UNITED STATES OF AMERICA

+ + + + +

FOOD AND DRUG ADMINISTRATION

+ + + + +

CENTER FOR DRUG EVALUATION AND RESEARCH

+ + + + +

CARDIOVASCULAR AND RENAL DRUGS

ADVISORY COMMITTEE

+ + + + +

94th MEETING

+ + + + +

THURSDAY,

OCTOBER 11, 2001

+ + + + +

The Advisory Committee met in Building 10, Clinical Center, Jack Masur Auditorium, National Institutes of Health, 9000 Rockville Pike, Bethesda, Maryland, at 9:00 a.m., Jeffrey Borer, M.D., Acting Chairman, presiding.

PRESENT:

JEFFREY BORER, M.D., Acting Chairman

JOAN C. STANDAERT, Executive Secretary

PAUL ARMSTRONG, M.D., Member

MICHAEL F. ARTMAN, M.D., Member

PRESENT: (cont.)

THOMAS FLEMING, Ph.D., Member

ALAN T. HIRSCH, M.D., Member

JoANN LINDENFELD, M.D., Member

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

GLORIA ANDERSON, Ph.D., Voting SGE Consultant

RAYMOND LIPICKY, FDA

ALSO PRESENT:

PETER CARSON, M.D. (Novartis)

JAY N. COHN, M.D. (Novartis)

LLOYD FISHER, Ph.D. (Novartis)

ROBERT GLAZER, M.D. (Novartis)

MATHIAS HUKKELHOVEN, Ph.D. (Novartis)

JAMES HUNG (FDA)

MALCOLM MACNAB (Novartis)

SHERI TARGUM (FDA)

C-O-N-T-E-N-T-S

PAGE

Conflict of Interest Statement 4

Introduction, Acting Chairman Borer 6

NDA 20-665/S-016 and NDA 21-283/S-001, Diovan

(valsartan), Novartis Pharmaceuticals Corporation:

Company Presentation:

Introduction, Mathias Hukkelhoven, Ph.D. 7

Clinical Efficacy Data, John N. Cohn,

M.D. 10

Safety Data, Robert Glazer, M.D. 112

Benefit-Risk Discussion, Jay N. Cohn, M.D. 130

Committee Reviewer, Thomas Fleming, Ph.D. 158

Committee Review and Discussion

P-R-O-C-E-E-D-I-N-G-S

(8:58 a.m.)

ACTING CHAIRMAN BORER: I'd like to call this meeting to order.

This is the 94th meeting of the Cardiovascular and Renal Drugs Advisory Committee.

We have a conflict of interest statement to be presented by Joan Standaert, and then I have a couple of opening comments about the format today.

Joan.

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

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

In accordance with 18 USC 208(b)(3), a full waiver has been granted to Dr. Thomas R. Fleming. A copy of this waiver statement may be obtained by submitting a written request to the agency's Freedom of Information Office, Room 12A-30, Parklawn Building.

In addition, we would like to disclose for the record that Dr. Paul W. Armstrong has an interest which does not constitute a financial interest within the meaning of 18 USC 208(a), but which could create the appearance of a conflict.

The agency has determined notwithstanding this interest that the interest of the government in his participation outweighs the concern that the integrity of the agency's programs and operations may be compromised.

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

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

That concludes the conflict of interest statement for October the 11th.

ACTING CHAIRMAN BORER: Okay. I'm going to first ask if there are any comments from the public. The meeting is open for public comment.

(No response.)

ACTING CHAIRMAN BORER: Okay. If there is no comment, we'll move on.

I want to point out that the schedule as denoted here on the agenda shows a 3:00 p.m. adjournment time. We're going to try to move ahead reasonably efficiently to meet that adjournment time because of the extraordinary problems that now exist with regard to air travel and the extended time that some of our committee members need to be able to reach their planes in order that they don't have to stay an extra night.

