itself or aspirin by itself, that there was an increased incidence of any of these events in the combination group. So, that leads us to the conclusion that the combination of pravastatin and aspirin is safe.

Since this large database was all based on a 40 milligram dose of pravastatin, it's appropriate to consider only a 40 milligram dose in this combination product. 40 milligrams is the approved starting dose for pravastatin. All prevention studies used the same pravastatin dose, 40 milligrams. This dose was extremely well tolerated and very safe and in the trials there was no down titration necessary for safety reasons. In addition, in a population like the elderly, there is no need for a lower dose of pravastatin.

In essence, pravastatin is only indicated at a lower dose in patients requiring complex management, such as patients with renal or hepatic impairment, of patients who have undergone a cardiac transplant who are on cyclosporine. We think that this combination product would not be a good idea to be used in these complex management situations.

With respect to aspirin, the label with respect to the efficacy of aspirin is clear. It advises that aspirin is effective anywhere between 75 and 325 milligrams once daily, and that therapy should be continued indefinitely.

The doses that we have chosen for this combination product are 81 and 325 milligrams. 81 milligrams was chosen because this is the most widely used dose for secondary prevention in the United States. The 325 milligram dose was chosen because this is the upper end of the approved dose range.

The key question for today may very well be whether or not we have data that show that pravastatin and aspirin is more effective than each of its components. This question can be broken down in two components. The first part is, is pravastatin/aspirin more effective than aspirin by itself? The other part is whether or not pravastatin and aspirin is more effective than pravastatin by itself.

For both of these questions we have evidence from the two largest placebo-controlled randomized trials, CARE and LIPID. I will address the first question on the basis of the LIPID and the CARE study. Dr. Berry will address the second part of the question, also on the basis of the LIPID and the CARE study, but also on the basis of the meta-analysis. In addition, Dr. Berry will address my part of the question also on the basis of the meta-analysis. So, in short, I will present to you the investigation of efficacy of pravastatin in aspirin users based on the data of the randomized controlled clinical trials, CARE and LIPID.

So, how did we define aspirin users in these trials? Aspirin users were defined as those subjects who were using aspirin at baseline. Aspirin was proactively collected as a concomitant medication in these trials, so we know whether or not the patient was taking aspirin at baseline. However, we did not rigorously collect the dose level that they were using.

We do know that adherence to the pravastatin regimen was very good. 97 percent of the patients who were using aspirin at baseline were still using aspirin at the end of the studies.

The endpoints that we evaluated for this investigation are, of course, the primary endpoints for the individual trials. For LIPID it was coronary mortality. For CARE it was coronary mortality or nonfatal MI.

In addition, we considered several other endpoints for this analysis. These endpoints are based on the overlap of the pravastatin and aspirin labels, and there are two endpoints that are relatively narrowly defined, fatal and nonfatal MI, and ischemic stroke, and a more broadly defined endpoint of coronary mortality, nonfatal MI, revascularization procedures, or ischemic stroke. Each of these endpoints were prospectively defined as endpoints in all of the trials that we included in the analyses.

Starting with the results of the LIPID study, this is a brief overview. The LIPID trial was a trial in 9,000 subjects who qualified on the basis of either myocardial infarction or unstable angina. The mean follow-up was 6.1 years. As said before, the primary endpoint was coronary mortality, and the patients were randomized to 40 milligrams of pravastatin or placebo. 83 percent of the patients were using aspirin.

These are the results for all subjects for the primary endpoint of coronary death. You can see here that pravastatin reduced coronary mortality by 24 percent, which was highly statistically significant, with a p value of .001.

We now investigate the effect of pravastatin on top of aspirin, so we're effectively investigating the combination of pravastatin plus aspirin, versus aspirin by itself. Here again, for the primary endpoint, coronary mortality, we see a 28 percent risk reduction, which was highly statistically significant.

For the other endpoints that we evaluated for this analysis, fatal or nonfatal MI, ischemic stroke, and the composite endpoint, again very similar risk reductions, all of which were statistically significant.

Of note, I would like to point out that despite aspirin use, almost 30 percent of these patients in the placebo group, despite aspirin use, still had an event, and adding pravastatin to the aspirin regimen cut that risk by one-quarter.

Now going over to the CARE trial, the CARE trial was a trial in 4,200 post-MI subjects. Mean follow-up was 5 years. Patients all had normal cholesterol levels in order to qualify for this trial, and the primary endpoint was nonfatal MI or coronary mortality. Patients were again randomized to placebo or 40 milligrams of pravastatin. 84 percent of the patients were also taking aspirin.

Again, we start with the primary endpoint in all subjects. We see here that for the primary endpoint, nonfatal MI or coronary heart disease death, a 24 percent risk reduction, highly statistically significant.

Now let's investigate the combination of pravastatin plus aspirin versus aspirin by itself. Again, here for the primary endpoint of the CARE study, a 28 percent risk reduction that was highly statistically significant. The other endpoints considered for this analysis, you can see that for these three endpoints there were similar risk reductions that were statistically significant for two out of the three endpoints considered.

The conclusion from these analyses is that the combination of pravastatin and aspirin is significantly more effective than aspirin alone, as evidenced by the randomized comparisons from secondary prevention trials, LIPID and CARE.

The second part of the question, as I already indicated, is whether or not pravastatin plus aspirin is more effective than pravastatin alone. Ideally one would like to have a database where aspirin therapy was randomized. However, the aspirin trials were conducted before the statins were used, so we couldn't look at these databases. A placebo-controlled trial with aspirin is not feasible because of ethical reasons. However, the pravastatin database, with about 94,000 patient-years of follow-up, provided the robust database to explore this question. Hence, I would like to hand over now to Dr. Berry, who has explored this question, to answer this part of the question.

DR. BORER: Blase?

DR. CARABELLO: You indicated that aspirin was safe. But we're talking now about buffered not enteric-coated aspirin. Is that correct?

DR. BELDER: That's correct.

DR. CARABELLO: And I'm not certain of that. I'd like to see the specific data that compares buffered aspirin with enteric-coated aspirin in terms of safety. So, I hope those data will be forthcoming.

DR. BELDER: Charlie, do you have any comments on that?

DR. HENNEKENS: Well, we didn't specifically study Bufferin against enteric-coated aspirin, but in the Physicians Health Study of 22,071 men, who were randomized to 325 milligrams of Bufferin or placebo on alternate days, after 5 years of treatment and follow-up, the rates of GI upset were virtually identical in the aspirin and placebo groups, a small excess of the Bufferin over the placebo. The rates of GI bleeding were only slightly higher in the 325 every other day versus the placebo, and finally there was only one fatal GI hemorrhage and that was in the placebo group.

Now, I think that while the formulation is important, I think the data suggests that it's the dose of aspirin that's more important with regard to the side effects. The UK trial of TIA, which randomized patients to placebo, 300 milligrams or 1,200 milligrams of aspirin, found that the rate of GI side effects in the placebo group was 24 percent. It was 29 percent in the 300 milligram a day dose and 39 percent in the 1,200 milligram a day dose.

With regard to GI hemorrhages, the rate was 1.6 percent in the placebo group, 2.6 percent in the low dose aspirin group, 300 a day, and 4.9 percent in the 1,200 milligram a day.

So, it's clear that the higher dose is significantly greater than placebo and significantly greater than the lower dose. So, I think that the doses that are prescribed here in this combination are well within the range where the rate of the side effects are quite low, and I think it's the dose that's more important than the formulation.

DR. BELDER: In the trials that we did obviously we don't know which formulation of aspirin that patients were using. Just aspirin as a concomitant medication was collected, so it could have been any formulation that's on the market.

DR. LORELL: Thank you for a very clear presentation.

You presented very clear data from the LIPID and CARE trials regarding efficacy on endpoints. However, since those trials were done, there are now guidelines from the ACC and American Heart Association that are followed across the country, that for secondary prevention, each of us should be trying to lower LDL cholesterol to a value of less than 100. It would be very nice to see today what the probability is of achieving that explicit goal, with the use of Pravachol 40 milligrams in your data sets. I didn't see that data clearly in either your presentation or the next one, so perhaps that can be brought back to the meeting a little later.

DR. BORER: Steve and then Susanna.

DR. NISSEN: I recognize that you don't have specific information about aspirin dosages in the trials. Do we have a range? For example, were any of the people receiving, say, 650 milligrams of aspirin? Do we have any information at all about the dose of aspirin that was used in those trials? And I'm specifically interested in whether there are significant numbers of patients who had substantially higher doses of aspirin.

DR. BELDER: We don't have information about that.

DR. CUNNINGHAM: I was noticing also that these studies are predominantly male, somewhere in the range of 85 percent, 84, something like that. Do you have any data on what happens with women?

DR. BELDER: Yes, we do have a subgroup analysis in women. These are the numbers of patients in the various groups, male and female. As you can see, the split is indeed what you indicated.

Here you see the results for the expanded endpoint. Here are men, pravastatin plus aspirin versus aspirin by itself. These are the comparisons that we have so far discussed. Dr. Berry will obviously discuss the comparisons that you see here indicated in blue, which are the observational comparisons.

The point here is that for both men and women there were significant reductions.

DR. BORER: Bob?

DR. TEMPLE: Maybe I should save this for the discussion, but I think one of the presumptions of this whole thing is that aspirin is approved for these uses at doses anywhere between 80 and 325. I don't think we're primarily asking whether aspirin is effective or safe at those doses. I mean, obviously there's some GI bleeding, et cetera. The question here relates to putting them together in what is essentially a fixed combination. So, some of those things I'm not sure need to be revisited.

The other thought was, if some people took more than 325 milligrams of aspirin and you still saw an added effect of the pravastatin, that wouldn't undermine the observation, the point they're trying to make, which is that when you add to an effective dose of aspirin or even maybe super-effective dose of aspirin you get a further effect.

DR. BORER: Can I ask you how many people in your data set were over 65 and how many were over 75? Just a number. I don't need a slide.

DR. BELDER: I'll show you the slide because I don't know it by heart. Above 65 you see the numbers here.

DR. BORER: And above 75?

DR. BELDER: I don't know. I believe none.

DR. BORER: None?

DR. BELDER: None.

DR. BORER: And we have a statement here that says there is no need for lower doses in the elderly. How many additional drugs were these patients over 65 and the 0 over 75 taking? How many other drugs were they taking?

DR. BELDER: I don't know it by heart.

DR. BORER: Well, I think we ought to know. And what were those drugs? Do we know that? What pathway of metabolism did those drugs use? Which ones interfered with the CYP 450 system?

DR. FIEDOREK: Well, Rene, you might comment about the PROSPER study, which we don't have finished.

DR. BELDER: We have currently in a study ongoing -- actually a study we'll have last patient visits in April. In 5,800 patients, on the age --

DR. BORER: But you have data now?

DR. BELDER: Let me answer one of the questions that you raised, is the CYP 3A4 interaction. Pravastatin is not metabolized by CYP 3A4, and therefore there's no potential for interactions with inhibitors of 3A4. pravastatin, with respect to drug-drug interaction pravastatin is extremely clean. In that sense our current label has a statement about the use of pravastatin in the elderly, indicating that pravastatin is safe in the elderly population.

DR. BORER: Okay. So, the statement here is that we have sufficient data so that we know there will be no drug-drug interaction, not only to alter the pravastatin level, but to alter the level of other drugs that could be concomitantly taken in the elderly. We know that.

DR. BELDER: And that was part of the original application with pravastatin, to make sure that pravastatin would not alter drugs like digoxin, warfarin.

DR. BORER: Right. And therefore, there's no need to be able to titrate the dose of pravastatin in these people.

DR. BELDER: In elderly, no.

DR. BORER: Is that a statement that the FDA is in concordance with, can I ask?

DR. LIPICKY: I do not know. I cannot answer that.

DR. BORER: Anybody here from metabolic and endocrine?

DR. KREISBERG: It's my understanding that as the drug is approved for utilization, there is no specific statement that titration is unnecessary.

DR. BORER: Unnecessary.

DR. KREISBERG: That it is unnecessary. I believe that the data that has been presented is impressive data that deals with a fixed dose, but it does not address the issue that was raised by my colleague down the table here about how this fits in with the NCEP adult treatment 3 guidelines, and whether it avoids or perpetuates the idea that the goals proposed by them are unnecessary.

DR. BORER: Yes. The efficacy issue is a very important one. I'm concerned with the relation of safety and efficacy here. Bob maybe can --

DR. TEMPLE: Well, I don't think anybody could say there's never a reason to use a different dose. I doubt the company would say that, and they've asked for and gotten approval of an 80 milligram dose, so obviously there are other doses that are useful.

Fixed combinations of this kind may very well say -- that all depends on what you all think -- that the fixed combination is appropriate only for people who need those relevant doses.

Now, one of the concerns that I guess you'll hear Ray talk about is that we don't want to have the convenience of the formulation constrain people unduly. So, as you see, there are two doses of aspirin because we don't want the existence of the combination -- and we talked to the company about this -- to mean everybody has to get 80 or everybody has to get 325, when both doses are currently recommended in labeling for aspirin.

And that's a fair question to ask about the prava dose. If the enormous majority of people need 40 then you might think that's reasonable. If that really keeps you from meeting some appropriate guideline because you can't go high enough, then you might consider that desirable, or you might handle that by saying the whole idea's a bad idea, or by putting something in labeling that says something. Those are all perfectly good things to think about.

But one of the principles that we've enunciated is that you shouldn't force people to use the wrong dose by having a combination. And for any hypertensive combinations, for example, we try to assure that there are dosage forms that have appropriate levels of each of the components. Not everyone necessarily, but a pretty good range.

DR. BORER: Alan?

DR. HIRSCH: Let me follow u, Bob, on your ideas a little bit. I'm going to ignore achieving guideline goals that I'm sure we'll get to later, but I just want to take a moment and stay on the safety issues. I think when we package things together, we're assuming obviously the patient should take them in that combination.

So far I think we were presented in slide B-4 with the pharmacokinetic crossover study, which looks very clean. But let me just tease this a little bit further for fun and interest.

We see no change in Cmax or area under the curve for these doses in the small study. The question, I guess, is, do we have any evidence in any way that prava affects aspirin's effect on platelets? In other words, I might hypothesize doing an aggregation study, and again demonstrating either with blood from the patient or in vitro that there is no effect on the platelet wall. Any thoughts? Platelet activation.

DR. BELDER: Well, that's a hypothetical possibility, and we think that is very unlikely. In addition, in the analysis that we did, we see a treatment effect of aspirin. Dr. Berry will, of course, go into further detail on that. That is very similar to the treatment effect of what one would have expected. So, in that respect we don't think that there is any diminished effect of aspirin.

With respect to the possibility of a potentiated effect of aspirin, we are fairly encouraged by the safety signals that we see. Perhaps we can show the slide with the hemorrhagic strokes. This is the fatal and nonfatal ischemic and hemorrhagic strokes. I haven't put them on a slide to put them in perspective with respect to how many hemorrhagic strokes we saw and how many ischemic strokes we saw. But clearly in this part of the panel, there's no evidence that the combination would lead to an increased bleeding. We have a similar picture for gastrointestinal bleeds.

You may think the fatal events look differently, but -- I think we have the next slide, fatal events. This is for fatal ischemic and hemorrhagic strokes, and again, you don't see any evidence of a signal here.

DR. HIRSCH: No, I agree. I've never seen, in the data sets you've given, that evidence of clinical signal, but I was looking for mechanistic interactions.

Let me take that another way as well, in vitro. We're obviously implying with this that 40 milligrams is the dose that should be used, but patients obviously don't comply with our recommendations. Sometimes they take too little, sometimes they take too much.

So, in these pharmacokinetic studies, again, do we have a dose response? If patients did take 80, or if we administered greater amounts of pravastatin, can we achieve an interaction with differing doses? In other words, how far have you tested the interaction between the two in a dose-response manner?

DR. BELDER: From a pharmacokinetic perspective?

DR. HIRSCH: Kinetic, and then --

DR. BELDER: We did it with a single dose. At the point that we did the study, 40 milligrams was the highest approved dose. We have not done a pharmacokinetic interaction study with the 80 milligram dose. However, based on the pharmacokinetic profile of pravastatin and aspirin -- they're both very short-lived -- one would not expect that at the 80 milligram dose the results would be different.

DR. BORER: One final question before you move on. This is really for Dr. Fiedorek, I guess. What data set were you referring to when you said that patients commonly take Tylenol rather than aspirin with a statin?

DR. FIEDOREK: Yes, I was actually foreshadowing to the fourth talk. Dr. Pearson will talk about that data. It's not in any data in the pravastatin data set. It's a publication on consumer use. Dr. Pearson can answer.

DR. BORER: Are we going to see numbers about that?

DR. FIEDOREK: Actually I'll refer to Dr. Hennekens, who actually did the study, even though Dr. Pearson is going to talk about it. I'll let Dr. Hennekens answer.

DR. HENNEKENS: Working with Nancy Cook at the Brigham and Women's Hospital, we had the opportunity to review a large national sample of people who had been prescribed aspirin for secondary prevention. In that data set that Dr. Pearson will speak about in detail later, fully 15 percent of people who were told that they should be taking aspirin by their health care provider were mis-medicated. They were mis-medicated either with acetaminophen or with nonsteroidal anti-inflammatory drugs. The other point in that survey is only 51 percent of the people who really should have been taking aspirin were taking it. So, there was both under-utilization of aspirin and mis-medication with aspirin in the very population for which this indication is being sought.

DR. BORER: Charlie, do you know how many of these people had statins prescribed concomitantly?

DR. HENNEKENS: No, but I can tell you -- I don't want to be stealing Dr. Pearson's thunder here. I think a major point is in recent databases suggesting maybe that 77 percent of people are really taking aspirin in secondary prevention who should be getting it, and only 37 percent are getting statins. So, if a combination product did nothing more than achieve that 77 percent of people who were on aspirin who needed the statin were also on the statin, narrowing that treatment gap from 37 percent to 77 percent, that translates to probably over 5,000 premature deaths prevented each year in the United States alone.

DR. BORER: Okay. Why don't we move along to Dr. Berry.

DR. BERRY: Thank you. Good morning, ladies and gentlemen. I'm a statistician and I work with cancer. I'm especially interested in and passionate about breast cancer, but I work on other diseases as well.

I'm interested in Bayesian statistics. The Bayesian approach is particularly appropriate for synthesis of information in the sense Bayesian analysis is meta-analysis. However, I will be presenting standard frequentist multivariate analyses and expanding the assumptions, dropping assumptions, expanding the model to consider Bayesian analyses as well.

Dr. Belder has addressed the question of pravastatin on top of aspirin, a randomized comparison. I'll address that comparison in the context of all five secondary prevention studies, and I'll also address the issue of aspirin use among those assigned to pravastatin, and finally I'll address the question of the persistence of the effect over time.

The possibilities. Pravastatin was randomized with placebo in all the trials we'll be talking about. Aspirin use and non-use was also measured, and so we have four categories. The combination. We'll be comparing the combination with placebo, the randomized comparison that Dr. Belder talked about. We'll also be comparing the combination with pravastatin alone, the observational comparison.

Placebo seems left out of this, and indeed, in most of the comparisons we'll be talking about the combination on top of a single agent, but at least once in the presentation I'll compare back to placebo. It's an important benchmark.

The question is, is the combination more effective than pravastatin alone. We have LIPID and CARE. The event rates in LIPID and CARE suggest that indeed that's the case, and you see that here. Both of these are observational comparisons. This is with respect to the primary endpoints in LIPID, which was coronary death, and in CARE, coronary death or nonfatal MI. The rates here are greater, but the effect of aspirin, the reduction among those using aspirin is about 35 percent in both of these studies.

Now, you're worried, of course, that the patients who took aspirin had different characteristics from those who didn't take aspirin. Perhaps they had better prognoses, perhaps they had worse prognoses. An approach to take into account the possibility that aspirin use was differentially applied in these studies, that patients took aspirin for a reason associated with the extent of their disease is to adjust for the various covariates, the patient characteristics.

You see here we adjust in the multivariate models for age, gender, previous MI, smoking, baseline lipids, baseline blood pressure. So, every analysis that I do and every comparison that I do will be taking these into consideration.

There are other variables that might affect aspirin use. For example, you'll notice if you have looked at the submission that among patients taking aspirin as opposed to not, those taking aspirin had a slightly higher incidence of revascularization procedures. So, that suggests that we take into account other things that might be used in assigning aspirin. Revascularization, diabetes, obesity, these variables we had in the two principal studies, in LIPID and CARE. We did not have them in the other three, the smallest studies. We've done separate analyses addressing specifically these, and also the use of ACE inhibitors, and I can tell you about that if you're interested.

The bottom line is that qualitatively there's no difference in the conclusion within LIPID and CARE considering these variables in addition to these as opposed to just these. So, we can talk about that if you'd like, but the rest of my presentation this morning will be focusing on those.

Now, no multivariate analysis can turn an observational comparison into a randomized comparison. However, if we look at subsets and we see the same thing from one subset to the next, which is in fact what we do see and you saw an example of that with the breakdown by gender, then that gives more confidence that in fact the result is real.

These are the five studies. Dr. Belder has talked about LIPID and CARE. LIPID and CARE consist of approximately 90 percent of the population and you see that here, 13,000 or so from the 14,500, the total being 14,600. The percent of aspirin use varied, approximately 83-84 percent, as Dr. Belder indicated, in LIPID and CARE, but somewhat less in the other studies varying down to 43 percent in PLAC II. Overall, about 80 percent of the patients were taking aspirin at baseline.

Now, in two of the models that I'll be talking about, we worry about the possibility that the trials are heterogeneous, that there are different characteristics of these trials somehow, even if we adjust for the covariates, that there is an additional trial effect that could affect the conclusions. So, we're going to allow for the possibility of heterogeneity.

However, the trials, the five trials, had lots of commonalities, and these are listed here: similar entry criteria, similar types of patients, of course a randomization of pravastatin versus placebo, long-term follow up, endpoints. We'll consider particular endpoints or others that you may be interested in and we can show you. These endpoints were all measured in the trials, the covariates recorded. The data analysis for each of these trials was conducted independently of the sponsor, separate from the sponsor. However, the sponsor has combined the data into a single data set with all of the variables in question to facilitate the meta-analysis.

These are the endpoints we're considering, three: fatal and nonfatal MI, ischemic stroke, and then a composite including these, but also including any coronary death and the vascularization procedures.

The first model that I want to talk about is the standard one, the one that is familiar to most of you, I suspect. It is a multivariate Cox proportional hazards model, which will include all of the covariates that I talked about before. The patients are combined across the trials. We're considering the single data set, but we also consider trial as an effect, so trial is one of the covariates that we are adjusting for in the model.

This is for fatal or nonfatal MI. This, the yellow comparison is the one that Dr. Belder talked about. It is the randomized comparison of pravastatin on top of aspirin. So, this is restricted to the patients taking aspirin. What is the benefit of adding prava? And you see that it is a 31 percent for fatal or nonfatal MIs, a 31 percent reduction.

This is the observational comparison. Among those patients who were randomized to pravastatin, 80 percent of them were taking aspirin. The benefit of aspirin amongst these patients was about 26 percent. This is the value 1. The fact that the confidence interval does not include 1 means that it is statistically significant in this multivariate analysis.

The next endpoint is ischemic stroke. The confidence intervals are wider because there are fewer events in ischemic stroke. Again, this is prava on top of aspirin, a 29 percent reduction. This is aspirin on top of prava, a 31 percent reduction. And again, statistically significant.

The composite endpoint, of course more events, smaller confidence intervals, the reduction due to pravastatin on top of aspirin, 24 percent; 13 percent aspirin on top of pravastatin. And again, statistically significant.

Now one of the questions of interest to the FDA is, is this one study? Is it two studies? And to address that, we've broken out into LIPID and CARE separately. So, the analyses that you've seen on the previous slide, I'll repeat on the next two slides. This is the randomization. See, all yellow? This is the randomization comparison, the benefit of pravastatin on top of aspirin for LIPID and CARE, LIPID and CARE, LIPID and CARE for the three endpoints that we're talking about. This is the number 1, so statistical significance if it overlaps the number 1 for these studies separately.