It shouldn't be a problem if we stick to the schedule. So it may be that at some point I'll cut off discussion not arbitrarily, but only so that we can stay within our agenda.

In addition, you'll notice that there's a change in the alignment of the end table here. The only reason for that is so that I as the Chairman can see all of the committee members and not exclude them from commenting in the appropriate way at the appropriate time.

With that having been said, we'll begin the discussion of Diovan (valsartan) for the indication of treatment of patients with congestive heart failure. The sponsor is Novartis Pharmaceuticals Corporation, and the presentations will be introduced by Novartis by Dr. Mathias Hukkelhoven, Vice President for Regulatory Affairs.

DR. HUKKELHOVEN: Dr. Borer, Dr. Lipicky, members of the Advisory Committee, FDA, and guests, good morning. My name is Mat Hukkelhoven. I am Vice President of Regulatory Affairs for Novartis Pharmaceuticals Corporation.

On behalf of Novartis, I would like to thank you for this opportunity to present and review Diovan data for a new indication, the treatment of heart failure.

Diovan or valsartan is an angiotensin receptor blocking agent acting on the AT-1 receptor subtype. It was approved in 1996 for the treatment of hypertension, and it has been widely prescribed since that time. It is now available in over 80 countries.

We are pleased that we are able to present data which demonstrates clinical benefit with Diovan in treating patients with heart failure. Diovan is the first angiotensin receptor blocking agent to achieve such results. These beneficial results were achieved on top of a background regimen that included an assortment of approved drugs for each participating patient as prescribed by their physician.

Our development program for the new heart failure indication consists of several studies. Val-HeFT or Protocol 107 is a key morbidity/mortality trial involving approximately 5,000 patients, and it was conducted at 302 centers in 16 countries.

In addition, we also conducted four shorter term control studies, Protocols 103, 104, 106, and 110. These studies evaluated a variety of endpoints other than morbidity/mortality, including quality of life.

Our clinical program was developed in consultation with the FDA. Importantly it was agreed that the Val-HeFT study could employ two primary endpoints, and a positive outcome for either would support an application. The two primary endpoints are all cause mortality and the combined endpoint of morbidity and mortality.

Based on our clinical results, the following profile emerges. Diovan improves morbidity since it reduced hospitalizations for heart failure. It slows the progression of heart failure. It improves the New York Heart Association functional class rating and ejection fraction. It improves signs and symptoms of heart failure, and it improves quality of life versus placebo.

The most common adverse experiences were dizziness and hypotension.

We propose the following draft indication statement based on our data. Diovan is indicated for the treatment of heart failure, NYHA Class II to IV, in patients receiving usual therapy, such as diuretics, digitalis and either ACE inhibitors or beta blockers. Presence of all these standard therapies is not mandatory.

Our discussions this morning pertain solely to a new indication for congestive heart failure. Dr. Jay Cohn will discuss the efficacy of Diovan in treating patients with heart failure. Dr. Cohn is Professor of Medicine at the University of Minnesota, and he serves as the study chairman for our Val-HeFT trial.

Dr. Robert Glazer will then summarize the safety of Diovan in heart failure patients. Dr. Glazer is Director of Cardiovascular Clinical Research at Novartis.

Dr. Cohn will then return to summarize our perspectives of risk-benefit in this indication.

In addition to the speakers this morning, we have the following advisors who are available to answer specific questions the committee may have: Dr. Peter Carson, Associate Professor of Medicine at Georgetown University and Chairman of the Endpoint Committee, if and when he arrives; and Dr. lloyd Fisher, Professor Emeritus at the University of Washington.

I would now like to ask Dr. Cohn to the podium.

Thank you.

DR. COHN: Thank you, Mathias.

Dr. Borer, Dr. Lipicky, members of the committee, it's a pleasure for me to be able to share with you this morning the data supporting the use of valsartan in the management of heart failure.

Let me provide you a little background for a moment on why we are here today. The management of heart failure has undergone considerable changes in recent years.

The pointer is where? There should be a pointer here somewhere, but that's all right.

As you can see from this first slide, there has been quite a development of drugs for the management of heart failure over the last -- thank you -- over the last 15 years or so, beginning with the first demonstration that nitrate and hydralazine could alter the course of heart failure. That was the first clinical trial carried out in heart failure.

And then subsequently the ACE inhibitors were assessed initially in Class IV heart failure and subsequently in more moderate heart failure, Class II and Class III, demonstrating efficacy on the long-term outcome.

Subsequently there were data to support the use of beta blockers beginning in around 1996. In 1999, a single study demonstrated that spironolactone had a favorable effect in very severe heart failure. That has never been submitted to the FDA for evaluation.

And most recently, the demonstration that one could achieve benefit with biventricular pacing.

So there has been a considerable expansion of our therapeutic armamentarium over these years.

Now, the rationale for an angiotensin receptor blocker is, I think, well known certainly to the committee. It's widely appreciated that angiotensin exerts a variety of adverse effects both on the vasculature and on the heart and on neural hormonal system that may contribute to progression of heart failure.

Traditionally, we have used ACE inhibitors in an effort to inhibit the formation of angiotensin II, and that was, indeed, the concept over the years that we've been widely using ACE inhibitors to treat heart failure, but it's becoming increasingly apparent in recent years that ACE inhibitors given certainly in the doses that are currently used clinically does not suppress very effectively the formation of angiotensin II, and that a good deal of the efficacy of ACE inhibitors may be related to its preservation of bradykinin by inhibition of the breakdown of bradykinin, and that bradykinin nitric oxide system may be an important contributor to the long-term benefits of ACE inhibitors.

Thus, if angiotensin II still persists, and it may well also persist because of the activity of alternate pathways to formation, particularly the

chymase (phonetic) system which is active in tissues, then we still may have circulating in tissue levels of angiotensin II which interact with the AT-1 receptor to subserve vasoconstriction, vascular and cardiac growth, and adverse consequences in the syndrome of heart failure, and this, of course, is where the angiotensin receptor blockers, such as valsartan, which are specific inhibitors of the AT-1 receptor, might further block the renin angiotensin system which we believe has deleterious effects in heart failure.

Now, what I'm going to present to you today is the clinical development program for valsartan in heart failure, and as Mathias has already described to you, there are four preliminary studies that were done that led to the major outcome trial called Val-HeFT that will spend most of the time this morning discussing.

Study 103 and 104 were sort of proof of concept studies, that is, can one get a hemodynamic effect when one administers valsartan in patients with heart failure.

Hemodynamics do not serve as an adequate surrogate for long-term efficacy of drugs in heart failure. They may well serve as a target for acute interventions because there is an acute response to hemodynamic response, which will influence acute symptoms. But the long-term course of the disease cannot really be predicted by hemodynamic effects of the drug.

Nonetheless, it is often important, particularly if you're using your drug with a known hemodynamic effect, that we demonstrate that the drug is exerting this predicted effect in patients with heart failure.

So two trials were done. Study 103 was a trial in patients receiving neither ACE inhibitors nor beta blockers, and this was carried out in Russia, and it was placebo controlled and also lisinopril controlled, and there were -- and I'll show you the data on valsartan dosing -- there were 116 patients in this trial. It was of four weeks' duration, and the major primary endpoint was hemodynamic effects from right heart catheterization.

Study 104 was carried out in the United States and Veterans Affairs hospitals. The patients were all mandated to be on ACE inhibitor, and they were on ACE inhibitor in doses that are recommended from the large scale trials.

They could not be on beta blockers, and there was a placebo controlled assessment of four weeks' duration of administration of valsartan in two doses in 83 patients.

Study 106 was an exercise tolerance treadmill exercise study, and as I'll suggest to you in a moment, exercise, again, is not a very useful surrogate marker for long-term efficacy, but it's one of the kinds of endpoints that one often carries -- attempts to study.