So, for example, you see in LIPID about a 24 percent reduction in the composite events for pravastatin on top of aspirin, about a 24 percent reduction, the same for pravastatin. This is pravastatin on top of aspirin in CARE and in LIPID.

The observational comparisons in blue, and the composite endpoint of 14 percent reduction of aspirin on top of pravastatin in LIPID, a 22 percent reduction for aspirin on top of pravastatin in CARE. And again, both statistically significant.

This is the second model I want to consider and it is an extension in the following way. It's a Bayesian hierarchical model. It allows for the possibility of heterogeneity in the studies, in the various trials. It treats really two experimental units. This is a hierarchical model. There are two levels of experimental unit. One is patient within trial, but trial itself is an experimental unit. There is more information in a trial with larger sample size, but the trial is counted as much as any other trial of the same size.

Now I want to show you the comparisons here. This is the cumulative proportion of events -- this is for fatal or nonfatal MI -- out to 5 years for the randomized comparison of the combination versus aspirin alone. So, this is prava adding to those patients taking aspirin. This is the 31 percent reduction out here at year 5. It's easiest to see the 31 percent reduction in event rates, as well as in hazard.

The other randomized comparison is for prava for non-aspirin users, prava versus placebo. And here the reduction -- actually we haven't shown you that -- is about 20 percent.

Any comparison of a dotted line with a solid line is an observational comparison because it compares aspirin versus not. I said I'd mention placebo. The effect of aspirin alone is a reduction here of this extent. The effect of prava alone is a reduction of this extent. If you add those two together, you get something, I don't know, about down here. What we're looking at in the combination is something that is at least additive.

This is for ischemic stroke. Again the randomized comparison of pravastatin on top of aspirin, and this was I think a 29 percent reduction. This is the randomized comparison for the non-aspirin users, and the benefit here, I think it was like a 29 percent reduction in risk for patients in comparison of aspirin alone versus those who were taking pravastatin plus aspirin.

And the composite endpoints, I think similar. This was like a 24 percent reduction, and this is like the 14 percent reduction that we saw a couple of slides ago.

So, the same thing is happening in model 2 as model 1. The analyses that we did in model 2, allowing for this study heterogeneity, reinforced the comparisons in model 1. So, the combination provides an benefit for all three endpoints, the benefits ranging from 24 percent to 34 percent comparing the combination to aspirin, and 13 percent to 31 percent comparing the combination to pravastatin. The benefit was similar in models 1 and 2. And this benefit was consistent within the studies, LIPID and CARE, considered individually.

Now, a possibility that you might worry about -- we're doing proportional hazards. And so these are cumulative proportion of events for model 2, very similar for model 1, and you see that these lines don't cross. Roughly speaking the hazards are the derivatives of the slopes of these lines. These are the hazards by year for the first year, the second year, up to the fifth year. And you see that these things are proportional for each of the treatment groups. That is one of the assumptions of model 2 as well as model 1. You see a drop in hazard. In the first year, all these have a higher hazard. Presumably the mixture of patients is heterogeneous and the patients are at high risk, at least some of the patients are at high risk, in the first year. And they recur. When we go to the second year, the hazard is calculated by redefining the denominator so that we're looking only at at-risk patients. The hazards drop and presumably start to increase with the force of mortality. People are getting older.

And so we introduce model 2. And one of the concerns that you might have is, well maybe one of these agents, say aspirin, works early on and then doesn't work anymore. And pravastatin works late on and doesn't work early on. So, maybe you can take aspirin first, and then after a few years convert to pravastatin. And so far, we've not worried about that possibility. I want to worry about that possibility. We want to extend model 2, all of the multivariate modeling aspects of model 2, to allow for the hazard ratios within treatment to vary over time.

This is the cumulative proportion of events from model 3. These are estimates. I can tell you what the probabilities are for comparing these curves at any of these time points if you are interested.

These are the hazards. The hazards, now you see there's a great deal more noise because we are modeling these things individually. We're modeling the hazard in year 1, separate from year 2, separate from year 3. So, there's a good deal more variability and crossing here of some of the hazard functions. For example, it happens in year 3 that the hazard for aspirin alone is actually slightly greater than placebo alone. You expect that sort of thing because there's a good deal of noise here.

There are several amazing things about this picture. One is that the combination is better in each one of these years. The combination is better than any one of the other treatment groups in every year. These are like five separate studies. The events in this group in the first year are distinct from the second year or distinct from the third year, etc. So, we sort of start over again. And when we start over again in the second year, again the combination wins.

Now, I can quantify that for you if you like, I can tell you what the probability is that in this particular year the hazard is better for the combination than, let's say, for aspirin alone. But the important thing to me is that the hazard is better for the combination group in each one of these years. It shows the persistence of the effect.

Another interesting thing about this picture is it shows what doesn't happen. I mean, one of the things that you see in the first year is that the combination lowers the hazard. What does it do? Does it extend the period of time before the event occurs? So, does it push it into the future a year or two? If that were the case then you would expect this bump coming later. That doesn't happen here.

So, the conclusion of the hazard analysis over time, the benefit of the combination over aspirin was present in each year of the 5-year duration of the trials and the same is true for the combination over pravastatin. The benefits estimated for model 1, the confidence intervals in particular, were confirmed by the more general models and fewer assumptions. When we dropped the assumption of proportional hazards, for example, we observed the same thing.

So, we've observed benefits in the meta-analysis. We've observed the same benefits within the studies considered separately. We allowed for heterogeneity in a number of ways, but in fact these studies are quite homogeneous with respect not only to the baseline characteristics but also the results.

And so now I'd like to turn the podium over to Dr. Tom Pearson who will discuss medical need.

DR. BORER: Are there any questions for Dr. Berry? Ray.

DR. LIPICKY: I guess I missed it when I first looked at the thing, but you actually think the analyses suggest that there's a super-additive effect or a synergistic effect between prava and aspirin and that you could --

DR. BERRY: Dr. Lipicky, between you and me, the answer is yes. I think there is a super-additivity.

DR. BORER: I'd like to extend the question I asked earlier. You had 1,600 people who were over age 65. 3 percent of your total population had liver enzymes that were at least three times the upper limit of normal or CK at least four times greater than the pretherapy level. How many in the above age 65 group had these abnormalities? Do you have that breakdown?

DR. BERRY: One thing. Not in direct answer to that question, but we have done a separate analysis of the over 65 with respect to what I've shown, if you're interested in seeing that. You don't care about that.

DR. BORER: I'm not because I believe you. And I don't disbelieve anything I've heard. You know, we're talking about a single dose to be mandated as part of a combination that could conceivably alter practice patterns, and I want to know about the safety of doing that relative to the effectiveness which we're going to hear more about. Bev has already raised that issue.

DR. BELDER: Perhaps I can tell you what currently the pravastatin label states about geriatric use. It says the following. Two secondary prevention trials with pravastatin, CARE and LIPID, included a total of 6,593 subjects treated with pravastatin 40 milligrams for periods ranging up to 6 years. Across these studies, 31 percent of pravastatin subjects were age 65 or older and .8 were age 75 and older. The beneficial effects of pravastatin in elderly subjects in reducing cardiovascular events and in mollifying lipid profiles was similar to that seen in younger subjects. The adverse event profile in the elderly was similar to that in the overall population. Other reported clinical experience has not identified differences in responses to pravastatin between elderly and younger patients.

DR. BORER: Okay. Do you have the numbers I asked for, or not?

DR. BELDER: So, in the two trials there was we didn't do an analysis of CK by age, no.

DR. BORER: Okay. Or liver enzymes. No.

DR. BELDER: Well, with respect to liver enzymes, the current pravastatin label does not require liver enzymes to be measured after initiation of therapy, and that applies to all ages.

DR. TONKIN: Perhaps if I could make some comments about the safety database and also about the issue around age. LIPID contributed 68 percent of the data that you're seeing. In fact, at baseline there were 1,511 patients age 70 or over. They were followed for a mean of 6 years, and then in fact after that, we approached all patients, including those who had been randomized to placebo, to see whether they would be agreeable to go on to open-label pravastatin, specifically to get more data about safety, including the elderly, more data about cost effectiveness. In fact, we have 95 percent of the initial cohort who had survived who hadn't died who agreed to that further follow-up. So, the safety danger in LIPID now goes out to where patients may be 83 or so. We see no signals.

But the important point, I think, is that what we did in LIPID was we said, what is the effect of pravastatin in a dose of 40 milligrams against placebo against the background of usual therapy. So, the individual clinicians had to make the decision about whether or not patients should be on aspirin. The trial didn't mandate it. We left that decision to the clinician. So, undoubtedly, a number of people who would not be treated with aspirin are not getting into the data set.

With respect to the overall data set with pravastatin, if one includes also the West of Scotland study with LIPID and CARE, there is 112,000 person-years of experience comparing pravastatin, a dose of 40 milligrams, against placebo. In fact, there are many patients who remained on pravastatin as remained on placebo at the end of the study. Extraordinary tolerance. There was not a single case of rhabdomyolysis in that 112,000 patient-years of experience.

If you took those patients who had abnormal liver function tests at baseline, there was no difference between placebo and pravastatin on top of that in terms of deterioration.

So, I think the experience with respect to safety is extraordinary. What really this is about is ensuring the patients would receive the dose that is proven in the studies, that against what would be the position of judgement, if you like, about usage of aspirin.

DR. BORER: Bev, did you have a comment? Or Susanna?

DR. LORELL: I guess one of the comments in the geriatric use paragraph that was read, in the next paragraph there actually is a comment that mean AUCs were slightly higher in elderly subjects.

DR. BELDER: That's correct. There is quite some variability in the AUC levels of pravastatin. However, as Dr. Temple indicated before, we have recently gained approval of pravastatin at 80 milligrams. In addition, we are collecting quite a substantial safety database on pravastatin of 1,260 milligrams, and so far we do not observe any safety signal with pravastatin.

Again, I don't think that the safety of pravastatin in whatever population is an issue. Pravastatin has been proven to be extremely safe in a variety of patient populations.

DR. BORER: Okay. Tom.

DR. FLEMING: Don, I had a couple questions. I appreciate and thank you for the very nice presentation of these three models. Certainly they are very informative.

As you note, the major challenge here is really trying to understand what aspirin adds to pravastatin in the absence of randomized trials. These models make an attempt to make adjustments for the imbalances that may exist between those who elect to use aspirin versus those who don't.

You have adjusted for a number of factors and I think you've really acknowledged this. What concerns most of us about observational data and analyses and models such as this is that they are informative and helpful, but we worry about whether we're adjusting for differences that are the tip of the iceberg.

You have noted that demographics -- smoking, revascularization procedures -- were important elements to adjust for. I understand you have adjusted for all of those in the model.

Some of the other things that we might think about are, for example, differences in other interventions, baseline treatments. We see, for example, in the FDA briefing document on pages 37 and 38, we see differences in beta blockers that are more frequently being used in those who are choosing to use aspirin and those who are not.

How have you addressed the potential impact of differences in concomitant meds between those electing to use aspirin and those not?

DR. BERRY: We did an analysis within LIPID and CARE separately for ACE inhibitors. I can't remember. Did we also do beta-blockers? Can you bring those slides?

All of these are within model 1, that is, the standard proportional hazards model. Model A is what we talked about. Everything that you've seen is model A. Model B includes these other issues of diabetes, the revascularization procedures, BMI, obesity, stroke, dyspnea, angina. Model C includes the same as model B, but also beta-blockers and ACE inhibitors.

And this is CARE. The variables in CARE are slightly different, as you see here. We didn't have some of the same variables. There are additional variables in LIPID as opposed to CARE. And so I'll show you these things separately. These are separate models and it's awfully busy. Let's see if we can focus on, say, the composite endpoint.

This is the composite endpoint and we are now talking about LIPID. And so in LIPID this is what you saw before. This is somewhat different now because it doesn't include all of the other studies. This is just LIPID separately. There was a 24 percent reduction in pravastatin on top of aspirin and a 14 percent reduction in aspirin on top of pravastatin. That was model A. If we incorporate the second tier of variables, we get something which is comparable. If we go to model C, which also includes the beta-blockers and the ACE inhibitors, we see something that is very similar.

DR. FLEMING: While we're here then, essentially model C is the direct answer to this specific question.

DR. BERRY: Right.

DR. FLEMING: But also let's look at CARE.

DR. BERRY: CARE, in fact, gets even stronger. The conclusion is even stronger.

DR. FLEMING: So, they go in a bit the opposite direction?

DR. BERRY: Right.

DR. FLEMING: With CARE, adjusting for beta-blockers and ACE inhibitors, there seems to be an enhanced effect. With LIPID, there seems to be a somewhat diminished affect.

DR. BERRY: Only slightly diminished. If you go back, you'll see that it's not changed. It is slightly diminished.

DR. FLEMING: 14 to 11 to 12 to 9.

DR. BERRY: Slightly diminished, 35 to 30.

DR. FLEMING: And this is using as covariates beta-blockers and ACE inhibitors as reported at baseline.

DR. BERRY: That is correct.

DR. FLEMING: Second question. If we look at the raw data, the conclusions, Don, that these analyses have presented, not too surprisingly, are fairly consistent with an impression you get when you just look at the raw data. One place that that is presented is in the FDA briefing document on pages 41 to 43 for each of the five major endpoints that were considered. And it's really worth perusing that data for a moment on pages 41 to 43 because it really shows an intriguing pattern.

What it does is it breaks the data out into groups by pravastatin plus aspirin, pravastatin alone, aspirin alone, and neither, which ideally we would have liked to have had in a true factorial design. Of course, what we know is that this is based on pravastatin to placebo randomization where aspirin use is observational.

As you scan through these three pages and you're looking at each of these endpoints, what you find, which is somewhat similar, Don, to your comment a bit early about there maybe being a positive synergy here, is for each of these five endpoints, you find that when you add aspirin to pravastatin, you get a much more vigorous or substantial improvement and outcome than when you're adding aspirin to control.

In a sense, that's reassuring because the really relevant question here is, what does aspirin add to pravastatin, not what does aspirin add to nothing. And yet, what we're dealing with here, as you've acknowledged, is we're out on the end of a limb here because we're really trying to determine what the effect of aspirin is in nonrandomized data. Where we do have randomized data is looking at the effect of aspirin alone.

So, what concerns me is, when I look at these five endpoints on pages 41 to 43, when I'm making the comparison from aspirin against nothing, I'm seeing essentially no effect on any of these five endpoints. Not only is it less effect than what aspirin does in the presence of pravastatin, but what aspirin does in the absence of pravastatin in these data is essentially nothing.

What concerns me is that's not consistent with what we've seen from randomized trials looking at aspirin. We do have evidence about what aspirin does in randomized trials, but it's in the absence of pravastatin. So, now that we're in this realm and we're using these data out on the end of this limb to say what aspirin does in the presence of pravastatin, and we look also at what these data are saying about what aspirin does in the absence of pravastatin, and we see an answer that's inconsistent with the randomized trials, how do we reconcile this? In your exploration of these data, can you tell us why aspirin doesn't add anything in the absence of pravastatin?

DR. BERRY: Yes. First of all, the group that you're looking at is the smallest group. It's the set of patients who were not taking -- let me start over again.

DR. FLEMING: It actually is half of the group.

DR. BERRY: Yes, it's half of the group. That's why I'm starting over again.

The effect of aspirin. If you looked at this study and said -- I think we have a slide on this -- let's look at aspirin alone, 80 percent versus 20 percent, what is the effect of aspirin? The effect of aspirin alone is the mixture of, or the average of the effect of aspirin for those patients who were taking pravastatin plus the effect of aspirin for those patients who were not taking pravastatin.

And so you correctly say that the benefit -- let's think about the composite endpoint where the comparison that you're making is most pronounced, and actually why don't we show that slide, the one where the composite endpoints, model 3. It's one of the late ones, like C-20 or so.

So, this is what Dr. Fleming is talking about. If you compare placebo and aspirin alone, there's very little difference. In fact, I think it was like 3 percent reduction due to aspirin. If you compare, however, the pravastatin, the affect of aspirin here, it's -- I don't know -- 13 percent or so. And so if you ask the question, what is the overall benefit of aspirin in this study, it's about a 10 or 11 percent reduction, 13 percent average with 3 percent, but with the greater weight on the other one. As to why this is not different, I'd give it to small sample size.

Let me say one other thing about that. In terms of the composite endpoint, the composite endpoint includes the revascularization procedures. And we have a slide, which we can show for the various endpoints, which indicates that in fact aspirin has no benefit on revascularization procedures. In fact, if you took these out, there would be a separation here.

DR. HENNEKENS: Can I make another point, Don?

DR. FLEMING: Oh, Charlie.

DR. HENNEKENS: I just want to make another point on Tom's question because this issue was troubling to me when I first looked at these data as well.

My own looking at it is as follows. If we look at the randomized comparisons in the Physicians Health Study, the time course to benefit, we began to see that over 40 percent benefit within 6 months of taking the aspirin. Then it persisted over the 5 years until the trial was stopped because of the statistical extreme nature of that finding, with more endpoints developing.

I think it's important to point out that in the CARE study, the time of randomization was 10 months after the event and the time to randomization in LIPID was 13 months after the event. So, I think one of the issues to consider is that the major benefits that aspirin conferred may have occurred already before these trials began.

DR. FLEMING: Charlie, are you suggesting then that this might be true here, that after you've been on a certain period of time, continued use of aspirin is not --

DR. HENNEKENS: No, I think that these data -- I think Rene was going to show that these data also show benefits among the aspirin users compared with the non-users. However, the ability to study this, I guess you'd call it, interaction would be best, as you point out, in a randomized, double-blind factorial trial where everyone is assigned to the agents at the same time. And here we have a disconnect because we have, in my view, predominantly anti-atherogenic effects of the statin drug that takes some delay until it occurs, and the predominantly antithrombotic effects of the aspirin, and the time course of that large benefit is within the first several months of starting it, which would be at the time these people were started, I think. It's just a methodologic point I wanted to add to the discussion.

DR. FLEMING: Don, while you're speaking, could you put that slide back on again that you just had?

DR. BERRY: Put it on again.

A bottom line that you can read from this is that the only way to get a benefit from aspirin is to take pravastatin with it.

DR. FLEMING: I like your color coding and your interpretation before.

Basically, as I look at this, where at least I feel most comfortable, I have to admit, is where I have randomized comparative trials.

DR. BERRY: Sure, of course.

DR. FLEMING: And as you note, when we're comparing these solid lines, and in particular the solid orange against the solid green, it's answering an important question and doing so in the context of a randomized trial. What does pravastatin add to aspirin?

The other question that I find very interesting is the dotted purple against the green, which is, what does aspirin add, the dotted purple there, against the green. Which is what does aspirin add to pravastatin?

What's encouraging, as you noted, is that that seems to be greater than what the orange does against the dotted red. If anything there is synergy here. And where my discomforts is I know something about the orange against the dotted red from sources that are much better than this, from randomized trials, and they don't agree with this. So, I'm just left with a sense that when I'm seeing something that I do know about that doesn't agree with this, then where I'm trying to use this, which is the dotted purple against the green, it just makes me a little uneasy.

The good news is, though, that this is underestimating what the effect of aspirin is. So, if I extrapolate that, then one might be willing to say that the green versus the dotted purple is underestimating, that's one positive way to look at it.

The negative way to look at it is, when I have randomized trials and I have historical evidence or observational evidence and they don't agree, then it makes me more worried about being out on the end of that limb when I'm having to use observational data for the green against the purple.

DR. BELDER: Don, we did a couple of other looks at the data, and perhaps a slide up.

This is all aspirin users and this is what Don already alluded to, that basically the effects on the lines that you see is an average of the aspirin users in pravastatin-treated patients and non. And here you see the effects on the various endpoints of all aspirin users versus non-users, and the treatment effects are actually quite consistent except for the composite endpoint that includes CABG and PTCA. All the other endpoints are very consistent with what you would have expected aspirin to do in this population.

Now, the question may be, well, why doesn't it show up in the previous slide that we had? And one has to realize that we had the non-aspirin users, who were a minority of the population, about 20 percent of the population. In addition, those patients who were not on aspirin at baseline, many of them started using aspirin as the trials went on. So, particularly in that group we see slowly a treatment effect of aspirin starting to occur. But these data, we believe, present the true effectiveness of aspirin in this population. That's the mix of the pravastatin and placebo users.

DR. FLEMING: Don, I had one more question. I don't know if this is getting at my answer or not. You had given the analyses on three of the major endpoints. The primary endpoint of LIPID was CHD death, and the primary of CARE was CHD death, nonfatal MI. Did you also do your analyses for those endpoints?

DR. BERRY: Yes.

DR. FLEMING: Can you just quickly show us?

DR. BERRY: Can we show those? CHD death, CHD death including nonfatal MI.

DR. BORER: Just a yes or no answer while you're waiting for that. Do you want to put a statement in the label of this combined product that says it shouldn't be used by people who've had a revascularization procedure?

DR. BERRY: No. You're going to answer this, I know, Rene. But I want to distinguish between revascularization procedures at baseline and what we're talking about here. This is an endpoint revascularization procedure. It is not a baseline.

DR. LORELL: A comment on that point. I think that there's even another possibility to interpret that data, if just for a minute we could go back to that slide that broke out bypass surgery and angioplasty. If you're doing an intervention of using aspirin and lipid-lowering therapy that dramatically reduces the risk of acute coronary syndrome, then by definition, in a large or small population, you are going to be doing many fewer interventions for that indication. And you have not shown us that data but I think it would be highly likely that that population of CABG and angioplasty events are enriched by a group for whom the indication was chronic stable angina and it was unenriched by loss of the population of people who had acute coronary syndromes.

In fact, I think -- and maybe Charlie Hennekens can correct me if I'm wrong -- but I don't think there is data that demonstrates that aspirin use alone prevents that piece of the indication for revascularization. In other words, this may be actually a confounding effect on actually changing the kind of pool of people compared to the trial that Dr. Hennekens was discussing.

DR. BELDER: We have thought about this as well, and actually we determined the endpoints that we were looking at before we actually saw the results of the meta-analysis. In retrospect, if we would define the endpoints again, we probably would take out the revascularization from the endpoint because it is clear that we do not pick up a treatment effect of aspirin. And it's clear that pravastatin has a treatment effect for these events. I think that what you said is a very plausible explanation.

DR. BORER: We have Alan, Bob.

DR. BERRY: Do we have those slides yet for the CHD death?

Okay. So, this is CHD death and nonfatal MI. I guess we don't have it, Tom, combined. Oh, second from the bottom, okay. These are broken out and so if you were to combine these, it would show something similar. If anything, it's a better comparison than including these procedures.

DR. BORER: Bob?

DR. TEMPLE: Well, the only observation I wanted to make is that the effect of aspirin in controlled trials is not perfectly consistent either. What we believe comes mostly from meta-analyses, as everybody probably remembers, the largest secondary prevention trial went the wrong way on survival and was almost dead even on most other things. The results in the Physicians Health Study are completely unmatched by what I consider a fairly similar trial in primary prevention. So, with a small data set, it's not entirely surprising that you might or might not see something in one of the components.

DR. BORER: Paul?

DR. FLEMING: Before we leave that point, that could be true. One could attribute this to the smallness of the data set. I'm looking at two sources of information. One is this data set here, which is 7,200 people, and then the 20,000 people that were reviewed in the FDA briefing document from the randomized trials.

DR. TEMPLE: Not in the no-aspirin group. I mean, most of the people here got aspirin. So, the comparison with the no-aspirin group is pretty small.

DR. FLEMING: It's 5,800 versus 1,500, right, in this study.

So, I'm saying that is a possible explanation, but across all five endpoints there is, I think, a very discernible difference in terms of lack of effect on any of those five, compared to some of these endpoints that when you look at it in randomized trials, certainly you show considerable effect.

DR. TEMPLE: Looking at the effect of aspirin alone, is what you're noticing.

DR. FLEMING: Correct. Aspirin alone.

DR. BORER: Paul?

DR. THOMPSON: Dr. Berry, could you address the possibility that these studies, done by very knowledgeable, sophisticated investigators, that the representation or the finding that both the lack of benefit and the super-benefit of aspirin is actually due to the fact that these doctors are making good decisions about who they put people on, and they're deciding not to put either frail or people with GI bleeding or other conditions on aspirin, and that that actually could be a possible explanation for both the high and the low, the over-estimation and the under-estimation?