These patients were allowed to be on ACE inhibitors and beta blockers, and the vast majority were receiving ACE inhibitors, and about a third of them were receiving beta blockers. There were 770 patients in this trial. It was a 16 week study, and the primary endpoint was exercise tolerance.

Study 110 was a study in which patients were not allowed to be on ACE inhibitors during the trial, but they had been on ACE inhibitors, most of them, until the randomization date. They were also allowed to be on beta blockers, and about 30 percent of them were on beta blockers.

And it was a comparison between enalapril and valsartan in patients who had been on an ACE inhibitor up until the day of randomization. So it's a protocol that basically asks the question: will valsartan exert the same benefit as continuing an ACE inhibitor in patients with heart failure?

And it was a six minute walk test, a 12 week duration study.

And then Val-HeFT, which will spend most of the time on, was carried out in 5,010 patients, and it was a morbidity/mortality trial, but in addition, there was a substudy in which a walk test was assessed.

Now, this is the study design of the two hemodynamic trials, Studies 103 and 104. This is the study in Russia. This is the study in the United States.

The Russian study examined three different dose levels of valsartan, 40, 80, and 160 milligrams twice daily, and they compared that to lisinopril, which was titrated up to ten milligrams a day, and there was a placebo group, and after a run-in the patients were randomly assigned to these five different treatment groups and followed for 28 days. They were catheterized at day zero and again at day 28 to assess hemodynamic effects.

Study 104 used two different doses, 80 and 160 twice daily versus placebo, and these patients, once again reminding you, were all on ACE inhibitor. None of these patients were on an ACE inhibitor prior to randomization.

These are the hemodynamic data in Study 103. The placebo group is in green. At the left are the bars at day zero, four to eight hours after administration of the drug. The values were meaned over that 48 hour period.

On day 28 assessments were carried out at zero time. That is, before drug was administered the patients had been on chronic therapy, and then the third set of bars is 12 hours after the dose was administered on day 28.

In green is the placebo group, and you can see very little change in pulmonary capillary wedge pressure during the follow-up period, and the little increase here.

The three valsartan doses are shown here, 40, 80 and 160 milligrams twice daily, and this is lisinopril, and you can see there is a clear hemodynamic effect of valsartan and probably also of lisinopril compared to placebo. These are least mean squares and some of the values are statistically significant and some not.

This is Study 104, and this again is the pulmonary capillary wedge pressure. Now the unique feature of Study 104 is that at zero time, patients were given a dose of lisinopril to maintain full ACE inhibitor effect throughout the study duration. So they got lisinopril here, and they got lisinopril again here before a dose was administered on day zero and day 28.

So the placebo effect in green is really a lisinopril effect in patients on chronic ACE inhibitor therapy, and then in addition to lisinopril, the valsartan patients were given valsartan 80 or valsartan 160, and what one sees is a trend for a dose response to valsartan, that is, a great reduction in pulmonary capillary wedge pressure here on day zero, here on day 28, and again on day 28, hour 12, there isn't much difference between lisinopril and the drug.

We also looked at diastolic pulmonary artery pressure because some patients didn't get their wedge pressure measured, and once again, a dose dependent reduction of pulmonary -- PA diastolic pressure.

This is systemic blood pressure, systolic blood pressure. Again, the appearance of a dose dependent reduction of blood pressure.

Now, we also measured hormones in Study 104, and this was plasma aldosterone levels. Plasma aldosterone levels were strikingly reduced with both doses of valsartan compared to lisinopril and appeared to be a bit of a dose dependent effect here.

And plasma norepinephrine also exhibited some decline in perhaps a dose dependent fashion, and in all of these studies, the one 60 milligram twice daily dose of valsartan exerted the greater hemodynamic effect. So that was the dose that was selected to be introduced into Val-HeFT when we designed Val-HeFT.

Now, the exercise studies I will review briefly for you. As I've already suggested, exercise tolerance, that is, treadmill exercise and six minute walk tests, have not served as a very reliable guide to efficacy and heart failure, but these studies were carried out just to determine if there was any demonstrable effect from valsartan.

This is the trial, 106 trial with treadmill exercise, and there were three different doses of valsartan studied in that trial versus placebo.

Remember most of these patients were on an ACE inhibitor, and somewhere over a third of them were on a beta blocker. There was no demonstrable difference among the four treatment groups on change in exercise performance and the p values were not significant. So no demonstrable additional exercise improvement when valsartan was added to background therapy.

And these are the two six minute walk tests. This is a substudy from Val-HeFT, and I put it in here because it was an exercise study. There were 633 patients in that substudy. There was really no striking change in six minute walk tests, essentially equal between the placebo and valsartan groups.

And this is Study 110, which was a six minute walk test. Now, this was a positive control study because it was valsartan versus enalapril in patients previously on an ACE inhibitor. There were 134 patients in this trial, and there was no difference at least between enalapril and valsartan. In fact, the trend was for a little greater improvement with valsartan than enalapril, but nowhere near statistically significant. So those studies are basically a wash.

Well, now let me go into the Val-HeFT protocol with you to review what we have done. This was the design of Val-HeFT. Entrance criteria, patients with chronic stable heart failure, and they had to have ventricular enlargement both by transverse diameter of the left ventricle at end diastole that is greater than 2.9 centimeters per meter squared, and by an echo ejection fraction of less than 40 percent, and they all had to be in New York Heart Class II to IV for eligibility in the trial.

Now, each of the echo labs that participated in this study were validated for their ability to both perform and read an echo, and it was a monstrous undertaking, I can assure you. Many of the centers were offended by the fact that they had to send three echoes in and demonstrate that they could do it right.

And I can assure you that most of them did not do it right, and the core laboratories that oversaw the echo quality had to go back and reeducate echo technicians and readers as to the importance of precision in the performance of the test.

They all eventually met the criteria, and we also monitored the quality control throughout by randomly collecting echoes and submitting them through our core laboratory. So we did the best we could do in a multi-center study without having all the echoes read in a single core lab, which would have been an unbelievable burden.

So these were the entrance criteria by echo, and then the patients were randomized. They stayed on their prescribed therapy, and we encouraged all of the physicians to get patients on optimal therapy for heart failure, and then they were randomized to receive valsartan 40 milligrams twice daily, which was titrated at two week intervals. It was a forced titration to 160 twice daily unless there were adverse events along the way that inhibited progressive titration, and I'll show you the data on that.

And then double blind placebo therapy in the other group, and we followed the patients. All of the patients were followed until 906 deaths were reported, and I'll show you how that figure was arrived at.

There were two primary efficacy endpoints as Mathias has already discussed to you. One was mortality, that is, the time to death, survival curves, and the other was a combined endpoint which we called morbidity, but it included mortality, all cause mortality. Plus episodes of sudden death with resuscitation were called an endpoint.

Patients who were not hospitalized, but needed therapeutic doses of intravenous inotropic or vasodilating agents for four hours out of hospital, that was equivalent we thought to a hospitalization, and that was counted as an endpoint. And the other one was hospitalization for heart failure.

In all of these events, the death and all the other primary endpoint events, were adjudicated by an endpoint committee, and they reviewed every one of the hospitalizations until a patient was identified to have had a hospitalization for heart failure, which gave them an endpoint for the trial. So this was a burdensome effort as well, but we thought it was very important to do.

Now, here were the statistical considerations in Val-HeFT. Since we had two endpoints, we divided our alpha into two and, therefore, assigned an alpha of a .025 to both morbidity and mortality.

The mortality alpha was further reduced by the interim analysis carried out by the Data Safety and Monitoring Board using the O'Brien-Fleming method, and that then came down to .02 as the level of significance.