Really the best utility of these data is to show that something that is a recommended treatment, which is aspirin and a statin, in patients with coronary artery disease doesn't appear to do a whole lot of harm.

DR. BERRY: Can I simply agree?

(Laughter.)

DR. BORER: Yes.

DR. BERRY: I agree. With respect to the frail, we did not have a measurement of frailty per se, but it might be reflected in some of the other covariates that we did measure.

DR. FLEMING: If I could just pursue that, if that's what one were thinking, and if I viewed these four subgroupings as real, then what I would say is the doctor should be saying, if I'm on pravastatin, certainly put me on aspirin. If I'm not on pravastatin, don't put me on aspirin. And yet, in exactly the same proportion of cases they chose to put you on aspirin, whether or not you're on pravastatin.

DR. BERRY: Of course they didn't know whether --

DR. FLEMING: They didn't know. What was striking to me is in these data, when you've randomized to pravastatin versus control in LIPID and CARE, and it was choice as to whether to use aspirin, the same fraction of people chose to add aspirin whether or not you were on pravastatin or control.

DR. BORER: Perhaps we can move on to the --

DR. TEMPLE: Jeffrey?

DR. BORER: Oh, sorry.

DR. TEMPLE: Just one thing. I thought, Tom, you were making the point you did because it made you wonder about the analysis; that is, the analysis failed to show something we all expect to see.

DR. FLEMING: That's correct.

DR. TEMPLE: The idea that these kind of data can show you, don't use aspirin alone -- maybe everybody was exaggerating.

DR. FLEMING: I will believe the 20,000 people from randomized trials. My whole point is, when I have a randomized trial telling me something about aspirin versus nothing, now I'm using this data set to answer a different question, what does aspirin add to pravastatin, but it also gives me the same information, imperfect though it may be, about what aspirin adds to nothing, and that information is now inconsistent with my randomized trials about what aspirin does to nothing, it makes me worry about being on the end of a limb when I'm using these data to see what aspirin adds to pravastatin. Not that I have any particular better source of data to use at this point.

DR. TEMPLE: I understand, but the real question is the methodological one. Does this admittedly nonrandomized comparison provide enough assurance so that we really do think that aspirin makes a contribution in the presence of pravastatin. It isn't really to go back and reinspect the advice everybody gives.

DR. FLEMING: Absolutely.

DR. TEMPLE: I'm not referring to what you said.

DR. FLEMING: Absolutely. The comments that I'm making have to do with the reliability of the interpretation of these data in an observational sense, to conclude whether aspirin adds something to pravastatin.

There's a good news and a bad news side to this. Just to summarize, the good news side is, the suggestion is that the effect of adding aspirin is even greater in the presence of pravastatin, and that's the question I'm really worried about here. The bad news is, where I do have an answer from a randomized trial -- i.e., aspirin versus nothing -- it's not consistent with that answer.

DR. THOMPSON: Dr. Fleming, I'm a little confused about something you said. It seems to me that actually a paucity of people are put on aspirin, if you look at and compare it to the number of people that are in these trials. So, somebody's making a decision. I'm always impressed that people in clinical trials always do better than what we tend to see in practice. I'm moved by the idea that it may be that the doctors that take care of them are doing a better job. So, somebody's making a decision here, and I wonder if that decision isn't what's driving us.

But you said that there was an equal decision to put them on aspirin or not. It doesn't look like it's equal. It looks like it's actually much lower. For example, on the top of page 41. So, I want to just make sure I'm understanding this. It looks like, you know, for example in that last column there are only 1,400 people that were not on pravastatin and not on aspirin compared to almost 6,000 people who were given aspirin when they were on pravastatin.

DR. FLEMING: In these two trials, when you add them together, you're correct that the largest fraction of people have been provided aspirin. If you break these people into four groups, pravastatin yes-no, aspirin yes-no, and if you believed these data as being true, what you would say is, certainly, use pravastatin. Also use aspirin if you are using pravastatin, but if you're not using pravastatin, don't use aspirin. I'm saying, if the clinicians in fact knew that, then why is it that when pravastatin is used, 80 percent offered aspirin, and when it's not used, still 80 percent offered aspirin?

They're making the right choice in the first case. They're making the wrong choice in the second. However, I want to emphasize what Dr. Temple is saying. I'm not interpreting these data as being the truth. In fact, I believe these data are not reliable in what they're telling us about the effect of aspirin in the absence of pravastatin.

DR. BELDER: Could I make one comment about this because it's only with respect to one particular endpoint, not with respect to the other endpoints. If you look at C-15, please.

The aspirin effect in this endpoint is much more prominent than in the expanded endpoint that includes revascularizations. It's what Dr. Berry indicated earlier, that we do not pick up a treatment effect of aspirin in revascularizations, and since they are the majority of endpoints that you have in the database, there's a significant dilution.

In addition, I would like to emphasize again that those patients who were not using aspirin at baseline, a significant portion of them started aspirin use as the trials were going on, so we did a very conservative analysis on, if you will, an intention-to-treat basis. So, that, again, would dilute the treatment effect that we would pick up.

DR. BORER: Just to save time here, I think we're being perhaps excessively obsessive in tearing these data apart. At the end of the day we're going to have to decide how convincing we are. We have questions that actually cause us to reason through this, and at that point I think we're going to hear a complete analysis. Speculation here is taking a lot of time.

Dr. Kreisberg, you had a comment?

DR. KREISBERG: Well, I was just concerned about the way I heard the conversation going, and maybe Frank Sacks could clarify it. These patients were not treated with aspirin by the investigators. They came to the study, either on aspirin or not on aspirin, and that's the basis of the analysis. Is that not right, Frank? So, it isn't that they get better management from the doctors who are involved in the study.

DR. THOMPSON: That's not my point. The point is that these are done at institutions that generally have quality of care. They're involved in research. Frequently the patients that are involved in controlled clinical trials appear to do better than those that are not in controlled clinical trials. There are some reasons behind that. One is that they're treated at medical centers that do research. Period.

DR. KREISBERG: I understand that, but most of the patients that are entered in these trials do not come exclusively from academic medical centers, and there are a lot of community participants. It's the academic medical center that serves as a coordinating center.

DR. THOMPSON: I do think we're over-analyzing, but I do think there's a degree of sophistication that goes along with doing controlled clinical trials that benefits patients.

DR. BORER: Alan?

DR. HIRSCH: Well, I don't want to over-analyze how the patients are treated by either academic or primary doctors, but I want to take one of Tom's points just one step further for later discussion. Which is, whenever I see that relative lack of efficacy on the fatal or nonfatal MI endpoint, which would be my signal that I would choose to look at for the aspirin efficacy, I choose to look at that to make sure that I have some sense that, again, these patients treated by their doctors actually took the drug. I want that signal of efficacy, not again in a small trial to prove that aspirin works -- I can look at the broader database -- but to make sure that in this database that I can then look at the crossover for safety, for a combination package.

So, I again look at page 41 of the FDA briefing booklet at endpoint 2, where I see no signal in the 1,460 non-aspirin treated and 5,833 aspirin-treated patients, no impact on MI rates at all. I say, well, who knows? It's too small a sample size, just the luck of the draw for statistics, the wrong model, or possibly really these are casual patients not really taking their aspirin. Maybe they mistake it for Tylenol. Later when I look over the safety database, I have a little bit of doubt.

I'm expressing this now so that later, when we talk about safety, I can come back to it.

DR. BORER: Dr. Pearson.

DR. PEARSON: Ladies and gentlemen, it's my great pleasure to present the medical need for the pravastatin/aspirin combination.

What I'd like to do is bring the perspective of the preventive cardiologist to this discussion. Certainly my interest has been in preventive cardiology and the treatment of high-risk patients for about 20 years. I direct a preventive cardiology clinic at the University of Rochester.

I've also been interested in the policy issues related to this and been involved with the development of the basis for the secondary prevention guidelines for the American Heart Association, as well as the first and second iterations of those guidelines. And more recently my research interest has been really in the implementation of these guidelines, as to the extent to which they're getting out to the patients who are eligible for them. So, I'd like to bring the preventive cardiologist's perspective to the medical need for this pravastatin/aspirin combination.

In the first place, to start this discussion, of course, is in the efficacy, and you've just heard these data. It sounds like everyone is a little bit remorseful for not having paid better attention to that multiple regression course in your statistics course, but I think what we've seen here is, I think, very good clinical trial data looking at two individual trials, the LIPID and CARE trials, as well as meta-analyses from three additional angiographic trials with clinical endpoints, that the combination adding pravastatin is more effective than aspirin alone. We just had a very nice discussion of the observational data, its strengths and weaknesses, as to whether the combination is more effective than pravastatin alone, that is, adding aspirin to the pravastatin, again with single and meta-analyses evidence.

So, I think also it is important, I think, with the second point, to put this into the backdrop of the large clinical trial evidence supporting aspirin use in the secondary prevention of coronary heart disease.

So, the question is, is this a large issue? Dr. Fiedorek presented the initial estimations, and these are the estimations used by the American Heart Association. 12.4 million Americans carry a diagnosis of coronary heart disease. This constitutes, for adults above the age of 45, 12 percent of men and 8 percent of women in the United States. And it's for this reason that many of the public health agencies now are starting to look at these issues of implementation of guidelines as a public health issue, not just a clinical health issue but public health issue.

Even if you were to exclude those individuals who might have contraindications to pravastatin or contraindications to aspirin, usually for GI intolerance, you're still left with about 10 million Americans who would be the eligible population for this combination. The other issue is whether or not this problem is going to be going away, and the answer is no.

This is a small working group that I had the pleasure of participating with for the American College of Cardiology around the end of the millennium, headed by Dr. David Foot, a demographer from the University of Toronto. This basically looks, taking into account the demographics of the United States, the baby boomers, et cetera, at the growth of the prevalence of patients who are going to carry the diagnosis of coronary heart disease. Here we are in 2001 with about 12 million Americans, and over the next 50 years we estimate that this is going to double. About 25 million Americans are going to be carrying this diagnosis. I think this is really the basis for us starting to call this a public health issue as well as a clinical one.

I had, again, the opportunity to head a writing group writing the basis paper for the first secondary prevention guidelines and participated in both iterations since then. I don't remember back in those writings that we ever had much of a question about adding these two issues as important components of those guidelines. First, lipid lowering to achieve an LDL cholesterol of less than 100 milligrams per deciliter, and second, antiplatelet therapy, particularly aspirin, so that these have always been cornerstones of the secondary prevention guidelines as put forth by the American Heart Association and the American College of Cardiology.

What these guidelines do and the wide consensus for their acceptance is for us to look at asking the question, how well are we doing with carrying out these guidelines. And obviously guidelines written, but not implemented, aren't really worth much at all.

I've really had an interest in what we call the treatment gap of the difference between what we recommend and what's actually being done for our patients. And I'd like to make three comments and talk about the relative need of this combination, the three issues.

The first issue I want to talk about is that many patients face a high uphill burden. Obviously, this is and should be a major concern for the medical and nursing communities. This gets at the issue of noncompliance and nonadherence with the recommendations.

Now, let's just consider the typical coronary patient here. And the typical secondary prevention patient might be taking, according to guidelines, a statin, aspirin, an ACE inhibitor and a beta-blocker. If this patient were to have diabetes -- and 25 percent of coronary patients carry the diagnosis of diabetes -- also oral anti-diabetic agents.

Let's consider some of the complications of coronary heart disease: atrial and ventricular arrhythmias, congestive heart failure. A lot of these patients, 60 percent or so by our calculation, will have hypertension, possibly not totally controlled by this. This is a group where we're going to have a large pill burden. I share with Dr. Borer the concern, particularly in the elderly patient, of drug-drug interactions.

But there's another problem with this, and we all know that one of the risk factors for noncompliance and nonadherence is the number of pills and the complexity of the regimens that patients have to deal with every day. So, this obviously is something that our guidelines are actually asking for, and the question is, what can we do to make this all simple.

The question is, is there any evidence to suggest that putting two agents together in a combination pill helps us with noncompliance and nonadherence? I'd have to say this is a relatively slim data set. Certainly I would like more. The American Heart Association has certainly been very interested in compliance in general. And we were able to find four studies in which combination tablets were compared with dual therapy; that is, individual tablets taken together on compliance.

A study in diabetes showed a 21 percent improvement in tablet consumption over a 6-month period in previously treated patients. There are two hypertensive studies. Obviously, there are some combinations available for this. One showed a 13 percent improvement in tablet consumption over a 12-month period, and in another kind of measurement setting, another study showed an 11 percent improvement in prescription renewal as a measure of compliance over a 12-month period.

Then finally, and perhaps the most archetypical polypharmacy kind of situation, HIV, there was a 9 percent reduction in missing even a single dose over a 16-week period if it was put together in a combination tablet versus dual therapy. I really think that it's our responsibility, as individuals who want to see our patients do well, to do all we can to improve adherence and compliance.

Let's talk a little bit about another part of the treatment gap, and that is that many patients fail to receive statins or aspirin. There, in fact, turn out to have been quite a large number of studies. We've been involved with a couple of these, but perhaps I'll show you one of the more recent ones, perhaps the largest.

This is from the national registry of myocardial infarction, with 167,000 patients nationwide from 1999 to the year 2000. Again there are many studies looking at this treatment gap. I picked this one because it's the most recent, and it particularly follows the HA medical advisory, which basically suggests that the initiation of lipid-lowering therapy, particularly statin therapy, in the acute coronary syndrome setting as part of the inpatient discharge regimen, is in fact appropriate. That recommendation predates this study.

Also to point out with this is that this includes coronary patients with no exclusion or contraindications to intolerance of this drug, so that this is in fact the true treatment gap. The treatment gap that is estimated is about 23 percent of patients are going home from their acute coronary syndromes without aspirin, and about two-thirds of them, 63 percent, are going home without a statin. So, the suggestion here then is that this is a large treatment gap, despite our best efforts in implementing our guidelines.

Finally, in addition to this yes-no, are they receiving therapy, there are also additional issues related to how many patients are not optimally medicated. This would include both inadequate and incorrect doses, and just incorrect therapy.

What about statins in this instance? This is a study that I've been involved with, called the lipid treatment assessment program, looking at the use of efficacious statin doses. This was a survey of 4,888 patients from 619 primary care providers around the United States, and of these, 1,460 patients carried the diagnosis of coronary heart disease.

All patients had to be receiving lipid-lowering therapy to be in this study. This was a study of the clinical epidemiology of lipid management in the United States. And in fact statins were used in 85 percent of the coronary patients.

It turned out that the doses proven to be efficacious in randomized control trials, secondary prevention trials in particular, as we noted in this, were seldom used. The vast majority of patients were not taking doses that the randomized controlled trials demonstrated efficacy in. This in fact was the single largest reason, in my opinion, for these patients not getting to their LDL goal. So, there's another issue in terms of not getting to the LDL goal, and that's the current state of therapy, that many patients are not at LDL goal because they're not even at the doses of therapy for which efficacy has been demonstrated. And this is obviously a big concern.

Well, what about aspirin? Is this any better with aspirin? This is the paper first authored by Nancy Cook for which Dr. Hennekens participated. This was a large consumer survey in which 3,818 patients actually carried the diagnosis of known cardiovascular disease, and only 51 percent of those patients reported taking aspirin or an "equivalent." I think that's worrisome enough, but of particular concern was of those who thought they were taking aspirin correctly for secondary prevention, actually 15 percent of them were taking a non-aspirin analgesic, especially acetaminophen, which as you know has no secondary preventive benefit. So, we have a concern about not only incorrect doses but incorrect drugs as well in terms of secondary prevention.

So, in summary, the proposal here is that the pravastatin/aspirin combination in coronary heart patients would provide one prescription with two proven therapies, with virtually unexcelled dual efficacy bases. This provides an advantage of making sure that we have proven doses and that we have proven products getting to our patients.

Just several other points in summary. We feel that this will enhance our implementation of the guidelines that we have, unfortunately, pretty good and recurrent evidence to suggest has a treatment gap.

Second, this would provide us the opportunity to assure the appropriate pravastatin dose, at the same time that those exact doses have 112,000 patient-years of observation showing no safety concerns.

Third is that this would provide us with the more appropriate use of aspirin and not provide particularly the elderly patient -- I share your concern, Dr. Borer, of people who are getting confused about what they should be taking. This would provide them more appropriate use of aspirin at a dose we know has secondary preventive capability.

Then finally, this would provide then enhanced convenience and reassurance for patients and their health care providers in that they are really in fact getting a secondary prevention package.

These I think are the main points that I wanted to cover in talking about the medical need for this combination therapy. Thank you.

DR. BORER: Thank you, Dr. Pearson.

Steve?

DR. NISSEN: Tom, thank you very much. I've long admired your work on the under-treatment of patients with statins, and I want to focus on that a moment.

You've got a lot of data you've looked at on getting patients to goal, and so my first question is a difficult one. What percent of patients in the secondary prevention population would you estimate would get to goal with 40 milligrams of pravastatin?

DR. PEARSON: Let's look at the -- I think the LIPID study would be the best one there. Can we have that slide from the LIPID study?

I might say, Dr. Tonkin, this was almost all the hospitals in Australia. Is that correct?

I just want to make one point here.

DR. TONKIN: Yes, indeed. In the combined populations of Australia and New Zealand, there were 21 million people, and it was 87 hospitals. So, this wasn't just purely the elite centers.

DR. PEARSON: Can we have that slide? It was about the percent LDL lowering in the LIPID study.

While we're getting there, let me also -- this has to do with the potency of pravastatin 40 and the population distribution of LDLs in coronary patients. Those two things were the two parameters.

I believe, in fact, the LIPID study, despite it's being in Australia, I think has something to tell us in the United States, and that is the average LDL was 146, 142. 40 milligrams of pravastatin, then, provided about a 28 percent LDL lowering, which got the average down to about 103 or so. So, on the average, patients were in fact around the LDL goal.

Now, we all know that there are subsets of patients that don't do so well on the diet, aren't implementing the therapeutic lifestyle change, which should give us another 15 percent reduction in LDL prior to pharmacotherapy. We also know that there are some patients with genetic hyperlipidemias that just need triple drug therapy in addition to this. So, we all know about this.

But in terms of almost a population-wide intervention as to how many patients are going to be getting to goal, it's my perspective that this is a pretty good look.

DR. TONKIN: It also indicates the dilution because of the drop-ins to those on placebo. There were 23 percent assigned placebo who commenced open-label lipid-lowering therapy and the 19 percent dropouts on pravastatin, which causes the upward drift over the trial.

DR. NISSEN: Would I be correct in interpreting these data to suggest, then, that something less than 50 percent of the patients in the secondary prevention population would get to the recommended guidelines using the 40 milligram pravastatin dose? Would that be correct?

DR. PEARSON: I would suggest that it would be around 50 percent, perhaps a little higher. We're really quite enthusiastic about the ATP-3 guidelines, therapeutic lifestyle changes. We're getting another 15 percent prior to pharmacotherapy. So, if you put all those together, you'll be a little bit more than 50 percent.

DR. NISSEN: There's other published data that would suggest that it perhaps is only as little as 30 percent of patients. Do you think that's possible?

DR. PEARSON: I think that depends on the population you're starting with, which is the reason why we wanted to look at essentially a community-wide issue here.

DR. NISSEN: An interrelated question, then. So, would you give this combination to a patient with an LDL of, say, 200?

DR. PEARSON: We always look at matching the potency with the intervention. In patients with markedly elevated LDL in my practice I use one of the more potent statins, particularly if the LDL goal is less than 100 or even more. But I would also tell you that we would look at a variety of other issues, including safety and efficacy, the ability to use combination therapies, and a variety of other issues, and we take it really on a case-by-case basis.

DR. NISSEN: Would there be a maximum LDL that you would consider to be inappropriate for the use of this product?

DR. PEARSON: I don't think so.

DR. NISSEN: So, you'd give it to somebody with an LDL of 200 then?

DR. PEARSON: I might, but I'm saying that I think usual practice would be, particularly if we're thinking that we're not going to be using combination therapy, that we'd be looking at probably a more powerful statin in this instance.

DR. NISSEN: If I could just follow up with one more short question, and that is, if you didn't get to goal with this product, if you chose a patient with an LDL of, say, 180, and you gave them this combination and they didn't get to goal, what would you then do?

DR. PEARSON: According to the guidelines, which I think we generally do follow, we would look at a variety of other issues relative to compliance, first of all, if they're complying, and the nonpharmacologic basis of it, but then thereafter the possibility of whether or not a more potent statin would give us as much benefit as perhaps adding another family of lipid-lowering agents to that instance, looking at the HDL and triglyceride and other issues related to that patient. Again, I would do it on an individual basis.

But the answer to your question, would I always change over to a more powerful statin, the answer is definitely no.

DR. SACKS: I'd just like to add a point.

DR. BORER: Wait one moment, please. We have a number of comments and questions from the committee. I'm going to ask the sponsor to just hold off until we hear the entire spectrum of our issues, and then maybe if you want to comment, you can.

Dr. Pedersen was first, and then Bev, Tom, Blase, and Ray.

DR. PEDERSEN: Tom, do you have any information about the reason why physicians do not prescribe these drugs? Is it actually the number of pills that is the main reason, or are there other reasons?

DR. PEARSON: I wish I could tell you the definitive answer there. It's kind of hard to kind of mind-read why physicians don't meet guidelines. I think certainly with the secondary prevention situation, I would have to say I've been quite optimized about U.S. physicians with increasingly using cholesterol-lowering agents. That 37 percent I think is a composite of a variety of things. But I think there has been some progression of use over time, particularly as efficacy studies come in.

We've looked at a couple of data sets, the American College of Cardiology evaluation of preventive therapeutics, the LTAP study, and our own databases, and there are some others as well. There's a variety of issues. One is a knowledge gap among physicians about whether or not there is efficacy of these drugs.

There continues to be a safety issue, which I think we've shown with the clinical trials. Really we don't exactly understand where that comes from because the safety of these drugs is quite extraordinary.

There is also the gap between the acute care setting and the picking up of that patient by the primary care provider. This is, I think, a huge abyss in which patients go in possibly, and this is one of the reasons why those guidelines about starting acute therapeutics in people with acute coronary syndromes, cholesterol-lowering therapeutics as part of the in-patient, was so important because then it's part of the coronary care package rather than something you can start 6, 8, 12 months later, which of course we know is not a good idea.

So, I think it's really a variety of issues having patient factors, physician factors, health care system factors. I think it's a worldwide phenomenon. You're seeing some of that from Europe as well. I think it's something we need to continue to look at strategies about how to overcome.

DR. PEDERSEN: The reason I'm asking this question is that I really doubt that there is a host of physicians out in the marketplace waiting desperately for a combination drug. To my knowledge, another pharmaceutical company, Merck, has already brought to the market a combination of simvastatin, which is their statin, with aspirin, tested on a European market in Sweden a couple of years ago. It may be due to lousy marketing, but they experienced a total flop. Swedish physicians didn't want to use this combination, and it was withdrawn again. This was a test market. I was wondering whether there is a similar experience from the United States, whether you have done some research about combinations of this kind, or whether the company has some experience about it.

DR. FIEDOREK: Well, we're only addressing really the clinical need here. I think we're trying to provide evidence to support the clinical need. If you approve this product, we'll find out.

(Laughter.)

DR. LORELL: Yes, let's return to the clinical need issue. I think you make a very cogent argument for both the need for increased usage of statins in this very high-risk population, as well as issues of the need for enhancing patient compliance once the drug is prescribed. But I'd like to return to the issue of the national guideline goal for this very high-risk population, at a risk for premature death, life-threatening infarction, unstable angina and stroke, for achieving a goal not sort of near 100, but below 100, for LDL-lowering.

I think it's very important for the public record and the public who is listening to understand that this is not sort of a petty adherence to a number, but that the data overwhelmingly supports -- doesn't prove but supports -- the notion that progressive lowering of LDL cholesterol is associated with progressive lowering of risk for these serious hard endpoints. I would really like to see the data presented by the company from both CARE and the LIPID study as to the percent of patients who achieved an LDL goal less than 100, and the percent who didn't.