The assumption of the sample size was based on a predicted placebo death rate of 12 percent per year. We didn't achieve that. It was nine percent. So in a way, we were under powered from the very beginning because the mortality rate was lower than we had predicted.

We were trying to identify a reduction in mortality of 20 percent with a 90 percent power and a two-sided significance of .025, and that's how we came up with the need for 906 events in order to achieve our target.

All patients were followed to the study end. They were censored obviously at the end of the study. At the time of loss to follow-up, and I can tell you that you'll see in a moment that there were very few of those, and at the time of heart transplant. So those were the censoring criteria.

There was an endpoints committee, as I pointed out that, reviewed all of the endpoints, and there was semi-annual interim analysis by the DSMB.

Patients were all over 18 years of age. They had chronic stable heart failure, and as I pointed out, their ejection fraction less than 40 percent and their left ventricle larger than 2.9 centimeters per meter squared.

For those of you who aren't used to the index, LVIDD, this means left ventricle well over 5.5, and usually over six centimeters in the trial.

And they all had to be on a stable regimen of prescribed heart failure therapy for at least four to six weeks prior to randomization.

The usual exclusion criteria, patients with significant valvular, obstructive valvular disease, patients with recent ischemic episodes or CVAs or recent reperfusion therapy, patients likely to need bypass or reperfusion in the near future. People with rapidly deteriorating heart failure were excluded. Those on the transplant list were excluded. Those with predominant right heart failure due to pulmonary disease were excluded, and those on drugs which we felt were contraindicated, such as Class IC anti-arrhythmics, patients who required IV inotropes or IV vasodilators in the previous three months.

There were 302 centers in 16 countries that participated in the trial, and this is the breakdown of the number of patients entered from each of these various sites. The United States entered a little less than 50 percent of the patients, and you can see the big contributors outside the U.S., Italy, the Netherlands, Germany, et cetera.

The follow-up averaged 699 days in both treatment groups. The mean daily dose of valsartan administered was 254 milligrams. Remember 320 would have been the target dose so that we came pretty close.

The mean dose of placebo was just slightly higher, 283.

The duration of treatment in days was somewhere over 600 days on average between the two groups, and 84 percent of the valsartan patients and 92.7 percent of the placebo patients achieved the target dose, which is, I think, pretty remarkable. This is a high dose of valsartan, and yet the vast majority of patients achieved that target dose.

Here's the disposition of the patients. Ninety-nine percent of them completed the trial either to death or a trial endpoint. Premature trial termination in the absence of death occurred in only a small number of patients, one percent in both groups; a small number of heart transplants, 18 and 23; loss to follow-up. I think this is a tribute to the quality of the performance of this trial. Three and four patients lost to follow-up out of 5,010. A few withdrew consent.

Four hundred and 48 or 18 percent of the valsartan patients and 14 percent of the placebo patients stayed in the trial, but discontinued trial treatment. Again, I think a very acceptable number.

The majority of those, the difference between the two was related to intolerable adverse experience, nine percent in valsartan, five percent in placebo.

Here are the baseline characteristics of the patients. There were 2,511 in the valsartan treated group and 2,499 in placebo. They averaged about 63 years in age. Eighty percent were male. Ninety percent were white.

We have an unfortunately small representation of black patients. Some of these are South African blacks, and a slightly larger fraction are African Americans. It is too small a group to make many conclusions about, and there was a small number of other racial groups identified.

Coronary disease was the etiology of the heart failure in about 57 percent of the patients. Thirty-one percent were identified as having idiopathic cardiomyopathy, and the causes are shown here.

Sixty-two percent of the patients were in Class II heart failure, 36 percent in Class III, and a small number of Class IV patients.

The ejection fraction averaged about 27 percent. The left ventricle was 3.6 centimeters per meter squared body surface area. So these are large ventricles.