The reason I think this is very important is there are definite advantages of combination agents for compliance and ease of use. The flip side of that is that there may be a reluctance and a bit of an impediment to changing therapy when you've got both of them packaged. So, I think we really need to see that data.

DR. SACKS: Just to give you the CARE experience, Bev. The average LDL in CARE on prava 40 was 98, so that would be certainly somewhat more than 50 percent of the patients in the CARE trial achieved goal.

Another interesting aspect of that is, in the CARE trial we excluded over 20 percent of patients because their LDL at baseline was under 115. In most of those patients, the LDL was between 100 and 115. In view of the advisability of getting LDL under 100, I would think in all patients, regardless whether their LDL is 115 or 120 or 150, that would add another pool of another 20 percent of coronary patients that with this dose would get under 100.

DR. LORELL: Frank, I appreciate that comment, but I think what this committee really needs to see are the hard numbers. The percent of people who achieved current guideline goal and the percent that didn't. And I think we need to see it both for the totality of the experience and broken down for CARE and LIPID because LIPID I think was skewed toward a somewhat higher cholesterol LDL population and CARE was a little bit lower.

DR. PEARSON: Just one comment to put this discussion into perspective, and that is that I think there have been several surveys as to what currently is achieved in terms of coronary patients being at goal in three or four large studies, certainly one of our own. And a number of about 25 percent pops up recurrently. That's basically how we're doing currently in the United States.

Part of the issue here is that part of those individuals aren't at goal, somewhere between one-third and two-thirds of patients, and they aren't being treated with efficacy-proven agents at all or at the levels of those efficacy-proven agents at which efficacy was shown. So, the other issue is we still have quite a large quantitative treatment cap in terms of LDL, and part of that, in fact, I think is approachable with a combination agent with increased convenience of use.

DR. BORER: Tom?

DR. FLEMING: Well, I'm glad I followed Beverly. She got exactly at the issue that I was concerned about. Steve raised this very important point. What is the amount of impact that we get in LIPID and CARE with the 40 milligram dose reduction, and we saw an average, but an average doesn't tell us specifically how many people in fact aren't going to achieve an acceptable level of reduction.

Exactly as Beverly said was my question. Maybe just to refine it a bit, what I'd like to see is an indication of what percent achieve 100 as a function of what they started at, and what percent achieved at least 110 as a function of what they started at, so that I would get a sense of at least what is the likelihood that if we had a packaged product, people would achieve levels of effect that they would be satisfied with versus needing a change.

Then the second question, for my own statistical sense here, not being a clinician. If in fact you don't achieve 110 or 100, what is the typical approach people would wish to use clinically. Do you switch to a "more potent" statin? Do you increase the dose? What are the consequences, and how would a packaged product impact the flexibility of implementing those alternatives. Two questions.

DR. BORER: Any or all of the above. There are lots of approaches if you don't hit the target. We don't need more information about what -- unless you have the percentages that were asked for.

DR. BELDER: Yes. For CARE, the percent of patients that were actually reaching goal below 100 was 75 percent. For LIPID, I don't know the number. We will not be able to find out during the lunch break either because we don't have access to the database, but that would be somewhat lower. I think it's bigger than the 50 percent, but it's somewhere between 50 percent and 75 percent.

DR. FLEMING: That seems a little bit surprising in view of the fact that the average was above 100. So, how could you have more well than half achieving below 100?

DR. BELDER: That depends, of course, on how the distribution was of the patients across the cholesterol range, and as Frank already indicated, there's a lot of patients with relatively low cholesterol levels.

DR. FLEMING: Maybe after lunch we can see an exact figure.

DR. BELDER: Well, I'm giving you the exact figure. I can put it on a slide, but it will be the same number: 75 percent for CARE. For LIPID, we don't have the number.

DR. PEARSON: And keep in mind that I believe those average levels were intention-to-treat. Right? So, that would include the noncompliant patients where their LDL is not changing at all. So, I agree with you that percent of patients with their LDL below is a more exact number than the mean.

DR. LORELL: Just to make a real brief comment, I think it's a really important issue because we're not being asked to approve here an escalating set of packaged products, and I think it's also very important because it's not the sponsor's job to defend or discuss other companies' products, but this is not a unique drug. There are other choices available that, in the current United States managed care environment, allow you to get to goal often with one prescription and documenting it with a single blood test. So, I think it's really important, if we can, to know the data from the LIPID experience.

DR. BORER: Blase, and then Ray.

DR. CARABELLO: Now that the issue of compliance in pill-taking is on the table, it would seem to me that this opens a Pandora's box. We're being asked to consider the co-packaging and co-production of two different pharmacologic agents that are focused on the same goal. As you point out, many of our patients should be on an ACE inhibitor, a beta-blocker, a statin, and an aspirin, and does that mean that we should co-package and co-produce three or four different agents in the same pill?

DR. BORER: If you want to answer, just make it with a yes or no. If you don't, it's okay, but the question remains.

Ray?

DR. LIPICKY: There are a couple of things, I guess, to talk about, and one of them might be a whole day. But you're not being asked to approve a drug that would be a product, a fixed-dose combination that would be used instead of the individual ingredients. The labeling would say, if you are on these doses of pravastatin or on these doses of aspirin, take me because I am convenient. That's what you're being asked to approve.

The questions that you will be addressing will ask you, do you think this will lead to bad practice? But you are not being asked to approve, put people on this combination product first.

There's a long line of fixed-dose combination products that are anti-hypertensive, ACE inhibitors and diuretics and so on and so forth. Years ago those products were labeled with black boxes that said, do not use me first. Titrate with individual components. Eventually that got to a place where that sort of got modified and changed, and there is a fixed-dose combination antihypertensive product that is in fact for initial therapy. That is, it says, use me instead of an ACE inhibitor or instead of a hydrochlorothiazide product, and it is a fixed-dose combination. Very special reasoning that got it there.

That's not what this is. This isn't use me first. This is use me if these doses are what your patient is getting because it's easier. That's the first thing to point out.

The second thing, I guess, which might be a day-long discussion, is I think it is inappropriate to think about guidelines, and it is inappropriate to think about percent of patients who would fall below something. Guidelines are okay for guidelines, but we, last time this committee met, looked at an antihypertensive drug where all patients were below guidelines for what blood pressure should be, but in fact, although all patients were there, there was a difference in blood pressure control below the guidelines, and that could have been the real clinical benefit.

So, a number is sort of inappropriate to look at, I think, and when we look at antihypertensive drugs, every single sponsor puts in data that say what fraction of patients are controlled, namely 140 over 90. And I have never looked at those numbers. I have advised all our reviewers to never look at those numbers because if it was 141 over 89, it would be a different fraction. If it was 142 over 92, it would be a different fraction. It's a totally arbitrary mind set.

If you in fact even thought that, people would start on the combination product first, and there were doctors looking after patients, there isn't any reason they couldn't add another dose of pravastatin, add more aspirin, add more diet because doctors have to look after patients. Right?

But you will be asked, would the existence of this thing in your judgment alter practice, and you'll be able to make a judgment. But I don't think you need to look at this as the initial therapy of people until doctors know what the response is.

DR. LORELL: Well, Ray, I think your comments are well taken, but in the spirit of this group having to ask the question, will it alter practice, I think it would be helpful as a component in our decision making to know the proportion of people who were or were not below 100.

I also think that although in an ideal world components are, indeed, titrated and used separately, the presentation that we just heard emphasized the clinical care component of initiating these agents at the time an acute life-threatening event occurs. In fact, in real world practice and in my practice as an interventional cardiologist, it is extraordinarily common for the dose of both aspirin and a statin that is started to be the one that is continued for a very, very long time. So, in terms of helping in our clinical decision making, I think these data would really be helpful.

DR. BORER: Let's go to Bob and then Tom and Steve, and then we'll stop because we have another speaker.

DR. TEMPLE: To some extent, the questions raised go to the entire existence of pravastatin. I am absolutely positive you'll find more people reach goal on a different drug. But the expectation is that people will actually measure the effect and see if they consider it adequate, and they might perhaps be influenced by the fact that this drug has much more outcome data than any other drug.

So, it strikes me there's some tension between meeting the guideline with a drug that's never been studied for outcome, or hardly, and instead trying first to get to guideline with one that has a lot of outcome data, and obviously doctors have to figure out what they want to do in that case.

But putting this in a combination with aspirin really doesn't change anything much. 40 milligrams used to be the top dose of this drug. Well, so be it. That didn't get everybody to goal, I'm sure, and then they'd have to decide whether to switch to something else or use it off-label at a higher dose or any of those things, and they would still have to do all this. As Ray said, you have to decide whether the existence of this will keep people from doing what's right, but our assumption is that you're supposed to check the cholesterol levels even when you use a combination, just as you would when you're using it alone.

As Ray said, it is important to us to know whether you think this will alter practice in a bad way, but some of the questions raised really go to the whole question of the drug itself.

DR. BORER: Tom?

DR. FLEMING: Ray, you've said it's not really integral for us to know what fraction of people, if they take the 40 milligram dose, will achieve targeted levels, will achieve a goal, and I understand what you're saying. You're saying the way you're going to label this would be, if in fact in your judgment a 40 milligram dose is what you should be taking of pravastatin and aspirin, then this is the pill for you.

And yet the way that I understand this has been presented to us as the motivation, as one of the critical motivations for doing this, is it's going to enhance accuracy and adherence. Adherence to what? Well, I assume adherence to an intervention that will allow you to achieve what the targeted goal is. If in fact the 40 milligram dose does that in the vast majority of people then I am persuaded that this will enhance accuracy and adherence. But if in fact a substantial fraction won't achieve targeted goal, then why is it I should still think that this strategy is going to provide enhanced accuracy and adherence?

DR. TEMPLE: I think that's the question I was addressing. If you think that this drug doesn't get enough people to goal at 40 milligrams -- maybe now at 80 milligrams it does -- I guess you're proposing to advocate that it be removed in favor of putting everybody on atorvastatin, even though there's no outcome data. What's the implication of your --

DR. FLEMING: No.

DR. TEMPLE: You're absolutely right. It won't get everybody to goal. That's true. So, are we in a position or are you taking a position that you want only the drug that gets the most people to goal?

DR. FLEMING: I'm saying when one thinks about what one is achieving here which, if I understand, is accuracy and adherence enhancement, it seems to me, to have a sense of what the level of that up side is, I have to have a sense of whether or not this packaged product is largely going to achieve the intended outcome. If in fact it's largely going to achieve the intended outcome, then I am persuaded that it's plausible to assume I'm going to get enhanced accuracy and adherence. If, on the other hand, it isn't then I'm thinking this is not necessarily going to have an up side.

DR. TEMPLE: But it can't be better than pravastatin alone at getting you to goal. How could it? It's going to have exactly the same effect on lipids as the drug does.

DR. FLEMING: My question is what happens. If you take the 40 milligram dose with aspirin, in what fraction of people do you achieve goal, or at least a level that care giver and patient would be satisfied, and if they're not, what would they typically do? And I want to have a sense of whether or not there is a large fraction of people that would be satisfied with this combination. If so, then it's plausible.

DR. TEMPLE: You're really asking -- and as Ray said, we are interested in this. If you think this would distort behavior because of the enormous convenience of this, then we would be interested in that concern.

DR. BORER: And we're going to get to that in questions, and the company has already told us they don't have the data we want, so we're going to have to go with what we've got.

DR. BELDER: We have one number: 75 percent for CARE. For LIPID, we're trying to get that number to you in due course.

DR. THOMPSON: I'd like to submit that we're talking about the wrong goal. The goal is to prevent coronary recurrent events and not necessarily a lipid goal. That's the data I think we've been presented to some extent. The issue is not whether pravastatin is an effective drug. The issue is whether the combination of them is better at reaching the true goal, rather than some guideline goal.

DR. BORER: Steve?

DR. NISSEN: Let me see if I can help make both Ray and Bob a little more comfortable. In a perfect world everybody gets titrated. You know what the goal is and everything is easy and your patient comes in, you check their lipids. If they're not there, you do some intervention and so on. But we know there's abundant data that the first dose that patients are started on is often the dose that they stay on.

What we're trying to get a feeling for is the concern that if a product is available and offers a lot of convenience, is widely marketed and available, that there's a certain inertia that's created. It's already a lot of inertia about up-titration and getting people to goal. We know from Tom Pearson's work that most people don't get to goal, unfortunately. What you have to do is, if you get them on this combination product and they're not at goal, you've got to stop that. You've got to start another statin, co-administer aspirin with it, get another set of lipid values.

I'm worried that, on balance, that the societal result, the public policy result will be that fewer patients will get where we want them to be than we get now. That wouldn't be a good decision if that were the case.

Now, I had one other question that I want to raise, and I think to me it's actually not trivial. The major side effect of both statins and aspirin is GI intolerance. A certain number of patients -- I think, Paul, you do this for a living. He can tell you that people come in, particularly with initiation of therapy. If patients are on the combination and they get GI intolerance, then they stop both agents. It gives me a little bit of worry here that patients may stop aspirin because they have GI side effects and continue their statin, or vice versa. But when you put things together, you may lose both components if a patient has a gastrointestinal side effect. It just makes me slightly nervous, and I wonder if anybody else is nervous about that as well.

DR. THOMPSON: You know, I do do this for a living and I'm impressed that we're making it tougher than it needs to be. We're not going to change slovenly medical practice one way or the other. There is this incredible failure to move. I agree with you. I don't think we're necessarily going to make it worse, nor do I think we're going to make it better with an agent such as this. But are we not making it tougher than it necessarily needs to be?

We have an approved drug that may not be the most powerful statin around, but does lower lipid levels, and people use it and there's evidence to support its use. We have another agent that all of us would agree with. Aspirin is effective in secondary prevention. Even though the meta-analysis we were shown raises some questions, none of us on the basis of that meta-analysis would stop giving aspirin after our angioplasties or anything else. So, that's making it more complex.

All we're saying is that there are a lot of people -- and I've left three charts undictated yesterday to get my plane, so I do this every day with a lot of patients. I can tell you that just yesterday somebody said, you're going to give me something else? I'm already taking seven drugs. There is a small group -- not a big group -- that this affects how they think about themselves and their medications, and combining two proven, effective drugs into one may not be a blockbuster seller or whatever, but it's not probably going to be more dangerous, et cetera.

Now, what about the GI bleeding and stuff like that? Even though I'm the one who's kind of picked on the general practice doctrine and said that doctors who do research do better -- I believe they do. But we have to give some credit to these folks out there to notice that if there is a GI intolerance that they're going to stop the combination and put them on the pravastatin alone. I think we also forget that you can add additional doses of another agent if you wanted to. I mean, you could do other things to potentially get these patients to goal.

DR. BORER: We're going to have just one more comment from Ray, and then I want to ask Dr. Pedersen to present his data because they really do get to the heart of the issue, no pun intended, that Beverly has raised several times here.

DR. LIPICKY: Well, you'll come back to this, I'm sure, but the difference, at least from my perspective, from what you're seeing here is that -- I ought to start from some place different.

Coming back to the antihypertensive model, ACE inhibitors and diuretics are approved drugs, and they are taken together, and it's reasonable to do so. My point of view: before one would advocate that one should take both by producing a fixed-dose combination, you ought to have the data you need that says that both are contributing to the good of the product -- and in fact that is a regulation that says that both ingredients have to be working in the product you're approving -- and that you ought to know something about the dose of each that you need to give when they are combined, because it might be different than when they are single.

We have accomplished that with antihypertensives, and we've accomplished that with, say, diuretics and triamptyrine, the potassium-retaining thing. Essentially we knew that both ingredients contributed to the product, and we knew roughly what dose you would need to use of each in combination.

Then it was sort of reasonable to advocate that this fixed-dose combination should be used, and it explicitly said whether you should titrate with individual components first, and if you turned out to be on the particular dose that was available, to then allow that to be used as a convenience. You know, there's a difference between doctors can use something together and saying they should use something together, and that's a subtle difference.

Now I've lost my train of thought.

DR. BORER: You'll find it again. It's okay. Let's go on to Dr. Pedersen, please, and thank you very much, Dr. Pearson. After Dr. Pedersen, any questions anybody has for him, we'll break for lunch, because the FDA has to eat, I'm told again, and then we'll come back and finish up.

DR. PEDERSEN: Thank you very much. I was invited mainly to present my view on whether a fixed dose of 40 milligrams of pravastatin would be appropriate in the majority of the target population, and the invitation came from the FDA. I have the feeling that this question has already been debated into exhaustion, but to justify the airline ticket, I will still give my presentation. It will be short.

As you know, there have been, until two months ago, five large scale, long-term clinical trials with statins in patients at high risk of CHD. I will not talk about the heart protection study because it hasn't been presented in writing yet and it's not important for this presentation either.

These five clinical trials included patients with a variety of LDL baseline levels. The 4S with simvastatin included the relatively high cholesterol level population. The two trials that are combined for the meta-analysis of this meeting, LIPID and CARE, ranged between 100 milligrams per deciliter in LIPID, approximately, and up to about 200 approximately.

The majority of patients were around 150, and for the total meta-analysis, the mean LDL cholesterol level was 148, with a standard deviation of 26. So, the mean plus two standard deviations would be exactly 200. Therefore, for the meta-analysis of the pravastatin trials in this context, extremely few patients have been studied with an LDL cholesterol level above 200 milligrams per deciliter.

Now, from a lot of epidemiological studies, it is known that about one-fourth to one-fifth of patients with acute coronary syndromes or acute MI coming into the coronary care unit have some inherited disorder of hyperlipidemia. The majority have familial combined hyperlipidemia, at least 20-25 percent, and they usually have LDL cholesterol levels about 200 milligrams per deciliter. The rest are made up by familial hypercholesterolemia and other disorders. So, approximately 20 to 25 percent, maybe a smaller proportion in the United States than in Europe, have very high levels of LDL cholesterol levels, which have not been studied with pravastatin.

Now, there is from the epidemiological data good evidence to suggest that the lower the cholesterol, the lower the risk of having a heart attack. However, there is very little data from randomized trials to support the concept of a target level. Neither the European target of 3 millimolar per liter or the U.S. target of 100 milligrams per deciliter of LDL has very good support from randomized data. As you may know, at present there are five large-scale, randomized clinical trials addressing this question, randomizing a total of 40,000 patients. But the results of these trials will not be clear until 2004, 2005.

There is, as I said, a lot of epidemiological evidence, and one European study suggests that once you get below 75 milligrams per deciliter of LDL cholesterol, other risk factors lose their importance. Whether you are a smoker, have hypertension, diabetes, once you get below 75, the risk is so low that you can ignore it.

The studies done with other lipid-lowering drugs like simvastatin in 4S indicated that in the internal analyses, the lower the simvastatin group got in the percent reduction of LDL cholesterol, the lower was the risk. The tertile in the simvastatin group who, after one year achieved an LDL cholesterol lowering of between 44 and 70 percent, had a lower incidence of coronary artery disease in the next 4 years than the two other tertiles. So, in 4S there was a linear relationship between the level reached after 1 year and the risk; the lower you could get, the better. But this is observational data.

In the two trials with pravastatin that have done similar analysis, the CARE study and the West of Scotland study, this finding was not confirmed. On the contrary, in the CARE study, there didn't seem to be much difference of risk reduction whether you had reached a level of 120 or 80 milligrams per deciliter in the pravastatin group compared to those who remained high. A similar finding was done in the West of Scotland trial, where it seemed like about a 12 to 24 percent reduction in LDL cholesterol was enough to achieve the same risk reduction as those who had greater reduction in cholesterol.

So, in an editorial where all these three papers were presented two or three years ago, Scott Grundy suggested that we now have three different models for whether there is a threshold or a target level or not. The evidence from 4S indicating a linear model, the evidence from the pravastatin trial indicating a threshold at approximately 130 milligrams per deciliter of LDL cholesterol, whereas all the epidemiological evidence seemed to indicate an exponential relationship between LDL cholesterol level and risk.

However, the meta-analysis performed with all types of lipid-lowering trials, including the statin trials, would indicate that there is an almost linear relationship between the percentage reduction in LDL cholesterol and the benefit achieved from the side of the patients.

I'm not going to talk about baseline levels, but the clinical practice to date is that patients with familial hypercholesterolemia and familial combined hyperlipidemia are actually rarely treated with less potent statins. They are usually treated with a high dose of highly potent statins or a combination of drugs. Therefore, for about one-fourth of the target population who are discharged from a coronary care unit with acute coronary syndrome, this type of drug would probably not be considered by physicians, or if they are considering this drug, the patients might not be given what is today regarded as the optimal treatment.

However, we will not know until three years from now whether the concept of a target level is correct or not. And until that, I will not press my point very hard about this.

But my final summary is that there is not very good clinical trial evidence on the use of pravastatin 40 milligrams and its efficacy in about one-quarter of the patients with coronary care unit disease.

DR. BORER: Thank you very much, Dr. Pedersen.

Are there any questions from members of the committee for Dr. Pedersen?

I have just one question that really is sort of not totally relevant here. If one were to measure the cholesterol at the time that statins commonly are begun now in the coronary care unit, if one were to do that, and recognizing that at least in acute myocardial infarction, there's an important change in cholesterol when measured immediately after the event, to what extent, if you can actually provide an estimate, would the estimate be incorrect that you were using as your baseline in cholesterol?

DR. PEDERSEN: I believe that most coronary care units today do measure cholesterol on admission into the coronary care unit. And that measurement would be fairly accurate as to what the usual level of that patient is. It is only after about 24 hours that cholesterol levels tend to drop, and they can drop quite considerably by more than 1.5 millimolar per liter over the next few days, and then gradually get back to the baseline level again after about 6 weeks. But if you measure within 24 hours of onset of symptoms, you get a fairly accurate estimate of what the actual level used to be.

DR. BORER: Steve?

DR. NISSEN: Professor Pedersen, difficult question for you, but it relates to your own practice. Is there a level of LDL cholesterol above which you would not use pravastatin personally?

DR. PEDERSEN: Well, first of all, I rarely use pravastatin at all because my experience is mainly with simvastatin. But if a patient has FH or familial combined hyperlipidemia, which means LDL cholesterol levels around 250, I start with a high dose of simvastatin or atorvastatin, usually at least 40 milligrams. And if it's very high, I start right away with 80 milligrams because the probability to get cholesterol levels down to target level, if you think that's important, is very small with prava 40.

DR. BORER: If there are no more questions for Dr. Pedersen, what we'll do now is break for -- dare I say it -- lunch, early. Let's be back here at 12:30 to begin again.

(Whereupon, at 11:32 a.m., the committee was recessed, to reconvene at 12:30 p.m., this same day.)

 

 

 

 

AFTERNOON SESSION

(12:37 p.m.)

DR. BORER: We'll reconvene. We have a little bit more discussion and some data to be presented prior to going to the formal questions that we've been given, but I think the discussion won't take us very long and then we can move on to the questions. It is to be hoped that nobody will have to leave early before we get through.

There are two issues. First, the sponsor has data in response to Beverly and Tom's question, and maybe you want to present that briefly, if you would, about the percentage of patients who achieved 100 milligrams percent of LDL cholesterol in the two trials.

DR. TONKIN: Essentially as you heard this morning, in CARE it was 75 percent of people who achieved an LDL cholesterol of less than 100. I should say that the exclusion criteria for CARE were an LDL above 175 milligrams per deciliter, and that's important.

In LIPID, a total of 53 percent of those on pravastatin achieved an LDL of less than 100, and the question was also asked about the LDL of 110, and that was 68 percent. The exclusion criteria for LIPID were a total cholesterol above 271 milligrams per deciliter.

The other comment that I'd make is that these are intention-to-treat analyses, so this does not -- for example, this 53 percent -- account for the 19 percent of patients who were assigned pravastatin who dropped out from that treatment limb.

If I could show a little bit more data dissecting the material around LIPID a little bit further. There was a lot of discussion about what was the value or the validity, if you like, and how should we look at LDL as against the other part of the argument of whom should be treated. And this is an analysis in LIPID, and I would stress that I do believe that primarily trials examine intervention and not the mechanisms by which they treat.