The blood pressure, 124 over 76, and here was their background therapy, and this is going to become of some importance as we go through these data. Eighty-six percent of the patients were on a diuretic. Two thirds were taking digoxin. Thirty-five percent were on a beta blocker, and 93 percent on an ACE inhibitor.

Now, this was far higher than we see in the practice of medicine in the community today, and remember one reason for that is the patients who are on an ARB were excluded from participation, and even though there are no data yet, there are many physicians who are substituting ARBs for ACE inhibitors because their patients coughed once.

And consequently those patients were excluded. So these are patients who are largely on ACE inhibitors, and only seven percent were not on an ACE inhibitor, and that's an important group, too.

Their quality of life was assessed by the Minnesota living with heart failure questionnaire, and the overall average score was about 32, which by the criteria of that questionnaire puts them in the moderate heart failure range, not severe, but not mild either, and that's broken down by the emotional and physical component of that Minnesota form.

Well, what kind of doses of ACE inhibitors were they on at baseline? And this is the list of ACE inhibitors that were being used. Remember these are physician choice.

The three biggest ACE inhibitors in use were enalapril, lisinopril, and captopril. The doses of these drugs are very close to the recommended doses. For enalapril and lisinopril, very close to the 20 milligram dose which is generally recommended daily dose. Captopril, probably lower than one would have chosen to use in clinical trials, but obviously this drug is often used more than once a day, and the mean dose was about 80 milligrams, and the other is incomparable doses.

What about the beta blockers being used? Well, the two most commonly employed beta blockers were carvedilol and metoprolol, and they were used in doses which are probably lower than one would choose based upon clinical trial data, but this is the real world, and I see patients coming referred to me on beta blockers, and for the most part, they're on low doses.

There's some concern about titrating up to the doses that have been used in most clinical trials, and then the other beta blocker uses are shown below.

Well, this was the primary endpoint of Val-HeFT. These are the two primary endpoints. Mortality was identical in the two treatment groups, a hazard ratio of 1.02.

The morbidity endpoint, which of course includes mortality, exhibited a striking reduction in the valsartan group compared to the placebo group, a hazard ratio of .87, a p value of .009.

So the study achieved its target endpoint with one of the two primary endpoints.

Here is the Kaplan Meier survival curve exhibiting superimposition of the placebo and valsartan arms over 30 months of follow-up.

And what about mechanism of death? These were all adjudicated by our endpoint committee, and you will see there's very little difference between the valsartan and placebo groups. Sudden death here, pump failure death here, sudden death with per monitory worsening of symptoms, other vascular causes, non-cardiovascular deaths very similar.

So there appeared to be no mechanistic difference in what led to death in the two treatment arms.

Here is the Kaplan-Meier curve for morbidity, which exhibits separation beginning at about three months and then widening over time. Once again, this was a 13.2 percent risk reduction, and the p value was .00852.

Now, in the morbidity endpoint, obviously we had four different possible contributors to morbidity, and here is the breakdown of those four contributors. The biggest difference was heart failure hospitalization occurring at 18.2 percent of the placebo group and 13.8 percent of the valsartan group, and that leads us to do an analysis of hospitalizations in more detail.

Here were the cardiovascular deaths, which are very similar. Here's the first nonfatal morbid event exhibiting a striking reduction in the valsartan arm, first heart failure hospitalization, a similar reduction of hazard ratio to .725. The first sudden death with resuscitation, very small numbers so not too meaningful.

This is the curve for incidence of worsening heart failure. Now, one has to censor deaths when one does this kind of an analysis, but it gives you some idea about the frequency in which patients are hospitalized for heart failure as an initial event, and you can see that the curves begin to separate at about three months ago, and they widen over time.

This is a 27.5 percent risk reduction, and the p value for that is .00001.

Now, the agency had raised some issues about overall cause hospitalizations, and let me just provide you that data because I think there's been a bit of a confusion about that.