With that caveat, this is analysis of the lipid parameters, on-study lipid levels, 12 months after recruitment to the study, and looking at the proportion of treatment effect which is explained by those on-study lipid levels at 12 months with respect to coronary events from 12 months over the next 5 years to the end of the study. Because this is a nonrandomized comparison, we adjust for baseline risk factors in all the models.

The proportion of treatment effect, and here's the 95 percent confidence intervals, is the proportionate reduction in the treatment arm effect when one factors in not only the other baseline risk factors, but the particular lipid parameters.

The point I want to particularly make -- a few points, one of which is that the proportion of treatment effect explained by LDL, although it's 38 percent, has very wide confidence intervals. It might even account for the fact that none of the benefit of pravastatin was related to the LDL lowering.

Also, though, the importance of HDL and other parameters that might be there, and I guess to me the most outstanding example of the fact that it is not just LDL lowering that's important is VAhit, which shows a benefit with gemfibrozil in secondary prevention when there is no effect on LDL. So, I think that what this says to me is that we really have to say that there is a lot of uncertainty about what might be the extent to which LDL lowering is important.

What is fascinating to me is the fact that the guidelines are based on this very endpoint data, data from 4S, data from West of Scotland, and these are the hard clinical endpoint data in LIPID, reduction in deaths, reduction myocardial infarction, stroke, need for revascularization. No evidence of any heterogeneity in treatment effect in any prespecified subgroup, and extraordinarily safe. No cases of rhabdomyolysis, et cetera.

DR. BORER: Steve?

DR. NISSEN: I take it from your comments then you don't agree with the guidelines.

DR. TONKIN: No, I think that guidelines are guidelines. I myself am involved in generating and chaired our own working group in developing the LIPID guidelines in Australia, but they are guidelines. I think what we also say in the guidelines is that these actually define what might be reasonable practice. They are not prescriptive.

But I really do believe that there is much more data about hard clinical endpoints, much, much more data about safety, and the two aspects of the lipid-lowering guideline, if you like, in terms of secondary prevention, should treatment be started, an extraordinary wealth of data in terms of what should the goal be.

I think we need to await the trials that have been discussed. We know, for example, there is an effect on inflammatory markers. We haven't discussed that at all. In CARE the benefits were restricted to those people who had high levels of CRP, serum amyloid A protein, et cetera. I think that we've got to be very careful in not going beyond the endpoint data in saying that this is the mechanism by which the treatment is working.

DR. BORER: Thank you very much.

There's one additional safety issue that we didn't touch on this morning. In the sponsor's book, it was suggested that combined treatment could be given at any time of day, and for safety with aspirin, it was necessary to take the aspirin with water. I'm not a gastroenterologist, but common clinical practice, at least where I come from, is to suggest that aspirin be taken with food. Until very recently, the labeling for pravastatin was that it should be given at night because of the somewhat greater efficacy at that time. So, there's some question about giving the combined product any time of day, or giving the two components at night, or at any other time.

Beverly, you had pinpointed this issue, and Beverly actually got a copy of the new proposed label to look through, and perhaps you want to say something about this.

DR. LORELL: Perhaps we could just hear a brief clarification from the sponsor. It's my understanding -- and correct me if I'm wrong on this -- that in the LIPID trial Pravachol was given at nighttime. The current brand-new labeling for Pravachol in the instructions to the patient comments now that it can be administered as a single dose any time of day with or without food. However, in the paragraph describing the pharmacokinetics and metabolism, there is a discussion that says explicitly, "The efficacy of pravastatin administered once in the evening, although not statistically significant, was marginally more effective than that after a morning dose." It appears that information was acted on in the design of the trial.

With this issue of giving both drugs at the same time, although some patients do take their aspirin at night, I think that in general practice in the United States, it is the practice to take aspirin in the morning, not on an empty stomach at nighttime, and usually to take it with food. I guess as the chair brought up, it would be helpful to have some discussion in using a fixed combination about both the issue of time of day and what it should be taken with.

DR. BORER: Does the sponsor have a brief response to that, or any other committee members after that?

DR. BELDER: Yes. The efficacy of pravastatin, when given at night or in the morning, there was a difference of about 2 percent in LDL-C lowering, and our reason to broaden the label to dosing any time of day was based on that very marginal difference, and the fact that perhaps some patients like taking their medications in the morning instead of at night.

When the statins were first developed, there was the at least theoretical thought that since cholesterol synthesis primarily happens during the night that you would expect a greater benefit, a greater efficacy when you would dose it at night, and that's basically how most of the trials were done. But if you then look back at the data that was actually generated, there was no evidence that there was a difference if you took it either twice a day, at night, or in the morning. The confidence intervals of the point estimates were all overlapping, and that's why we asked the FDA to change our label and we were granted to do so.

DR. BORER: Are there any other questions about this or any other issues before we go on to the formal?

DR. LORELL: [Question off microphone.]

DR. BORER: Well, the issue of the aspirin you mean? It becomes moot if you can actually take pravastatin any time of day with food. Then we can tell people to take aspirin however we want them to take it.

Alan, you'll be happy to see that you'll be opining again. The Cardio-Renal Advisory Committee is asked to opine on the benefits and risks of a fixed-dose combination product consisting of pravastatin and aspirin for use in patients who are prescribed these two products as individual entities. It's common knowledge that FDA will accept applications for fixed-dose combination products when two or more approved drugs are commonly prescribed together for convenience and perhaps for better compliance.

In discussion of such products, we've said that availability of such convenience formulations should not alter health care providers' prescribing practices, that is, by not providing a full range of useful doses. Generally that means that a full range of dosing strengths of each individual entity should be available for the combination product, thereby providing convenience but not influencing selection of doses or dosing regimens of individual entities. And we've heard some discussion specifically about that point, and we will again in responding to the questions.

Further, the division has asserted that it should be well established that both entities should be taken concomitantly, since the existence of a fixed-dose combination product implies that they should be taken together, not just that they can be taken together. Generally speaking, the division has required for fixed-dose combination antihypertensive products that the effects of the combination, A plus B, must be greater than the effects of either one alone, A or B. Moreover, the effects of several doses of A in combination with several doses of B must be evaluated, often in a factorial trial, so that some description of the use of A plus B can be compared with either A or B alone.

The sponsor has chosen a single dose of pravastatin, 40 milligrams, and two doses of buffered aspirin, 81 and 325 milligrams, to combine. Thus, there will be two formulations of the fixed-dose combination marketed, 40 milligrams of pravastatin with 81 milligrams of buffered aspirin, and 40 milligrams of pravastatin with 325 milligrams of buffered aspirin. Although initial marketing will be accomplished by co-packaging, formulations of fixed-dose combinations have been prepared and are awaiting completion of stability studies. The fixed-dose combinations will be marketed as soon as data are available. Although the application is for a co-packaged product, the advisory committee is asked to consider the issue the same as that of marketing a fixed-dose combination product.

Pravastatin is approved for use in: A, primary prevention in those individuals at increased risk for atherosclerosis-related clinical events as a function of cholesterol level, the presence or absence of coronary I guess in the presence or absence of coronary heart disease and other risk factors; B, for secondary prevention of cardiovascular events, total mortality and stroke; and C, for the treatment of hyperlipidemia.

Aspirin is for use in the following patient populations. Secondary prevention of death and stroke in patients who have had transient ischemic attacks or stroke, all CNS indications related to thrombotic events. B, secondary prevention in patients who have survived a myocardial infarction, and C, patients who are suspected of having an acute infarction, patients with unstable angina, and patients who are having revascularization procedures, coronary or carotid, who have underlying occlusive vascular disease. Aspirin is given for life, according to the dosing and administration section, for patients who have had unstable angina or PTCA.

Given that preamble, can we define a patient population for whom pravastatin plus buffered aspirin would be indicated, and if yes, we need to define the population or populations. Second, are there populations where there might be net harm from giving both pravastatin and buffered aspirin together, and can we define some of those populations.

The committee reviewer is Alan Hirsch. Alan, why don't you go ahead and we'll see if everybody buys into your answers.

DR. HIRSCH: The reason Minnesotans like to opine is because we have lots o' pines in our state.

(Laughter.)

DR. HIRSCH: So, to start this off, it's easy to define the population. I think the sponsor has helped us with that. This combination would be used in those individuals with established coronary heart disease, and though not explicitly stated, I think there is an assumption that it is also patients with established heart disease with an elevated cholesterol level.

DR. BORER: Okay. Is everybody reasonably in agreement with this? I think that's pretty much agreed upon.

And 1.2, are there patients where there might be net harm from giving the two?

DR. HIRSCH: The contraindications are, I think, in this case the same as the individual contraindications. There's no additive contraindication. So, no specific population beyond the individual.

DR. BORER: Susanna?

DR. CUNNINGHAM: I have a question. What I'm wondering, and I don't know that we have an answer to this, but I'm wondering if there is actually a combined preparation, that people will actually know that they're on aspirin. And therefore, if they were to have surgery or some other event where someone might say, are you taking aspirin, I'd like you not to take any, or the surgeon might request that, will people know? Because I think patients don't always understand what medications they're on, and so it's kind of a puzzle.

DR. BORER: So, that might be a population for whom at least transiently there would be net harm.

Steve?

DR. NISSEN: Actually, Susanna, it's a very perceptive comment. You know, it's been our practice to withhold aspirin for a period of days prior to cardiac surgery because there's very good prospective data to suggest that if you're on aspirin, your chances of having a major or even catastrophic intraoperative bleed are increased. That is a down side of the fixed combination, that both the physician and the patient -- it may tend to obscure a little bit what the components are. It's one of the reasons why in practice I tend to avoid fixed combinations because you may lose track of the individual components that you're giving.

Is it a huge issue? No, but it could be a problem and it could be even a lethal problem under the wrong circumstances.

DR. CARABELLO: Well, in that same vein, or artery, what we don't know is then what would be the down-side risk of discontinuing the statin agent, let's say three or four days ahead of time of surgery, considering its endothelial effects and other effects. I'm making this up, but it's possible that there would be risk involved.

DR. HIRSCH: This sounds like a labeling question, but so we can go one step further, I want to make sure everyone heard me. The sponsor's proposed population was long-term management to reduce the risk of cardiovascular events in patients with clinically evident coronary heart disease. I added the phrase, with elevated cholesterol levels.

Based on the discussion, I thought someone was going to stop me and say, moderately elevated cholesterol levels. Does anybody want to add a population based on that?

DR. LORELL: Dr. Hirsch, I would welcome thoughts of others on the committee on that issue. I think the other issue on which I would welcome some comments from Ray is, he made a very, I think, clear statement that he would view appropriate use of this medication in line with the FDA's opinions that fixed combinations are usually used after safe and efficacious titration of the individual components.

Since this affects so many patients in the United States, is this an indication issue or a labeling issue that clinically evident coronary heart disease following successful titration and safety in the use of Pravachol and aspirin?

DR. BORER: This is not an indication issue really. The fact is that this is a convenience preparation. There's nothing, as Bob said earlier, that would prevent you from giving an extra dose or changing the statin, adding a statin, doing something else, if you thought that your cholesterol goal wasn't being hit. So, I don't think it's an indication issue.

It might be, however, a concern in terms of altering usual practice, and that's something we're going to have to consider. Does the presence of the fixed dose combination make it less likely that doctors will do the titration? That's something that we're going to have to think about and give an opinion about.

Yes, Alan?

DR. HIRSCH: I'm going to try to take your point, Bev, which you've come at fervently, and take it one step further. We're asked as a committee to define the population and the approvability based on a really very elegant, well-done series of meta-analysis we'll get to in a minute, but in the absence of a prospective trial. So, another way of ignoring guidelines is to say, if we decide later that we have safety and efficacy data that should be applied to the population studied. In other words, the sponsor said there's 12.4 million Americans at risk and 10 million who might not have contraindications. The question is, is that the population that this is going to be used for, or really is it the set examined in CARE and LIPID? We'll come back to that later.

DR. LIPICKY: Just for the sake of the record, what you were asking about usually goes into the dosing and administration part of a label. So, it is always included, but it is in a different part.

DR. BORER: Yes, Bob?

DR. TEMPLE: I think the assumption on these convenience preparations, where you're hoping to get the effect of each drug in an appropriate population, is that the indications for pravastatin are unaltered. You do have to say something about using the two drugs together, for example, titrate separately or things like that. But the pravastatin-receiving population here should be the same people who get pravastatin any other time, one would think, except that in addition they need aspirin. Or someone thinks they need aspirin.

DR. BORER: Mike.

DR. ARTMAN: Along those lines, Bob, though, pravastatin is indicated and approved for primary prevention. I think there's controversy about the use of aspirin in primary prevention. So, I wonder about that population. And are we going to extend the use of aspirin in primary prevention?

DR. TEMPLE: No, we're not, until aspirin gets that claim.

DR. BORER: The fact is that people can choose to take the two sets of indications and find out where they intersect and give the drug that way. I think that's fine. I don't think that's our big issue.

Do we believe that the data support the concept that you could define such a population? And even though a randomized, controlled, prospective trial hasn't been performed, I think what we're hearing here is yes, you could define such a population. We might argue a little bit about what the edges are, but you could define such a population.

I'd like to focus just a little bit more about populations for whom there might be net harm. We've heard a couple, and I want to ask the opinion of the committee about another, and then Dr. Kreisberg will have another opinion.

That is, in the elderly on polypharmacy. I would suggest that we don't really have a lot of data about that population, and it's a relatively high-risk population, so I can't say that there's net harm in the whole population or subset of the population for whom aspirin and/or pravastatin might be indicated among people who are over age 75, if we accept that as elderly, or whatever we accept as elderly now. But I think that it's something that we ought to talk about a little bit, again, in part because the average number of drugs that people at that age receive is at least five prescription drugs. I'm a little concerned about that, and I think we don't have enough data to be able to say.

Dr. Kreisberg?

DR. KREISBERG: Well, I'm not sure that this is directly relevant to what you've just discussed, but it seems to me that there's a lot of uncertainty vis-a-vis the NSAIDs and the cox-2 inhibitors that will also play into this, although that's not part of the issue that needs to be considered.

I'd like to amplify on Steve's comment about patients undergoing coronary artery bypass. That's actually a small subset of patients who are having procedures. There are dental procedures and there are minor dermatologic procedures and there are colonoscopies and a variety of other things that occur in these patients that will require a specific set of instructions or understanding about the inflexibility of being on this particular preparation when it comes time to temporarily discontinue a component of the pill.

DR. BORER: Ray?

DR. LIPICKY: Well, Jeff, your concern seems to me to be part and parcel of the individual entities. That's true whether it's a combined product or not. So, I'm not sure that it's a specific concern for thinking about a fixed-dose combination.

DR. BORER: Yes, that's true. I don't think that the concern is specifically because of the fact that the drugs are combined, but if given in a combined product, we do have to be concerned, where we might not be so concerned if we could just give one or the other, which we can because this is a convenience product.

DR. LORELL: Question. I guess I'd like to hear your comments a little bit further. I hadn't thought about this point until it was brought up today about the notion of withdrawal of drug, but I think one of the points that your comment made me think about is that in the older patient there are many instances for procedures, when integrated over time, over six months or a year, where aspirin may be stopped for a period of anywhere from three days for the dentist, or for two weeks or more for a major operation.

It's an interesting comment, given the meta-analyses that we showed today, and were shown, which indicated a very persuasive effect of Pravachol in isolation without aspirin, that there is a protective effect there in those meta-analyses, although the effect of both is clearly greater.

So, I really hadn't thought about this until today, but it raises the issue over a long period of time, the indication of long-term management, that there may be quite a substantive amount of time in some patients' existence where they would lose the protection even of Pravachol in isolation, in addition to aspirin. So, I think it's an interesting point to think on.

DR. BORER: Okay, let's go on to the second question. There are no data from any trial prospectively designed to test the hypothesis that pravastatin at any dose, plus buffered aspirin at any dose produced a better clinical outcome measured by any clinical endpoint than either pravastatin or buffered aspirin alone. Therefore, is that sufficient reason to cease consideration of approval of the fixed-dose combination product? In other words, is it necessary to have the results of specifically designed controlled clinical trials to consider approval of this fixed-dose combination product? If not, what might be sufficient. Alan?

DR. HIRSCH: I think this is an easy question. We wouldn't have a whole day of discussion if we didn't believe that we could look at the database that exists and consider it, but obviously it's preferred to have a prospective trial.

DR. BORER: Tom, do you have any thoughts about this particularly?

DR. FLEMING: I'm a little uncertain about the lead-in paragraph.

DR. HIRSCH: Where is he taking us in this question? Is there something we're missing?

(Laughter.)

DR. FLEMING: It's my sense that, of course, we'd all love to have had a two-by-two factorial design. It's clear, though, how things evolved in time. We had comparative trials of aspirin against nothing, and then with that being accepted, when pravastatin came along, we had comparative trials of pravastatin, yes versus no, allowing for what was in this case the majority of patients being on aspirin. So, those don't provide for us a randomized comparative assessment of one critical issue, which is, what does aspirin add to pravastatin.

But I would say they do provide us a randomized comparison of what does pravastatin add when you have a population of people who would be on aspirin. So, I would think at least one of the dimensions, we do have randomized trials.

DR. LIPICKY: That's correct. When I transmitted these questions by e-mail to get published, I chose the wrong file. And what you just said was part of the edit that I missed.

DR. BORER: Before we get to you, Steve, though, why don't we deal with the sense of the question, though. We don't have a randomized, prospectively designed trial to test the effect of aspirin added to pravastatin. Is that a show-stopper, or can we deal with this?

DR. FLEMING: I don't believe it's a show-stopper. Certainly we strongly urge randomized trials to give us far more interpretable data and a much greater sense of confidence in the results, but there certainly are settings where adequate evidence can be provided in the absence of randomized trials.

DR. BORER: Steve?

DR. NISSEN: I think as usual, Tom, you offer lots of wisdom there. I would suggest, however, that there are some issues, and that is that if we aren't going to have randomized data, prospective data, the data should be very solid, well documented, and fairly compelling.

And there's something we didn't talk about very much today that does bother me. We really haven't the faintest idea what dose of aspirin was used, even what the range of doses were. We only know what the aspirin administration was at one time point, which is the time that it was assessed at the beginning of the trial. We don't know if people dropped in and dropped out of aspirin use during the course of the trial.

So, I think the data is actually weakened substantially from the level of evidence that I would like to see by the fact that we -- I mean, if they had annual assessment of concomitant medicines and could tell us at each year of the trial who was on and who was not on aspirin -- I didn't see any of that data today.

DR. BELDER: We have the data.

DR. NISSEN: Well, you didn't provide it us.

DR. BELDER: You didn't ask for it. I did say that 97 percent of the patients who were taking aspirin at baseline were still taking aspirin at the end of the trial. I said a couple of times that the patients who were not taking aspirin, there was significant drop-in rate. If you want to see the data, we can show you the slides.

DR. NISSEN: There was or was not a drop-in rate?

DR. BELDER: There was a drop-in rate. We can show you the data.

DR. NISSEN: I don't know whether we can do this now or not, Jeff, but to me, if there is a lot of drop-in and drop-out, it's a significant confounding variable. I don't know, Tom, if you could help me with that, but does it confound the data?

DR. FLEMING: It's certainly relevant when I think in terms of what we didn't get by not having a randomized trial. Two of the features are that, on the one hand, we don't have the assurance that those on intervention, in this case aspirin and those not aspirin, are comparable in ways other than they're receiving aspirin.

The other feature is one you're getting at, Steve, and that is we would ideally like to have had a better managed adherence to the interventions. My own sense about that, though, is if we're relying on these 14,000 patients from LIPID and CARE to not only address the question they're obviously designed to address, which is what does pravastatin do, and in most cases in addition to aspirin, but we're also going to use it to try to learn what does aspirin do in addition to pravastatin -- my own sense is if we had actually designed that as a factorial design, we probably would have had more adherence to the distinction between being on aspirin versus not being on aspirin.

In this setting if there is -- we're hearing that in fact those on aspirin, to a great extent, did continue to adhere. We're hearing it was 97 percent. We're hearing, though, that the aspirin patients did have cross-ins. Wouldn't that dilute the effect that we would be looking for? As a result, if you ended up seeing an effect of those on aspirin and pravastatin versus those on pravastatin alone at baseline, wouldn't the sense be that this is then good evidence that there is an effect? It would have been even larger had there been better adherence to non-aspirin?

DR. BERRY: Mr. Chairman, may I address that?

DR. BORER: No, not at this moment, please. I'm sorry, it's my fault. I should have pointed out that once we begin the questions, this is a committee discussion. If we need more information, we'll ask for it.

Dr. Pedersen, you wanted to make a comment here?

DR. PEDERSEN: Well, under these circumstances it may not be appropriate that I comment, but I was just thinking that it would be really too much demand a large-scale randomized clinical trial with a combination, considering the cost and the resources required to do such a trial.

However, since the main argument for bringing this combined treatment to the market is that it will increase the compliance with treatment and also the proportion of the population to be treated, one would think that a trial to prove that might be appropriate. And a trial of compliance wouldn't need more than maybe 100-200 patients, looking at proportion of patients reaching certain LDL targets, proportion of patients actually taking aspirin. You could randomize to the combined treatment or to the usual care, and that would be a simple and inexpensive trial.

DR. BORER: Alan, and then Bob.

DR. HIRSCH: I just want to reemphasize a point for the completeness of the discussion, I think, that Tom made, and in a sense defend my colleague to my left, which is, we looked very carefully at the data for treatment, and I must say I also was not quite aware, other than hearing the 97 percent number, that I knew the aspirin compliance rates.

The question we're asked is, in the absence of a prospectively designed trial, can we consider approval of a combination product with these kind of data. I think that we are going to be, especially if we vote yes, increasingly faced with questions. There are these two anti-ischemic or anti-atherosclerotic interventions. Can they be combined? That's where we started today. Increasingly, there will not be prospective randomized trials. So, this question will, I think, arise again.

So, I think the sense of the committee, despite an elegant presentation and a wonderful data set, is that when there's two treatments in a trial that are going to be expected to be combined, I think this committee probably would like to see compliance rates clearly prospectively collected and presented, so we can have a higher level of confidence.

DR. BORER: I think Bob has a comment about this, but I do want to make the point that nothing in any law or regulation says that you need to have randomized, prospectively designed, placebo-controlled or any other controlled trials. It just says you have to have adequate evidence, and I think that what Tom is suggesting.

And what I think I'm hearing from around the table is that this issue is not a show-stopper. You could use this kind of evidence, but the confidence that we have in the precision of the conclusions that we draw or the accuracy of the conclusions we draw is less than what we might have or would like to have, certainly with prospectively designed trials.

Bob?

DR. TEMPLE: We actually, in most cases where it was considered an issue, have asked for randomized trials showing the contribution of each. But sometimes, for example, you already know that one of the components doesn't contribute. So, for Sinemet, you don't really need to show that carbidopa doesn't have an effect in Parkinson's disease. It's not intended to. So, all you have to do is do the two components, showing that one adds.

So, one of the questions here is, where are we? Are we at the level of, well, we know that, as the question is asked, because it's obvious, because they've done these meta-analyses that strongly support the argument, or do you actually need a trial? Of course, the difficulty here is nobody is going to do that trial. You're not going to leave aspirin out.

I want to make a couple of observations. One, to the extent you think compliance is a problem, as Tom said, that weakens the association. If you still find the association, that's not an argument against it, although we may need to inquire just who was counted as being on aspirin. Does that mean aspirin once, aspirin ever, aspirin all the time? I can't tell the answer.

DR. HIRSCH: But Bob, the noncompliance issue works against efficacy but also impedes the safety analysis.

DR. TEMPLE: Well, again, it's not that you might not want to worry about it, but all the advice people give everybody in the world is, if you need these two drugs, take them. So, why do you have a new safety question about low-dose aspirin? It's the same aspirin that 90 percent of the population is supposed to take. So, I'm not sure why that's a new question.

I just want to make an observation and see what you think about it. It's sort of a problem. It's the difference between doing something under the FDA rules and doing something just because you're a knowledgeable expert. The whole world tells everybody, take aspirin, take a lipid-lowering drug, take ramipril if you need it, get your blood sugar controlled. And they just do that and they give advice and everybody follows it because it seems sensible.

When someone comes to us asking to put those into a fixed combination, we say, well. And I think that's appropriate because marketing something for a particular reason does mean that you have a particular reason for using those drugs together, and we've always taken that position.

I do just want to point out that that raises a problem when it becomes impossible to demonstrate the effect in a formal randomized trial. I don't think you'll find anybody who will do a trial leaving aspirin out of an appropriate population. I don't believe I'd allow myself to be randomized, and I usually take that to mean that most people wouldn't like that trial.

So, the question is what we do in a situation like that. Does that mean you just can't do it? Which is not an impossible conclusion. Or do you find other data that you do your best to probe?

DR. LIPICKY: You've said that you can do it, and so it's time to go to question 4.

DR. BORER: Right, but it's question 3 we're on. Thank you.

DR. FLEMING: Just briefly, though, before we go on. In 2.2 I'd just like to briefly add. Whereas in 2.1 as we've said, it isn't a show-stopper, I would like to reinforce what Steve was saying, in a sense as an answer to 2.2, if not, what might be sufficient? In my own sense, what might be sufficient, of course, is something that would be somewhat setting-specific, but if one has randomized trials for certain elements, and one has for other elements randomized trials in sufficiently closely related settings, and if one has observational studies and properly conducted meta-analyses where, by properly conducted, I mean using a choice of studies and a choice of endpoints that all of us would accept are an appropriate representation of relevant data, and if those analyses provide very strong evidence of benefit, and if in addition to that, one has very strong biological evidence based on complementary mechanisms of action, that's an illustration of some of the types of information that could be persuasive in the absence of formal randomized trials for all of the elements.

DR. BORER: Okay. Now we'll go on to 3.0. One could argue that for the patient population you've defined since the purported mechanisms of action for the demonstrated clinical benefit of each agent are very different, something to do with lipids for pravastatin, maybe even something more than that, and something to do with platelets for aspirin, and maybe something more than that, showing that there were no important pharmacokinetic or pharmacodynamic interactions using surrogates would be an adequate basis for approval of the fixed-dose combination product.

Do you agree with this, and if so, are there -- well, first let's see. Do you agree with this, Alan?

DR. HIRSCH: I found the question again to be intriguing because I think we were told at the beginning that we should be thinking about fixed combinations in the context not just of the lack of interaction but also in the context of finding some evidence that there's beneficial clinical synergy or benefit in compliance.

DR. BORER: No.

DR. LIPICKY: No.

DR. TEMPLE: Synergy is too much to expect. It's rarely encountered. You just want to know that the two drugs do something that neither drug does alone.

DR. LIPICKY: This was written in the sense that you know aspirin works and you know pravastatin works, that you have the trials that demonstrate that. If you know those two things, which is the basis for people prescribing them both, would you be satisfied for purposes of a fixed-dose combination, with something less than bodies? Namely, there's no pharmacokinetic interaction, and the platelet effects of aspirin aren't blocked, and the lipid-lowering effects of pravastatin aren't blocked. This is a hierarchical question, to try and find out what's enough.

DR. HIRSCH: Well, so I'll answer that, and I was ambivalent. I was probably trying to dodge an answer.

DR. FLEMING: Before you do, could I just also ask just to make sure that I'm understanding this, Ray? My understanding of this question says, suppose you have done properly controlled trials that establish each of the individual components is effective individually.

DR. LIPICKY: No. It is in the sense that Bob is saying now.

DR. TEMPLE: Let him finish. He's going to get to where you want him to.

DR. FLEMING: And if in fact you have properly controlled studies that show each of these components is effective individually, and if you then have data on PK and PD that indicates there's no interaction, are those sources of information alone adequate without knowing anything about combination efficacy?

DR. LIPICKY: That's correct. Without ever doing the meta-analysis.

DR. TEMPLE: Together with the fact that they work in a completely different way, which you might choose to believe means that their independent effects will be manifested, even without measuring that. That's the question.

DR. HIRSCH: I'm going to keep the discussion going by simply charging in and saying I waffle. It might be under certain circumstances.

DR. LIPICKY: Under this circumstance.

DR. HIRSCH: This very circumstance?

DR. LIPICKY: Yes.

DR. HIRSCH: I would not be personally prone yet.

DR. BORER: That means no.

DR. HIRSCH: It's a no.

DR. BORER: We'll go around and get some comments about this because this is an important point.

And I'll tell you that I certainly wouldn't agree with this either, that this alone would be adequate because there are several other issues that we're going to get into, specifically one that was highlighted in your preamble about affecting practice patterns that would be necessary to make some judgment about in order to determine whether the specific combination on the table was approvable. There may be other issues as well, but the point is, I wouldn't agree with this.

Steve, you want to make a comment?

DR. NISSEN: Yes. I would emphatically think it's not adequate, and I'm going to give you an example, although it's a controversial one. There's some data out there that suggests that aspirin works, that ACE inhibitors work, and there's also some data that suggests that when you give ACE inhibitors with aspirin, it reduces the effectiveness of the ACE inhibitor.

Now, I think that's controversial, but there certainly are plenty of examples out there where two drugs that independently are active, that when combined, if studied carefully enough, would show less combined effect than the individual components. So, I would feel very strongly that we should not set the standard so low.

DR. BORER: But this question specifically states there are no important pharmacokinetic or pharmacodynamic interactions.

DR. NISSEN: Yes, but it doesn't have to be pharmacokinetic or pharmacodynamic. It can be biological. That somehow or other, that biologically, when you combine two drugs, it does something you didn't anticipate to the biological system that makes one or another of the components work less well, and it has nothing to do with PK or PD.

DR. HIRSCH: It could be biobehavioral. It could be, again, how the patient actually then, therefore, is taking the two tablets and how they're given or withdrawn in real life.

DR. THOMPSON: Anything is possible, but we don't have any evidence to suggest that, and do we put any credence into the fact that this is a common, widespread clinical combination, that every one of us as clinicians would do? How does that figure into this?

DR. BORER: Just one second, Paul. I think that the issue here is first a generic one, and second, applying it to this particular concern. I think what we're hearing from Steve and from Alan and from me so far is that no, for perhaps different reasons, just knowing that there are no pharmacokinetic and no pharmacodynamic interactions of the two entities isn't sufficient by itself as a basis for approval. It might be, but it isn't sufficient.

DR. TEMPLE: Can we just tease two parts of that? You're going to come back to the question of whether the fact that they're in a fixed combination screws up your ability to use them properly. Perfectly good question. I think this was intended to ask, do we know that these drugs, used properly, will have an additive effect because they work differently and because they're well studied alone.

And I've heard several different answers. Your particular answer was, well, I might believe they would work, but I'm very worried about whether I'm going to change people's behavior. Perfectly good question. See, you could have a well-designed factorial study and still worry about that.

So, they're really a separate question. One is, do I know that the two will work. A second very interesting question is, will people use these properly if they're available in a fixed dose? But I think that's a separate question.

DR. BORER: Beverly?

DR. LORELL: Well, I think that in adhering to the strict wording of this question, and in responding to Ray's comments, I think we were provided with very clear data regarding the peak levels of the two drugs when given at the same time, and the area under the curve. I don't think we were quite provided with one of the things that Ray mentioned that I would like to have seen, and that was that giving the two drugs at the same time does not modify standard indices of aspirin's effects on platelet activation and aggregation. It may not, and my best guess, if I had to make a guess -- does it? The answer is probably no. But we weren't shown that data.

In argument that it should be shown, there are studies, not with this combination but with other agents that have been in the literature recently, suggesting that the timing of when one gives common other used drugs, in addition to aspirin, can profoundly modify the pharmacokinetic activity of aspirin on the platelet.

So, this is not a minor point if in the trials Pravachol was given in the evening, and in large amount of patient practice, they can't tell us how and what time aspirin was given, but it is widespread practice for aspirin to be given about 12 hours apart in the morning.

So, I would like to have seen pharmacokinetic data demonstrating very clearly that when you give both together you're obligated, you're getting them there at the same time, that you don't alter the pharmacokinetic activity on the platelet.

DR. BORER: What if you had had those data? What if you had, by whatever standard you wanted to set, adequate pharmacokinetic and pharmacodynamic data indicating no interaction? Would that alone be sufficient?

DR. LORELL: I think this question is asking something different. It's saying, are there sufficient data to support the presence or lack --

DR. BORER: No. That's 3.1. 3.1 is, do you agree with this, which is the statement that the surrogate pharmacodynamic data and pharmacodynamic data, if you had them, would be adequate. Then we go to this product.

So, would you agree with the idea that you could approve a combination of two different drugs, drugs that presumably act differently, if you know that there are no pharmacodynamic and pharmacokinetic interactions. Then we get to the issue of whether that applies to this drug. Would you accept?

DR. LORELL: It would have helped me to say there are both sufficient data to support the lack of a significant interaction, as well as it would have helped me think about answering number two.

DR. LIPICKY: But the question that was answered was I don't care if they had it, that wouldn't be enough. So, it's a matter of would that be enough if they had whatever it was you wanted. Then you deal with do they have this and do they have this.

DR. LORELL: I would say it's a component of additional data I would like to have.

DR. LIPICKY: No. Enough, enough.

DR. LORELL: It wouldn't have been enough in isolation.

DR. LIPICKY: That's the question.

DR. HIRSCH: Let's take each part of the question in turn and come around.

DR. BORER: Okay. What Beverly has just said is that these data alone would not have been enough.

Susanna?

DR. CUNNINGHAM: I'm just curious. I believe that this says a lipid effect and the platelet effect, but I believe, but I don't know for sure -- and somebody else can give me more information -- that both drugs also have an anti-inflammatory effect. And what do we know about their interaction in terms of enhancing each other's action as in anti-inflammatory drugs?

DR. BORER: My guess is we know very little. But again, I think just in terms of getting through this thing efficiently, let's say we knew all that stuff. Would knowing that there were no important pharmacokinetic interactions and no pharmacodynamic interactions in two molecules that act differently be enough to allow approval of putting them together in a fixed-dose combination to be given to people? Several people have said, no, that's not enough.

Then we go on to are there sufficient data here. If it wouldn't be enough, you don't have to go on to ask if we had sufficient data here.

DR. CUNNINGHAM: I don't know if I know enough to answer that.

DR. BORER: Okay.

How about the others at the table who haven't commented yet? Mike?

DR. ARTMAN: I think I agree with the sort of sense of unease that I've heard so far, and I would say no.

DR. BORER: Blase?

DR. CARABELLO: Yes, I also would say no.

DR. BORER: Tom?

DR. FLEMING: I say no, but I would like to be real precise about what I'm saying no to.

My interpretation of this question, right or wrong, is if we know we have two agents and individually we know that those two agents are effective, and in addition to that now we're adding that we know that they have purported different mechanisms of action and we have done PK and PD studies to show no interaction, is that information in its own right adequate to approve a fixed-dose combination?

By the way, I would say all of those pieces are very important to ultimately having what's adequate, but those pieces themselves aren't sufficient in my view. There is additional insight I would like to have directly clinically about what the combination does as the additional piece to add on to those important elements to come up with what is sufficient. Hence, with that interpretation of the question, my answer is no.

DR. BORER: Paul?

DR. THOMPSON: Yes. I would say that there's sufficient data to answer 3.1.1, that there's a lack of significant pharmacokinetic interaction, and I would suggest that --

DR. BORER: But we're not at 3.1.1. We're at 3.1. Do we agree --

DR. THOMPSON: No, I don't agree with that.

DR. BORER: Okay. Then we don't have to go to 3.1.1 because if you don't agree those data would be enough, then we don't have to determine whether they have those data or not yet.

Dr. Kreisberg?

DR. KREISBERG: My answer is no.

DR. BORER: No, okay. So, I think it's unanimous. Everybody said no.

Now, Ray, do you want a response to the subsidiary questions?

DR. LIPICKY: No.

DR. FLEMING: Could I have a clarification of that?

DR. BORER: Yes.

DR. FLEMING: At least in my own answer, I said those conditions aren't sufficient, but they are certainly relevant to ultimately what I want to consider as what may be sufficient.

DR. BORER: So, you may get back to 3.1.1.

DR. FLEMING: Eventually we're going to have to answer 3.1.1 and 3.1.2. At least I want to answer 3.1.1 and 3.1.2.

DR. LIPICKY: Well, we don't want your answer, Tom.

(Laughter.)

DR. BORER: And as advisors we can only give the advice we're asked for.

DR. TEMPLE: Jeffrey, it is important and it's important to other potential convenience preparations because there's always going to be a rationale like this. You know, one lowers lipids, one lowers this, one lowers that. And so, figuring out how far you think we should go with that information alone is of considerable interest to us. But I'm sure, as Tom was about to say before he was interrupted, you can keep those things in mind even while you consider the adequacy of the other data. I'm sure everybody will.

DR. LIPICKY: And there isn't any question about the importance of all that stuff. I'd just like to get to question 5.

DR. BORER: That's good. Well, we're at question 4 right now actually.

The sponsor has provided three different meta-analyses, data from five placebo-controlled trials, the total number of randomized patients being 14,617, that address whether or not administration of pravastatin plus buffered aspirin has a greater effect than either buffered aspirin or pravastatin alone. Some of the selected trials required that patients have greater than normal levels of serum cholesterol; others did not.

4.1. Do these 14,617 randomized patients represent a reasonable approximation of the patients for whom this combination product would be indicated?

Alan?

DR. HIRSCH: Yes, but I was again very bothered by the relative lack of women and minorities, and when we talk about generalization to the American population, we've got to do better. But knowing that the general database we always look at is not much better than this, I'll say yes.

DR. BORER: What about the fact that there were upper limits on cholesterol levels?

DR. HIRSCH: My understanding of the word "reasonable" is not all-inclusive, broadly representative, but let's hear everybody else's opinions.

DR. BORER: Steve?

DR. NISSEN: It weakens it a little bit, not a huge amount. You'd like to have all comers, but these trials, at the time they were designed, were designed around, some of them, fairly narrow ranges. We heard from Dr. Pedersen that a quarter of the patients that come in with myocardial infarction have LDLs of over 200, and those people would have been excluded from at least some of these trials.

So, I think when you pull all these patients together in a meta-analysis and you've restricted the lipid range for some of those components, it's a source of uncertainty. I don't think it's a huge source of uncertainty, but there is some uncertainty related to that.

DR. BORER: But the sense is that this is not a show-stopper either I take it. Does anybody disagree with that or have any other opinion about this?

(No response.)

DR. BORER: Let's go on to 4.2 then.

From the results of the meta-analyses, do you conclude that the data show that pravastatin plus buffered aspirin has a greater effect than either buffered aspirin or pravastatin alone? And there are two subheads to that, and I'll read them first because I think the answer is all-inclusive here.

Using as a standard of two trials at a p less than .05, is the strength of evidence from the meta-analysis as strong as this standard?

Using as a standard of one trial at a p less than .05, is the strength of evidence from the meta-analysis as strong as this standard?

Alan, why don't you go ahead, and if there are some technical issues, we'll ask Tom to comment as well.

DR. HIRSCH: I've learned to give the yes/no first and then to opine. So, I think the answer is clearly no, but let me just say why.

Both in the FDA briefing document, as well as I think what Tom said initially, there are many reasons why meta-analyses cannot hold the weight of a prospective trial in general, and for me, reviewing the application, although I clearly see an efficacy signal for aspirin with pravastatin and not as strong by itself, there's always this weakness in being able to interpret data in a meta-analysis form which I think is also evident in this very robust meta-analysis. So, the answer is no.

DR. BORER: Tom?

DR. FLEMING: Well, this certainly is a difficult issue, difficult question to answer because those of us who believe strongly in the greatly enhanced interpretability of randomized trials struggle mightily when we're faced with a situation such as this. And there is substantial evidence here and there is a strong motivation or rationale for why the more complete access to fully randomized data would be difficult to achieve, although I'll also argue the fact that it's difficult to achieve something doesn't mean having less than reliable evidence makes it any more reliable.

But I guess my overall sense here, when I look at the data that's been provided, is we look at the progression in clinical practice that led to the nature of the trials that were conducted that provide the evidence that we need to answer this question. So, initially we began with aspirin in placebo-controlled studies and my sense is even though there is some diversity in the level of effect that those trials have established for aspirin, when compared to controls, that when one looks at the aggregate of evidence, I think there is substantial evidence establishing the effect of aspirin in randomized trials when looking at it as aspirin versus control versus nothing.

Aspirin then became quite widely used, and then when pravastatin came and the trials that were being done were assessing the effect of pravastatin, even though in a sense I would have loved to have seen a factorial design conducted at that point, where patients were randomized to aspirin yes-no/pravastatin yes-no, I can understand the rationale by those who were designing the trials to believe that aspirin would be important to provide at least for the clinical caregiver to choose, whether or not they would use it. As a result, those studies, and most specifically CARE and LIPID, provide I think a very proper and reliable assessment of what pravastatin adds to aspirin, but obviously aren't designed to provide a reliable conclusion about what aspirin does.

We're left, in answering this question, with the need to look at the aggregation of available evidence to answer two questions. What does pravastatin add to aspirin? What does aspirin add to pravastatin? Are they both integral to the combination?

I think doing some kind of meta-analysis formally or informally is a very appropriate way to proceed. Of course, there's always the challenge, as has been clearly and appropriately identified by the FDA review, that when you do a meta-analysis, it's important for us as consumers of that information to be confident that this is a representative summary of relevant information rather than a retrospective choice of those specific studies, subpopulations, and endpoints that might best defend or achieve a conclusion that those that are conducting the analysis would like to achieve.

My sense is that if we're looking at the question that I defined, it is certainly relevant to focus on those studies that the sponsor has put forward, but I'm open and very interested in comments from my colleagues if they think a different choice would have been more appropriate. I believe the focus on CARE and LIPID is a very logical and appropriate focus here.

I also think with the cross section of endpoints that we've been presented here, which are basically CHD death, fatal/nonfatal MI, ischemic stroke, and the combination that includes revascularization, that that is the array of relevant clinical endpoints as well. So, I'm not particularly troubled by either of those features.

So then I'm, as a result, comfortable in looking at these data particularly in the sense that they were designed. They were designed to address specifically whether pravastatin, in a randomized fashion, adds, and in most cases, adds to aspirin. There are many summaries, but if we look at the model 1 analysis that the sponsor provides, which is the traditional Cox regression analysis, and we see the data on C-11, C-12, and C-13, we see a summary in the yellow bars on those slides as to what the data are showing us about what the effect is of pravastatin when added to aspirin. And we see, I think, consistent evidence of benefit across all of the endpoints.

In particular, when we then divide this in the next slide into LIPID versus CARE to see, whether or not, in the spirit of are there two studies that are adequate and well-controlled at the .05 level, I see evidence that I view to be adequately convincing. So, when I look at this source of information, I'm persuaded that the standard for strength of evidence has been met for establishing that pravastatin adds to aspirin.

Well, that was the easy part. The tougher part is if we have to rely on this same source of information, is this adequately convincing that aspirin is integral and it adds to pravastatin.

I struggle greatly with this when one looks at the information that's presented here, which are the blue lines -- and in particular, slide C-13 presents for these three primary endpoints, what is the strength of evidence for what aspirin adds to pravastatin, looking separately at LIPID and CARE -- I see evidence, which if I can view this to be reliable -- i.e., if this were from randomized comparisons -- I would view that this strength of evidence is definitely convincing to me that aspirin is, in fact, integral as well to the effect of the combination.

So, that leaves me then with one final dilemma, and that is, these aren't from randomized trials. What is the plausibility that these differences, in fact, could be more due to the systematic differences in patients who chose to use aspirin versus chose not to use aspirin as opposed to the actual effect of aspirin itself? That's an incredibly difficult question to answer.

I could be readily persuaded that those people who would be put on aspirin wouldn't be randomly done, but I could also be persuaded, although there's really no evidence in the covariates that we have, that those people who were put on aspirin might, in fact, have been more ill.

The other feature here -- and it comes back to something Steve said, but actually it makes me a little more confident in these data -- is we don't have the same level of confidence and adherence to the aspirin versus non-aspirin, and it's the point that was reiterated by Bob Temple. If anything, that would lead me to think that we might be underestimating the effect.

What we have in these analyses is the ability to look not only at what aspirin adds to pravastatin but what aspirin adds to nothing, although it's not as reliable because it's not in a randomized trial. But what's interesting is when you look at what aspirin adds to nothing, you're getting an underestimate of effect. And this was an issue that I was probing at some length earlier, the cup half full/half empty.

It made me, in the cup half empty, a little more skeptical about what we could say about what aspirin adds to pravastatin when we see evidence about what aspirin adds to nothing as being less than what we would have known from the randomized trials.

But the cup half full says to me, well, but this is consistent with an underestimate of effect that could readily be achieved if those that are being administered aspirin are, in fact, if anything, somewhat more seriously ill or at higher risk, and if those on the control arm had a greater propensity or likelihood of crossing in.

So, when I think about all of those features, it actually leads me to think that the evidence here from this nonrandomized comparison surely is far less reliable than that I would have from a randomized comparison, but the things that I can think of that are the likely systematic biases would tend to make me think that we're getting an underestimate of effect, and the levels of effect that we're seeing, if they were from randomized trials, would meet my sense of standard for strength of evidence.

So, when all is said and done, as rarely as it is for me to be able to say something that isn't randomized probably is adequately convincing when one considers all of this and the fact that you have different mechanisms of action, I think I am persuaded, when I look at all of this information, that yes, each of these components is contributing.

DR. BORER: Ray, it really isn't necessary, is it, for us to answer precisely 4.2.1 and 4.2.2? I don't think we can provide an equivalence answer. I think either we'll all agree with Tom that the data are adequately compelling to convince us that both components are integral to the combined effect or we're not.

DR. LIPICKY: But Tom gave sort of a binary qualitative answer. I'd like a little bit of a quantitation with respect to the confidence you have in the conclusion you drew.

DR. BORER: Okay.

DR. LIPICKY: And that's what 1.1 and 1.2 are devoted to. It doesn't need much discussion. He just has to say it's sort of one trials, sort of two trials, in between, or it's even better than two trials.

DR. BORER: Okay. Steve?

DR. NISSEN: I wasn't going to take that so much, but I want to remind everybody of something. A few years ago, it was just absolutely clear and obvious from nonrandomized, sort of observational data that estrogen was very good for cardiovascular protection in women. In fact, many women I know in my practice were pressured heavily to take estrogen by their family practice physicians because huge, enormous observational databases showed that women that got estrogen had a lower incidence of coronary heart disease. And now, as it's tested prospectively, we find out it isn't so.

Now, is it exactly the same situation? Of course, not. But what happened was that women who chose to take estrogens were different from women who didn't. And you raised the question, Tom -- and I agree with you completely -- that people that chose to take aspirin or whose physicians chose to give them aspirin in this trial -- could they have been sufficiently different to account for some of this?

And, Ray, I don't know in any given situation how you ever can correct for that. It's a huge hazard. So, the only question then you have to do is look at it and say how plausible is that possibility. And, boy, we've been wrong. On the estrogen story, we've been as wrong as we could possibly be.

DR. LIPICKY: You can be wrong with a p of .05 in a prospective randomized trial. Okay? So, I just want to get a feeling of how wrong you think you can be. That's all.

DR. BORER: Let me ask Steve then, since he was the last to speak, and therefore it's easiest to keep him going, do you find these data sufficiently compelling so that you can conclude that both components add to the combined effect of the combination product?

DR. NISSEN: I do, but I'm nervous. And I gave an example of a situation where people who thought they really understood this very, very well turned out to be absolutely dead wrong about another form of therapy.

DR. BORER: So, are you nervous enough to say this is as good as one trial at p less than .05, or are you even more nervous than that? Because nervous I think means you're not willing to say it's as good as two trials at p less than .05.

DR. NISSEN: I'm going to think about that before I answer.

DR. LIPICKY: The point estimate is in the right direction. No question. No one is going to argue about that. The question is do you think that's real, and then how certain are you of that? And are you going to put us in the position of saying, well, you ought to prove things with a single trial of .05?

DR. HIRSCH: Ray, this is about like a single .05 trial, meaning that we have really quite good data with quite good fidelity, with a p value that looks sort of appropriate, but we could be wrong. That's where we are. It's about equivalent to one well-designed trial.

DR. BORER: Blase and then Bob.

DR. CARABELLO: I think the data are compelling, but I don't think you can compare the apple and orange of randomized trials to a meta-analysis.

But unlike the estrogen situation, all the toothpaste is out of this tube. There is never going to be this trial. There is never going to be a double-blind, randomized, two-pole trial of these two drugs. It's not going to happen. Millions of Americans are already on this combination, and unless and until some data to become available to suggest that maybe they shouldn't to be, to then throw the whole issue back, it just isn't going to happen.

DR. BORER: I'm not sure that that's what we're being asked to do or not to do. I think what we're sort of being asked is should the FDA put its imprimatur behind the combination if we don't have the data that normally -- the evidence of the strength that we normally would require to allow the FDA to come to that conclusion.

DR. LIPICKY: I want to emphasize that Bob said yesterday that a p of .05 single trial was good enough to get approval. So, saying a p of .05 single trial isn't the death, I just want to get a feeling for the strength of evidence. And so far you're telling me, well, I don't know how to tell you what I think. That's what you said so far.

DR. BORER: No, no. Very clearly Alan said, one trial at p less than .05.

Bob?

DR. TEMPLE: Well, you really have to make the same distinction Tom made. On the question of whether prava adds to aspirin, A, you don't need a meta-analysis. Both trials showed it. They showed it for all endpoints. The p values were pretty extreme. And I must say, although we asked it, I find it hard to imagine that anybody doesn't find that part of it convincing. Those trials were mostly done in aspirin users. All the evidence you have on the effect of prava is from trials in which most people got aspirin. So, that doesn't seem hard. It's the other part that seems hard because you're into epidemiology or something.

I just wanted to go back to something Tom said before, which was that he would like to be allowed to think about the fact that the two drugs work in completely different ways and factor that into his thinking, which I'm sure he did. When Ray wanted to pose the one study at .05 versus two, I said, why don't you cross that out? They can't answer that because it is, to a degree, apples and oranges. It seems fairly obvious that one is bringing one's impression about how things work, with all the flaws that that can induce, just as everybody appears to be wrong substantially on what estrogens do. So, that does seem part of it.

I want to mention one other thing. I somewhat hesitate to do this. There are actually trials in which antiplatelet drugs have been given to people who are on statins, not trials of aspirin, but it's not out of the question we could take a look at trials of clopidogrel and things like that to see whether there was an effect of an antiplatelet drug. It raises some of the same issues as yesterday when you're already on a lipid-lowering drug. We haven't done that. We haven't ask the company to do it, but those data are in the public domain. It might be possible to do that to test the hypothesis.

DR. BORER: I think would think that clopidogrel would be the wrong choice since the drug was approved because of its putative superiority not only to a placebo defined on aspirin, but to aspirin.

DR. TEMPLE: Well, see, that's a good question. It depends on what you think the question is. And I'll tell you what my thought was. We weren't trying to answer the question whether you get precise estimates of what the exact effect of these things are together. It was really a qualitative plus or minus thing, answering the question, if your lipids are under great control, does doing something to your platelets make a difference. So, maybe in that case, another antiplatelet drug, even one that was better than aspirin, might be pertinent.

DR. BORER: I'd like to disagree with that. I think -- and you'll have to correct me if I'm wrong -- that in asking the FDA to approve a combination drug for prevention of events, which is what we're talking about when we add the aspirin on, we'd like to have some sense that there is a quantitative effect, any quantitative effect. And looking at the quantitative effect of a drug that's more potent than aspirin may not tell us that aspirin really adds in an important way to pravastatin in prevention of events. Now, it may add in other ways. It may have platelet active effects, et cetera.

DR. TEMPLE: Jeff, might it not tell you that even if your lipids are just perfect, fixing your platelets makes a difference?

DR. BORER: Yes, it might well.

DR. TEMPLE: In some ways that's the question.

DR. BORER: Well, I'm not sure.

DR. TEMPLE: Not the whole question.

DR. BORER: I'm not sure I'd agree with that.

DR. THOMPSON: Dr. Borer, I'd like to come back to the question --

DR. BORER: Wait, wait. Just a minute, Paul.

I don't want to carry this discussion ad nauseam. The points have been made. I think what we're trying to do here is to determine how compelling we believe the evidence in favor of an additive effect either way is. And I think we've heard that in general there's evidence, and so far everybody has been willing to accept that there's evidence and both components do add to the effect of the combination of drugs.

The data, as we've heard from Steve and from Alan, aren't as compelling as we usually expect to see, but they're there. That doesn't mean that they're adequate or inadequate, but they're there. And Alan suggested, in terms of the degree to which he's convinced, he's as convinced as he would have been if he had seen one trial at p less than .05.

Let's see if we can sort of narrow the answer to that question, and let's hear from the voting members. Mike?

DR. ARTMAN: What's the question, Jeff?

(Laughter.)

DR. BORER: From the meta-analyses --

DR. ARTMAN: Okay, so you want to answer 4.2.1 and 4.2.2?

DR. BORER: Yes.

DR. ARTMAN: I would agree with what Alan said, that the level of confidence I have in this would be comparable to a single trial at a p less than .05.

DR. BORER: Tom, do you want to finalize your answer now, or do you want some time to think?

DR. FLEMING: Well, I think I've already said the essence. I'm very comfortable to say that the contribution of pravastatin to the combination has been established by the standards of two studies at the .05 level.

The combination of aspirin is where I struggle greatly. It's extremely rare for me to find nonrandomized data as adequately convincing. The basis that I have judged in this case that it is is essentially built on, A, the fact that I think the evidence is adequately convincing that aspirin, in the absence of pravastatin, is beneficial according to the standards we would usually have for strength of evidence; B, that the biological plausibility that it would maintain that effect in the presence of pravastatin because of different mechanisms of action is relevant; and, C, because the evidence that we do have, flawed as it is because it's not from randomized comparative studies, gives us very favorable point estimates where the best judgment that I can make about where the biases would be, in terms of selection factor as to who got aspirin versus who didn't, and in terms of lack of adherence diluting differences, would if anything dilute the estimates that we came up with, which in fact does appear to be what we're seeing when we look at these data from the aspirin versus nonaspirin. It's the aggregation of all that that provides me a sense that this is adequate, as incredibly rare as it would be for me to arrive at that conclusion in the absence of randomized comparisons.

And I'm not comfortable, though, stating numerically whether or not this is the same as one or two. It's not nearly as convincing as what you would have if you had had the data from two randomized trials that provided strong evidence of benefit. Nevertheless, it's my sense, for the reasons that I've given, that the aggregation of this evidence is adequately convincing to conclude that both elements contribute to the combination benefit.

DR. BORER: Dr. Kreisberg?

DR. KREISBERG: My answer is yes. I think the preponderance of the data supports the fact that the combination is better than either one alone.

DR. BORER: Are you as convinced as you would be if we had two randomized controlled trials?

DR. KREISBERG: Well, with the proviso to Dr. Nissen about there are things in medicine that make perfect sense but are absolutely wrong, the answer is yes, I'm satisfied with the evidence.

DR. BORER: Beverly?

DR. LORELL: The answer to 4.2 is yes. I thought the data was very compelling that pravastatin plus buffered aspirin has a greater effect than either buffered aspirin -- either aspirin -- not buffered, but aspirin -- or pravastatin alone.

And for the record, I will say that as a nonstatistician, I cannot feel comfortable answering .1 or .2, but I will say that in this case, there were really two things that I think made this meta-analysis compelling to me as a clinician, not a statistician.

One was that the meta-analysis involved a very large number of patients who were quite well defined.

And the second thing is that I'm always nervous as a nonstatistician when I hear a statistician use a single meta-analysis approach to try and persuade me of something. And I thought it was very valuable in the analysis we heard today that there was an effort to approach this meta-analysis dilemma from three different models.

So, the answer to 4.2 is yes. I can't answer .1 or .2.

DR. BORER: Susanna?

DR. CUNNINGHAM: I would also say that I'm convinced that the pravastatin has a greater effect than aspirin alone, and I also cannot answer the subquestions.

DR. BORER: Yes. I think the question is a little confusing in that it refers in 4.2 to the meta-analyses. In fact, the meta-analyses really I think properly refer to the pravastatin on top of aspirin rather than the aspirin on top of pravastatin, which I think would be hit in 4.3, but I think the answers have referred to both, if that's okay.

And is it okay if we try not to answer 4.3. I don't think any of us other than Tom can --

DR. LIPICKY: Fine.

DR. FLEMING: Just a simple answer. I find the results from the models as qualitatively consistent.

DR. BORER: Okay.

I think for the record, then, everyone has agreed that there is reasonable evidence that both components contribute to the effect of the combination with varying degrees of enthusiasm, perhaps in general, less than would have been the case had there been two randomized controlled trials to look at, each meeting the p less than .05 standard.

Let's go on to 5.0. Upon what basis was the dose of buffered aspirin chosen for use in the fixed-dose combination product? Do you consider this reasonable? What alternative doses can you recommend? And should one wait prior to approval on settling the question of buffered aspirin dose?

Alan?

DR. HIRSCH: Do you want one or all three?

DR. BORER: Just do all three.

DR. HIRSCH: Well, I think the basis of the choice of antiplatelet dose was retrospectively a combination of the primary aspirin trials, the antiplatelet trialists collaboration, and other meta-analyses and obviously the clinical practice that was both valid in the pravastatin trials.

Do I consider this to be reasonable? Absolutely, acceptable and reasonable.

What alternative doses besides the 81 and 325 would I recommend? I wouldn't. Those would be the appropriate doses certainly in the United States market in any case.

And should one wait prior to approval on settling the question of buffered aspirin dose? I think that in this real world, we have adequate data to be happy with those two choices of doses.

DR. BORER: Is anybody unhappy with that answer? Steve is unhappy.

DR. NISSEN: Well, not completely unhappy, but I must point out that there's an enormous meta-analysis, just published within the last few days, from the Oxford Group that shows that there is a higher risk of the 325 milligram dose and that the dose that seemed to have the best combination of safety and efficacy was the 81 to 160 milligram dose. So, it's brand new data. It wasn't available to the sponsor when all this was done, and I haven't had a chance to fully analyze that manuscript, but it ought to be at least mentioned.

DR. TEMPLE: That depends a little bit on whether you want to look at a particular dose that was used in a particular setting or, like the collaborations have to do, lump them all together. The current labeling for aspirin says you can do either of those things. Be my guest. And I think that's what we urge: cover the range of doses that are used. There are some things where 150 is the recommended dose.

DR. HIRSCH: Just to come back to the point, our goal was to make sure that were doses available, not to follow another meta-analysis, another guideline. I think the sponsor has done that.

DR. BORER: I think in general then everybody is happy with 5.0. We know how the doses were chosen. We think it's reasonable, no alternatives to recommend, and with the caveat that was just made by Steve and amended by Alan, we don't think it's necessary to beat this one any further.

But 6.0. Upon what basis was the dose of pravastatin chosen for the use in the fixed-dose combination? Do you consider this reasonable? What alternatives can you recommend, and should one wait prior to approval on settling the question of pravastatin dose?

Keep in mind, in answering this, that putatively this is a convenience product. So, it doesn't mean that everybody has to give this dose.

Alan?

DR. HIRSCH: Yes. I'm collecting my thoughts here.

We spent less time, I thought, than we might have on the discussion of dose, even though we did circulate there. So, this is a question which I would like everyone to weigh in on.

The basis of the choice of dose I presume was a combination of the initial dose-response data the sponsor had, the application of that in the PLAC I, PLAC II, REGRESS, LIPID, and CARE trials.

And do I consider this to be reasonable? Yes, that's reasonable because that's the database we're presented in the meta-analysis.

The hard part is when we get to 6.2 and 6.3 when we're asked what alternatives can you recommend. I suspect there will be some diversity of opinion.

When you lead the question, Jeff, and say it's a convenience dose product, actually there's no need to recommend alternatives. We're asked for a single dose based on the clinical trials for convenience and we can stop. But I do suspect that the panel members will want to discuss the potential for alternative doses in the prava arm, as well as in the aspirin arm, though many may not want to go there.

I'll charge ahead and go to 6.3. Should one wait prior to approval on settling the question of pravastatin dose? I think not, but I bet you there's diversity of opinion. I could justify that if you'd like.

DR. BORER: I think we've said that the choice of the dose was based on the fact that that was the dose that was used in the prevention trials, and it was reasonable.

There could be alternatives. I don't know if we want to discuss this at this point. There could be, but this is being suggested as a convenience product for people who come to the conclusion that this is the dose that ought to be used.

Yes, Beverly.

DR. LORELL: Had there not been the preamble in the text before that question, I would have answered 6.1 yes. I think there's reasonable logic as to why 40 milligrams were chosen. And for the answer to 6.2, what alternatives would you recommend, I would say none.

But I am concerned that in fact in the text preamble in paragraph 2 that you read was the comment that generally that means that a full range of dosing strengths of each individual entity should be available for the combination product, thereby providing convenience, but not influencing selection of dosing or dosing regimens of individual entities. So, with that preamble to guide us as members of the committee, I would say 6.2 probably does merit some discussion, and the alternative I would recommend would be including consideration of also 80 milligrams with the two options for aspirin. I don't understand the logic for offering a range of aspirin and not offering a range of titration of Pravachol, unless it is the intention to argue that it doesn't matter what your LDL is.

DR. BORER: That issue that you just so beautifully outlined is the sum and substance of question 8.0, and so perhaps, with your permission, we'll wait until 8.0 and discuss that more fully because I think this is really, so to speak, the heart of the matter. Is that okay, Ray?

DR. LIPICKY: Sure.

DR. BORER: 7.0. Assuming that you have concluded something about the strength of evidence that pravastatin and buffered aspirin should be taken together and that the doses to be available in the fixed-dose combination product are appropriate, what is the strength of evidence that a fixed-dose combination product, taking a single pill, has increased clinical benefit with respect to taking two pills not necessarily together?

To clarify that further, should we require better demonstration of additional benefit provided by convenience, and what kind of demonstration would be better?

Alan, do you want to start out?

DR. TEMPLE: Jeffrey, we really have never asked people to show that. It doesn't mean we couldn't change our view, so it's worth listening. It's my belief that it would not be easy to do in a controlled trial setting where people tend to be compliant. You'd have to establish so loose a setting that people could just ignore the drugs they're supposed to take. I wouldn't say that's not possible, but there's not a lot of track record on it.

DR. HIRSCH: But you asked the question.

DR. TEMPLE: Yes. I probably tried to cross it out.

(Laughter.)

DR. LIPICKY: Yes. It has to be put in context. The assumption was you were going to come to different conclusions in some of the questions above than you did and that then you would not be able to assert that you knew that the two ingredients contributed to the effect or that you had more major reservations than you did.

So, the crux of the argument, in part, that was made by the company was that compliance was the benefit here, that this fixed-dose combination -- they didn't think you'd reach the conclusion that they had shown A plus B is better than A or B -- offered compliance advantages. And there was absolutely no data regarding compliance at all for this fixed-dose combination. So, this question was written to find out whether you want data to support your judgments. It probably is out of place now since you answered the questions above in a different fashion.

DR. HIRSCH: I think it's still in place.

DR. TEMPLE: Well, it's still pertinent to the end game where you're going to worry about the fact that they may not know how to stop it properly before surgery, a perfectly legitimate question, and maybe the potential for better compliance --

DR. LIPICKY: Well, okay, fine.

DR. TEMPLE: -- certainly not the documentation of good compliance might be part of what you think in it. But ordinarily in other senses we don't really ask that question.

DR. LIPICKY: But in fact I guess the presentation disturbed me because what was shown for this general knowledge that combinations are better for compliance was a total of four trials. So, I don't think it is general knowledge or that you can assume that taking one pill instead of two leads to better compliance. And so, it's sort of pertinent to the question of if that were important, what do you think we should have seen.

DR. HIRSCH: Can I rephrase the question a little bit because I think this does merit some discussion based on everything I've heard from the panel? I think the lead-in to this whole discussion was obviously convenience and compliance. So, the question I'd phrase is, when does perceived convenience, really driven I think by patient demands we all hear, actually work for or against a perceived or demonstrated, I should say, health benefit? In other words, does having a convenience product per se, which might potentially improve compliance with one or two drugs, work for or against hitting the endpoint ultimately of a decreased cardiovascular risk?

DR. LIPICKY: I guess the discussion should be would you require seeing data that there is something to that. The question is totally out of context now. So, the subparts are: would compliance data be enough, or would you need to have body counts?

DR. HIRSCH: So, just as my friend Tom occasionally tells us how we might best think about the statistical considerations, I'd like to just throw some ideas out there for the panel to consider because the sponsor did and I found them intriguing.

There are a lot of hypotheses about why this might be a good thing, and the ones I listed were things like potentially adherence to guidelines because people would actually take their aspirin, for example; perhaps use of the correct dose because it would be formulated correctly; perhaps a decreased pill burden and again a greater daily compliance. There were many others, but I think all of these are basically conjecture at this point. There is not adequate data, having read those primary references as well, to suggest that we achieve that in this particular population with these particular products.

So, with that in mind, when I thought of these questions, I actually did think that we're going to be faced with many potential combinations. From this very data set, we could look at beta-blocker pravastatin in the future. We could look at "prilstatin." This is a large data set. There are many combinations that would be very beneficial.

But before I'm having to face this as a panel member, I actually would like to see some additional data, and I was thinking of a compliance study, Ray, but that may be unreasonable.

What does everyone else think?

DR. BORER: Are there any other opinions? Dr. Kreisberg?

DR. KREISBERG: Well, there's another element to this and that is cost, which we haven't discussed and maybe we can't discuss. But aspirin is dirt cheap and it's about a penny a day for those people that take it. And we're talking about a combination now, and when you consider that 80 percent of the deaths occur in people over the age of 65 and they have Medicare, the payments that they have to make for these medications become crucially important.

So, one of the questions that I have in my mind is what is the intent of the sponsor with regard to this preparation because if it turns out that a drug like pravastatin will, in the near future, will be a generic and will be allowable with a $10 co-payment through most health plans, but this product is not a generic and it requires a $25 co-payment or a $35 co-payment with the plan, then what is that going to do for the proposed adherence rate that we're contemplating would be of benefit to this type of combination?

DR. TEMPLE: It's not that that's not a perfectly cogent question, but I think we don't consider it FDA's province to do that.

DR. BORER: Should we require demonstration of additional benefit from a convenience study? I don't know that we need to spend a lot of time on this, but does anybody have any thoughts that they want to share here? Blase?

DR. CARABELLO: I don't think we should because I think that ultimately the sponsor bears that burden. If it turns out that the drug is easier to take and that people like it and use it more, then it will be used. And in fact, if that's not the truth, then the drug will die on the vine and it won't be a consideration.

DR. BORER: We don't have compliance data, and I think speculating about it probably isn't going to be very useful at this point.

But let's get on. The last two questions I think are where the action is.

8.0. How likely is it that the availability of a fixed-dose combination product would encourage inappropriate use of the doses of any of these drugs?

This was the issue that Beverly was getting at earlier and it's what we danced around all day. We heard data from Dr. Pedersen and from the sponsor, and I think this is where we really want to concentrate our discussion.

Alan?

DR. HIRSCH: Do you want us to take these one by one or again the whole packet?

DR. BORER: Take them together.

DR. HIRSCH: Inappropriate use of buffered aspirin for primary prevention. I think the risk is very low.

Inappropriate use of a dose of 40 milligrams of pravastatin. Actually low.

Inappropriate use of a dose of 325 milligrams buffered aspirin. I think equally low.

And the same thing for 8.4.

DR. BORER: I'm sorry. We were given these 8.1 through 8.4. I want to expand a little bit.

DR. HIRSCH: Okay.

DR. BORER: Inappropriate use of a dose of pravastatin. I don't care what the dose is. If only one dose is offered, is it likely that the practice pattern will be that the drug is not used in the way that it otherwise might be used?

DR. HIRSCH: Yes, I think there's a real chance of that.

DR. BORER: Does anybody else want to talk about that? Steve?

DR. NISSEN: Yes. I'm troubled by this, and let me see if I can help.

First of all, I do think that there is a moderate risk of inappropriate use of buffered aspirin, and I think Susanna really was the first to point this up. And I hadn't really thought about it, but the more I've thought about it and heard from other people, I am concerned that people undergoing both minor and major surgical procedures may accidentally -- much more likely accidentally -- be given aspirin as part of a fixed-dose combination.

And what's important for us to understand is that aspirin is not a completely benign drug, that it has very serious consequences in the wrong circumstances. Therefore, when you put it together in a fixed-dose combination, I do think you increase the likelihood that either the patient or the physician will be unaware of the fact that they're taking a potent antiplatelet agent and that someone will forget about that in a circumstance where the patient may be harmed. So, I've got to give that at least some credence.

Similarly, because the dose of 40 milligrams of pravastatin, according to current guidelines -- this speaks to medical practice -- is unlikely to get, in my opinion, the majority of patients to the recommended goals, then I think that encouragement of use of this fixed-dose combination will, in fact, increase the probability that some patients will be undertreated with respect to their lipids.

So, I would also say -- and we can skip the dose issues -- that both for pravastatin and for aspirin, there is moderate risk here that the agents will be given inappropriately when here in the fixed-dose combination. And the more relevant issue then is how does that risk equate with the benefit that might accrue from this combination, and I will speak to that a little bit later.

DR. BORER: Beverly?

DR. LORELL: I agree with what was just said.

DR. BORER: Mike, go ahead.

DR. ARTMAN: I would just like to get back to the issue of inappropriate use of buffered aspirin for primary prevention, and I raised it earlier and got pooh-poohed a little bit. But I disagree with Alan. I really think that there is that risk, and I think there's a lot of controversy about the use of aspirin for primary prevention. I don't think it's been proven, and I think if this fixed-dose is approved, I can see a big campaign and the detail people talking up this combination for secondary prevention, and oh, by the way, you know pravastatin is approved for primary prevention as well. And I think there's going to be a lot of leak and a lot of bleed over to that, and that concerns me as well as these other issues that Steve and Beverly have raised as well.

DR. BORER: I agree with what everyone has said here, and I would like to add just one additional point. I think that over and above the other aspirin issues that have been raised, aspirin may not be the appropriate drug for every patient who requires a platelet active agent. Clopidogrel is approved for certain, specific situations. Most people who receive an antiplatelet drug like aspirin or clopidogrel require or should receive or do receive lipid-modifying therapy in the form of statins. If a combination product is available that has aspirin attached to it, I'm concerned that in some, admittedly small, segment of the population, the more appropriate drug, which might be clopidogrel, won't be used in favor of the less appropriate drug because of the convenience of giving the aspirin together with the otherwise necessary statin. So, I would just add that to the mix, but other than that, I agree with what's been said here.

Blase?

DR. CARABELLO: Just to reiterate what I said earlier, we have no idea what the sudden withdrawal of the statin agent prior to surgery might do, and so one can easily foresee the cumbersome nature of withdrawing the aspirin part of the product and continuing the statin part of the product by prescribing a single agent. I doubt that anybody would do that. And while I have no knowledge that the withdrawal of statins might be dangerous, it's open to question. So, that does concern me.

DR. BORER: Just to put this in proper perspective without moving beyond where we are here, right now the sponsor is proposing co-packaging, in which case the issue of stopping aspirin independent of pravastatin wouldn't be a big issue, it wouldn't be a big deal. But you asked us to consider a pill that has both of them in it. You didn't tell us about the form of the pill. Is the aspirin part something that could be broken off? Or are we going beyond, in talking about that, where we should be going?

DR. LIPICKY: I can't answer the question. Clearly aspirin is not going to be in half and prava in the other in your single tablet.

DR. FIEDOREK: Yes. We're still working on the formulation for that, but it will be a combination tablet.

DR. LIPICKY: But you're not going to put aspirin in one half and prava in the other.

DR. FIEDOREK: No.

DR. BORER: No, okay.

So, I think the general sense here is that we have real concern about the range of doses that's available for this product because of the various reasons that have been raised, because of the likelihood that this selection, this range that's been offered, will potentially adversely affect clinical practice.

DR. LORELL: I think that there was also the separate concern raised about packaging a potent antiplatelet agent with another drug.

DR. TEMPLE: We need to understand this because these are probably the reasons you may give a particular opinion. So, it would be very helpful if we understood those. Let me tell you what I understand.

You haven't addressed the question of whether labeling could overcome this and you might want to think about that.

But one major concern was that having two together really makes it difficult to stop one of them, and you can think of quite bad consequences if people don't realize they're supposed to stop their aspirin prior to surgery and you're not sure that stopping the whole combination is the right thing to do. So, I understand that part. That's pretty clear.

I'm a little foggy on the dose thing, unless you just don't believe people will do it, which might be your explanation. This is going to be labeled as if 40 milligrams is the right dose, you can add them together. So, you must suspect that that will not, in fact, happen.

DR. BORER: Yes.

DR. TEMPLE: I assume that's the reason.

DR. BORER: I think from the discussion that we've had around the table during the day, that would be the presumption.

DR. TEMPLE: Just let me continue. The first one, you might think of ways to label around that, but you'd be suspicious about whether they'd work, I imagine.

This one, is this susceptible to appropriate labeling injunctions, you know, be sure you get the right dose, not everybody needs 40. Or is that just not a possibility?

DR. BORER: We can ask everyone around the table their opinion. My opinion is that the label really won't mitigate that potential problem.

Steve?

DR. NISSEN: Bob, the evidence -- and again, this is an area that I happen to be an expert in -- is that most patients end up on the dose of statin that they're started on, that unfortunately, despite all of our efforts to get people to titrate, they tend not to titrate. So, my concern is that to the extent that this will happen, the inconvenience to the physician and the patient of having to stop the combination, switch to a different statin and then co-administer aspirin will be enough of an impediment that more patients will not be titrated to goal than would be titrated to goal if these agents weren't available in a fixed-dose combination.

DR. TEMPLE: Now, how is that different from starting on 40, which is what the recommended starting dose is, and having stopped that and go buy the 80 milligrams or take two of them or something? You think this makes it materially worse because stopping the aspirin and going to the drug store and getting a few aspirin is really hard.

DR. NISSEN: Yes, I do.

DR. BORER: And I would agree with that, and I'd even amplify on it but I don't think I have to.

Tom?

DR. FLEMING: I'd like to probe on this a bit more as well, maybe somewhat along the lines of what Bob was just asking.

Quite frankly, I'm inclined to vote yes in question 9 unless there is a substantial reason given in question 8 to there being a concern with this packaging, and in the context of question 7, the way I've been thinking about this is what's this all about. Why are we considering this packaging?

It needs to be at least reasonably well motivated that there will be an enhanced likelihood of achieving the most favorable benefit-to-risk profile that these agents are capable of providing should you have a more optimal level of adherence. There needs to be a perception or a reality, better than that, that we have effective agents here that are being underused and as a result, globally we're not achieving the level of clinical benefit that they could potentially provide. If that's not true, we're done.

Now, if in fact that's true, and presumably that would be argued to be true in the sense that when you have a packaged delivery, you're going to make it more convenient and you're going to have a higher level of consistency and accuracy in the delivery of these doses to a larger fraction of the population than currently is getting access to this. That's the pro argument and we've heard that from the sponsor.

The con argument that we're hearing that also seems to be relevant is there's a "yes but." There are circumstances in which this packaged delivery can be nonoptimal, examples being there are certain types of patients that shouldn't be getting antiplatelet therapy and it may be, in fact, not recognized as readily when they are taking a single tablet, or maybe they're getting a nonoptimal antiplatelet agent. Or it may be that we're not achieving an adequate level of targeted goal of lipid reduction and it may become less likely that we would have an altered dose or altered selection if people are choosing this. I see all of those as relevant concerns, but it is true, somewhat, that when you label things, or at least when you make it as clear as you can that this intervention, this particular package combination isn't necessarily the right choice for all people, that some of these concerns can be offset by intelligent use of this.

So, what I'm left with in the end is the subjective weighing out of what are the benefits that we're going to get when we're using it correctly and it's going to give an enhanced, widespread use of agents that have potential of delivering benefit but they're being underutilized against these instances that we've given where there are concerns. The fact that there are concerns doesn't mean that the benefit-to-risk, so to speak, is negative.

What I'd like to get a sense of from the committee is, is there really an unmet need that's sufficiently substantial such that if this is more convenient, we really will achieve a higher level of adherence and thereby a better benefit-to-risk in the broadest population, but that there are these concerns and in fact they're real but they won't counterbalance the benefit, or are these concerns so real that they will give us no net gain? That's the answer I'd like to hear.

DR. BORER: Tom, let me begin responding to you by pointing out something. You said get these agents to people. I don't believe we're talking about improving compliance with two agents. I think we're talking about compliance with one agent. The frequency of prescription of statins I don't think is going to be altered one bit by making this combination available. The associated use of aspirin may well be increased. I mean, it may or it may not be. I don't know, but it might be. I don't think the use of statins is going to be altered at all. So, it may be that we might alter the use of aspirin by providing the combination.

I think, though -- and I'd like to hear from everybody else to respond to Tom as well -- that the sense of the people who have spoken, about this among the voting members of the committee, is that perfect is the enemy of good, and we don't know what perfect is, and being on some statin, if you have any of the conditions for which the drug is indicated, is better than being on no statin. We're not going to alter the statin part of this equation here.

What we are going to do, however, we believe, by limiting the dose range of the combination, is making it very likely that people will use this dose even if, left unfettered, their judgment would cause them to use different doses for targets that they may wish to achieve based on clinical guidelines or whatever else. And it's not that they couldn't make the change, but that it becomes so inconvenient and costly to make the change that they're unlikely to do it and that the patients may not benefit from that.

Now, Beverly and Steve and the others here, maybe you may want to add to that.

DR. NISSEN: Well, I think you said it very well. Tom, let me see if I can help you with this.

There are two sides to this equation. Will the combination improve safety or make safety more of a concern? We're talking about weighing, pluses and minuses. I think for that one it's pretty clear to me that the likelihood that people will get aspirin inappropriately and therefore raise a safety problem goes up not down. I think the likelihood that a primary prevention patient who shouldn't get aspirin will and the chances that a patient will get aspirin when having a surgical procedure that shouldn't will are both worse with the combination. So, on the safety side, it's not a plus.

What about efficacy? Efficacy would mean that more patients that ought to have co-administered aspirin would and that more patients would be more likely to get to those lipid targets that we've spent the last 7 or 10 years trying to get people to follow. And I have concerns that the net balance on the efficacy side will be worse. Yes, a few more patients may get aspirin, but the burden of then having to go through this process of stopping a combination product, switching to a different statin, and co-administering aspirin will be enough of an impediment in primary care and in other settings that fewer patients will actually get to the goals that we want, and so therefore, this combination will have an adverse effect on overall efficacy.

So, I think for me it loses on the safety side and it loses on the efficacy side.

DR. THOMPSON: Jeff, I'm a non-voting member, but can I comment?

DR. BORER: Yes, sure, Paul.

DR. THOMPSON: Because I'd like to make two comments.

First of all, I always tell people that there are two devils in this world. There's the devil you know and there's the devil you can speculate about. When we look at the devil we know and we look at slide D-13, we see that among those patients -- now, this is 1999 data -- with known cardiovascular disease, only 51 percent reported that they were taking aspirin or an equivalent. So, that's a large gap. And we also know that 15 percent of those who said they were taking aspirin were taking something else. They were taking a non-aspirin compound. So, I think we're actually speculating a fair amount about things that are possible, devils that we're actually imagining, as opposed to facing devils that we can see and have data in.

Now, a couple other small comments. We're talking a lot about the safety of withdrawing a statin, for example. I do that in my lipid clinic all the time because people always doubt whether the statin is responsible for their effect. So, I stop it to show them that indeed the statin was effective.

We're also talking about the negative stuff of aspirin in primary preventions. Well, it's news to me. It's one of the few things I do for myself. I think most cardiologists do take aspirin for primary prevention. So, perhaps there's some data that wasn't presented that I didn't know.

Then finally, what I always do and I tell my children to do when I have to make a decision is I make up a list of pros and cons, and I'm glad to share that with you, if you'd like. But I've listened to the cons during the day that aspirin might be taken by those who are at risk to bleed, and there are all those sorts of risks. But when you take and you make up your benefit-risk thing, even with the fact that convenience for a patient is a benefit -- I thought this was actually a slam dunk when I read this stuff at home before coming here. I had no idea.

But when you look at the pros and the cons, I find it hard to worry so much about these devils unknown as opposed to the devils we actually see in the data and the efficacy of the combination shown from the meta-analysis, which we all agreed that we liked.

I'll go through my pros and cons, but I don't want to take more time. I'm a nonvoting member, but I've listed to everything such as that you lack flexibility. But that lack of flexibility with a 40 milligram dose may get doctors start with 40 milligrams as opposed to 20 milligrams. So, for most of the things we've talked about as a negative, there's also the potential benefit.

No data on aspirin effect when given together, but actually we do have some data. It looks like it decreases the event rates. And this was the stuff on platelets. In other words, people were talking about the fact that there was no effect on the platelet effect. Well, there's no in vitro studies, but there are in vivo studies showing that that combination looked better. Now, if you give me a choice between some Bademan's platelet chamber versus some human being, I'm going to take the human being studies.

So, I've listened to the two sides of it, but I'm surprised a little bit.

DR. BORER: Tom, did you have some comments here?

DR. FLEMING: Jeff and Steve, I wanted to just follow up for clarification on your insights.

I understand, Jeff, from what you're saying is that you see the most substantive impact of this would be potentially increased adherence to the aspirin, that the lipid-lowering intervention will essentially, in terms of its overall coverage, largely stay the same.

DR. BORER: Yes. Anything else would have to be a result of marketing, and I don't think we can make any guess about the impact of marketing, how it's going to be done, or whatever. But logically there should be no impact on the use of the statin. It really should be only on the impact of the use of the aspirin.

DR. FLEMING: So, if I go with that, Steve, what I heard was a concern in terms of the safety side in particular being that if there isn't an as flexible an adjustment of the regimen for a lipid-lowering agent, that could have adverse effects on efficacy and safety.

DR. NISSEN: No, I'm sorry. The safety issue related to the aspirin. I'm very comfortable with the safety of pravastatin and, frankly, the other statins as well.

DR. FLEMING: All right. So, basically it would be a concern about potential loss of efficacy on the lipid-lowering side, the statin side. When we see data, as we were just referred to, D-13, and we see estimates of about 50 percent underuse of aspirin, even though I appreciate that there are settings you've described where aspirin use could be inappropriate and hence larger use could lead to then some instances of adverse effects, if it truly is 50 percent, isn't the up side for increased adherence to aspirin far greater than the down side?

DR. BORER: I don't know. I really don't know. First of all, we have no idea what the impact on aspirin use would be. We heard that in Sweden there wasn't much impact, not that those are primary data, and yes, this is a convenience product.

DR. FLEMING: I am persuaded by what was said earlier and that is if you make this available and it's not in fact used, then that's a marketing issue. That's not an issue as to whether this had the potential of providing benefit. In fact, there's no harm, in a sense, done if you make this available and no one uses it. We just didn't provide any benefit.

My concern is if we make this available and it is used, is this in fact going to achieve more net benefit than harm, and the benefit would allegedly be, if I focus on the way you think this is going to go, Jeff, it has the potential at least of substantially increasing adherence to aspirin which, if we believe the efficacy data, has the potential of providing meaningful benefit on major events. But then there is the offsetting aspect that in some of these cases, there may be some safety concerns, and in the lipid-lowering case, there may be lack of as effective adjustments that could provide some offsetting.

So, is that the benefit-to-risk that we would have to say when we --

DR. BORER: As far as I'm concerned, you've pretty much noted the elements, and I would go one step further. I'd be a lot happier, because I think the problem would be obviated, if the sponsor provided pravastatin at several doses with aspirin attached to it so that I could give the dose that I want to give.

DR. NISSEN: We don't know what percent of that 50 percent not on aspirin also require co-administration of statins. You understand that that data is not statin data. That's aspirin data. So, there's an unknown here, and that is how many of those people would be even eligible for a combination product, and we have not been provided with any information about that.

DR. FLEMING: A very important point. Is there anybody that has that data? If this 50 percent doesn't apply to our population and we need to know what percent of the people on statins are not using aspirin, which is very relevant.

DR. ARTMAN: Just go back to D-10, the sponsor showed us data from 167,000 patients who had had an MI at discharge, and 77 percent of them were receiving aspirin and only 37 percent were receiving statins. So, I don't see this as an underutilization of aspirin. I see it as an underutilization of statins. And I think the sponsor is probably not going to go out promoting aspirin. They're going to go out promoting pravastatin.

DR. BORER: Which they could do anyway.

DR. ARTMAN: Which they could do anyway.

You know, this was prefaced to us with this treatment gap and it's a public health issue, and it's not because the drugs aren't available. These drugs are available. It's educating the physicians and, these days, educating the consumers directly. So I think, Tom, this maybe is more reflective of the proper patient base.

DR. FLEMING: I was going to throw something out here because part of where I'm struggling here is I prefer to be looking at quantitative results and we're having to speculate and that's somewhat unavoidable.

We had data presented to us which wasn't certainly as reliable as what we would have had from randomized trials, but it's the best evidence at least that I have that I can use in a subjective judgment here that says you're going to have on the order of 25 to 33 percent reduction in major events with proper administration of these interventions. If in fact making an intervention that's more convenient does allow for a 25 percent -- and that's a pretty substantial increase -- but a 25 percent increase in adherence from 50 to 75 or whatever, that leads to what one could at least speculate a pretty substantial reduction in the overall rate of occurrence of some of these events.

On the other hand, there are these down sides that occur that relate to safety. It has to be a pretty strong argument I would think, though, that those down sides would be so substantial that they could offset this potential for a very substantial gain by having greater adherence to effective agents.

DR. BORER: To preempt Beverly for just one second here and to try to answer your question, I can't answer the question.

As it was presented, I recognized the same concern that Steve just verbalized, that is, that we were shown the gap for all aspirin users. If you look at the data in the book from LIPID and CARE -- it may not be representative but that's what we've got -- pages 14 and 16 in the briefing book, patients were randomized to pravastatin or placebo. They may have been on aspirin. 83 percent of the patients in LIPID were on aspirin; 84 percent of the patients in CARE were on aspirin. That's a gap of 16 to 17 percent, not 49 percent.

And I'll tell you my opinion. I can't substantiate it. My opinion. Once somebody is taking a statin drug, their perception of the importance of drug-taking is heightened, and I think it would be more likely rather than less likely that they would take their concomitant aspirin. I don't know if that's true or not, but that's the way I would interpret these data.

DR. FLEMING: Let me reiterate my question then from five minutes ago, which is if it's appropriate to suggest that this 50 percent underuse of aspirin isn't relevant to our context, we need to know, in the context of somebody who would be on lipid-lowering agents, what is our best estimate of that. I understand, from the clinical trials for aspirin, when you have people joining a clinical trial to be randomized to a lipid-lowering agent, 80 percent were using aspirin, but does that represent the real world, which was my question. And then we had reference to data from D-10.

DR. HIRSCH: Tom, you're not going to get an answer to your question. There are various surveys with various aspirin use rates in different populations, and you can quote figures from 50 to 90 percent. Somewhere between 70 and 80 percent is probably accurate. And you're not going to be able to make your decision based on that. Although you'd like that hard data, sorry, you're not going to get it today.

And I'm going to charge in and make an impassioned plea here. I'm someone, Tom, who also deals with this public health issue of use or not use and filling the gap in underuse of medications. I'm a little surprised at us today. I'm a little surprised that we're going to try to make a decision at this last minute to try to fill this gap with this single product combination because any hope of us changing the outcome of use of aspirin and lipid-lowering agents in America with this decision -- that happens with very long, large-scale NIH, American Heart Association 5- and 10-year, Healthy People 2010 plans. It doesn't happen in a deliberation like this.

I would like to constrain the conversation a little bit and bring it back down to, again like you were saying earlier, is there some evidence of efficacy, what is the safety issue, and do we think physicians can prescribe more or less appropriately, never perfectly.

And with that, Bev, go ahead.

DR. LORELL: Thank you.

I think that in thinking about the data that we've seen today, we've seen very compelling data from the meta-analysis that the use of Pravachol plus aspirin is efficacious. But I think a very interesting point is that the use of aspirin in the trial data that we have to look at to make our decision was discretionary. It was a high frequency of use that we saw, but it was discretionary and what people were really doing in those trials was taking two separate pills.

I am concerned about the issue that having a single fixed dose carries a risk, as was put well by one of my colleagues, of increasing the likelihood of under-treatment with statins, but I think that my much larger concern that I had not even thought about until I heard the discussions today, is the very real risks of packaging a potent antiplatelet agent with something else. And I would go on record as saying that I have grave concern about packaging aspirin, clopidogrel, or any other antiplatelet agent with anything, whether it be a beta-blocker, an ACE inhibitor, you name it. I use these drugs every day in very high-risk patients. Aspirin is a powerful drug. It carries risks of bleeding. Patients and doctors have to be extremely attentive to its use and the subtle nuances of when to temporarily stop and when to restart it. So, I think my largest concern is I don't think the issue of convenience is outweighed by the very real risk and lack of data about packaging a potent antiplatelet with something else. To me that's a very major issue.

DR. BORER: Blase, do you have any last comment here?

DR. CARABELLO: No.

DR. BORER: Then let's move on to question 9. We'll go around the table because this is a vote that has to be counted, and we have to have a verbal statement.

Do you recommend approval of a fixed-dose combination product of pravastatin plus buffered aspirin -- I assume the product that we've been given, which is the doses -- yes, right?

DR. LIPICKY: Yes.

DR. TEMPLE: Jeffrey, one thing. We didn't think of this. You ought to address the question of whether it would be different if they made it available with two doses of pravastatin, which they may not want to do, but which is certainly possible.

DR. BORER: Alan, why don't you start?

DR. HIRSCH: Am I'm allowed, Bob, to answer your first question first? I mean your dosing question. Say yes.

DR. LIPICKY: [Inaudible.]

DR. BORER: Why don't we answer this one first? Then we'll go around the room and ask the second one.

DR. TEMPLE: That would be okay. Would your answer be different if they had an 80?

DR. LIPICKY: If you say no, then you can ask what if you had two doses of pravastatin, would that be good?

DR. HIRSCH: I'll need to explain later. I'll say no.

DR. BORER: Let's start with Dr. Kreisberg.

DR. KREISBERG: No.

DR. LORELL: No.

DR. BORER: That was a no from Beverly.

Susanna?

DR. CUNNINGHAM: No. I'd really like to have the information that allowed me to vote differently. I would like to know that this was safe and that people wouldn't be put in trouble, and I'd like to know that this actually helped people to take their drugs more. We don't have that information. I'd like to have something that says if it were available as a combined, one, people would take it more often. I don't have that. I hate having to guess, and so I'm going to say no.

DR. BORER: Mike?

DR. ARTMAN: No.

DR. BORER: Tom?

(Pause.)

DR. KREISBERG: Yes or no.

(Laughter.)

DR. FLEMING: Well, there is a third and I'll take that option. I'm to abstain from this. I actually was inclined to vote yes, and I've been persuaded by a lot of comments over the last 20 minutes that this is extremely difficult.

Basically my sense is that if it is, as I would have thought, much more plausible than I'm hearing, that we really would enhance the overall adherence, then I would think that the potential for benefit exceeds the very real and relevant concerns. But what I'm hearing from my clinical colleagues, who have much better insight than I do about the actual real-world reality of this, is it's not very likely we're going to have enhanced adherence, and if that's the case, then I don't believe that it would be a positive step. So, in the absence of being able to make an informed judgment about that, I think I'm going to abstain.

DR. BORER: Blase?

DR. CARABELLO: Yes.

DR. BORER: Steve?

DR. NISSEN: No.

DR. BORER: And I'll vote no. So, we have 7 no, 1 yes, 1 abstain.

Now let's go on to the next issue, which is the one that Bob just asked, and I'm going to change the question just a little bit. What if we had not two doses but a broader range of pravastatin doses? I don't want to pick the doses, but what if we had a broader range of pravastatin doses available together with aspirin? Would that change the vote?

Alan, why don't you start?

DR. HIRSCH: Yes, and the reason is because pravastatin is safe and effective, aspirin is effective and rather safe. Actually the biobehavioral issues of how patients take drugs and physicians prescribe them are the main ghost in the room. That's what this is all about. For that, I think with a dose range, I would certainly have voted yes.

DR. KREISBERG: I'll vote no and the reason that I'll vote no is that the only alternate dose is 80 milligrams, at least that's what I'm hearing. And that adds an additional 3 percent LDL cholesterol lowering, and consequently I don't think that's enough of a therapeutic option to make that an attractive one. So, my answer is no.

DR. BORER: Beverly?

DR. LORELL: I would answer no too. I think it still does not address the issue of potential increasing the probability of undertreatment. We were shown no data addressing that.

And secondly, I still have a major concern about combining, in a fixed-dose combination, with any other drug an antiplatelet agent.

DR. BORER: Susanna?

DR. CUNNINGHAM: No, because it doesn't change the issues I'm concerned with.

DR. BORER: Mike?

DR. ARTMAN: Yes, I would say no as well. I think that again as sort of a clinical pharmacologist, this really goes against one of the principles of rational drug therapy and that is to try and tailor your therapy to each individual patient. And that may change on a month-to-month basis in that given patient.

The other issue is related to the compliance issue. For those who have tried to study compliance in populations who are taking a medication that really doesn't make them feel any different, compliance is a very difficult issue. It's very complicated. And the inference is that just because this patient who is taking four or five drugs now will take three or four drugs instead of four or five will have better adherence to their regimen I think is just a lot of hand-waving. There are absolutely no data to support that, and I think this is the wrong way to go.

DR. BORER: Tom, you already abstained from number one. Do you want to abstain from two?

DR. FLEMING: Yes.

DR. BORER: And Blase, you said yes the first time. So, that's not an issue.

Steve?

DR. NISSEN: No.

DR. BORER: Alan? I'm sorry. Alan already said --

DR. HIRSCH: I did but I could change my mind if you like.

(Laughter.)

DR. BORER: No, no.

I have a very difficult time with this question. I think that it would be reasonable to approve a combination of these two products as a convenience as long as the uncombined drugs remained available and as long as a range of doses and combinations sufficient to suit my purposes as a prescriber were met.

I agree fully with what Bev said. I am concerned about potential problems associated with adding an antiplatelet drug that might not be used appropriately because it's attached to something that the patient may not remember about. But I think I might be able to deal with that or someone might be able to deal with that in certain situations. And the convenience of having that available might outweigh my concern in some situations. What I was not happy about with the drug that was proposed was that it limited my capacity to manipulate cholesterol the way I might choose to do it.

So, while I still have the safety concerns that have been raised here, and I agree fully with Mike's comment that I don't think the right way to give drugs is in fixed combinations, that you should titrate individually and be able to change doses individually, nonetheless I think the convenience of having the two pills in one would be reasonable and would be sufficient to outweigh my concerns about safety for approvability as long as my capacity to prescribe the doses I want to prescribe wasn't limited. So, I would vote sort of a tepid yes on this last question.

So, I don't think we have to take a big tally on that one. It's not one of the official questions here.

DR. TEMPLE: But you did take a tally, didn't you?

DR. BORER: Well, we have everybody's answer. Do you want me to summarize it for you? We have 3 yeses, 5 noes, and 1 abstain.

DR. TEMPLE: It's late to point out what I'm going to point out, and we obviously should have built it into the questions. But you may recall that aspirin is available as an over-the-counter drug and whether people take it and under what circumstances they take it and whether their doctor knows they're taking is is not exactly a known quantity at the present time. I don't know if that makes any difference, and we really didn't mention it. But that seems to have something to do with how much anxiety one should have about someone taking aspirin when he shouldn't. I mean, they already are.

DR. BORER: Not to me, for the reasons that Beverly verbalized very well.

I think that we've completed our responses. We'll call the meeting adjourned.

(Whereupon, at 3:05 p.m., the committee was adjourned.)