Heart failure hospitalizations, now these are investigator assessed because all the endpoint committee did was to adjudicate the first heart failure hospitalization. Once a patient had been hospitalized for heart failure, they no longer adjudicated hospitalizations.

But of course, hospitalizations occurred, and the investigator was busy assessing hospitalizations; the investigators were doing this on their own.

Well, what did the investigators find about heart failure hospitalizations? Well, the investigators identified 266 fewer hospitalizations in the valsartan group compared to the placebo group. This was all cause hospitalizations, a similar reduction. So this is statistically significant. This is obviously not because it is influenced by a large number of non-heart failure hospitalizations which were equal in the two groups.

So this is a tribute in our decision to adjudicate heart failure hospitalizations as an endpoint for the trial. They had no reason to think that valsartan would reduce the number of other hospitalizations, but we hoped it would reduce heart failure hospitalizations and not increase non-heart failure hospitalizations, and that's indeed what we found.

Now, there's also been some question raised about the days in the hospital. Not only are we interested in how many patients get hospitalized or how frequently they're hospitalized, but what about the number of days in the hospital?

Well, highly significant reduction of days in hospital, mean days in hospital during the trial, 3.5 for valsartan compared to 4.8 for placebo. All cause hospitalizations also tend to be reduced, not quite significant, and non-heart failure hospitalizations once again, identical.

Well, what about days alive and out of the hospital? Well, this is an attempt to get at that. This is not an easy number to get at, but once again, as you might expect, there's more days out of the hospital and alive in the valsartan treated group than in the placebo group. And these are, again, the data I showed on the previous slide.

Well, you can't do a statistic on that very easily because the number of days out of the hospital, alive and out of the hospital, varies tremendously based on when the patients were entered into the trial because there's a wide range of duration of follow-up. So you get this wide standard deviation, and then you get a p value which, of course, is nowhere near significant.

So it's important to correct the number of days alive and out of hospital for the number of years of follow-up or months of follow-up, and we've done that on this slide. And this is the mean days per year alive and out of hospital, and now, of course, the standard deviation gets much lower, and the heart failure days in hospital is highly significant. All cause is not, and of course, non-heart failure remains essentially identical.

Well, we monitored a number of secondary endpoints. Signs and symptoms of heart failure were assessed by the investigator, and new York heart class was assessed, and this is the change from baseline to endpoint in each patient of these measurements.

Here's New York heart class. More patients with valsartan improved and fewer worsened than in the placebo group, and that was highly significant, .001, and that's true also of jugular venous distention, of edema, of rales, not quite of third heart sound, of paroxysmal nocturnal dyspnea, of dyspnea at rest, dyspnea on effort, fatigue, and not quite orthopnea.

This is actually a remarkable demonstration of efficacy. I've never seen a trial before in which all of these secondary endpoints exhibited a benefit from a drug. It's hard to come up with this kind of data, and so this was remarkably congruent.

DR. FLEMING: Could I just -- you're making a key point. If we could go back, Jay, if I could just interrupt you for a second.

DR. COHN: Sure, by all means. I love to be interrupted, Tom. So don't hesitate.

DR. FLEMING: When you point out how remarkably congruent it is, I look at that and wonder as a statistician if it's even more congruent than random chance would anticipate. Yes, these are all significant. They all show about a one to three percent more favorable result in the percent that improve and a one to three percent more favorable result in the number that worsen, almost exactly the same across all of these endpoints.

There must be a lot of correlation between these endpoints.

DR. COHN: Well, sure, there are. New York heart class and symptoms all go together. Jugular venous distension is an observation which shouldn't really relate to such things as fatigue or dyspnea, but they are all going to go sort of together, and your point being then what?

DR. FLEMING: My point being addressing your point that it was remarkable that they were all significant. Well, if in fact you have a general quality of life phenomenon and you have many variations of measuring the same phenomenon, then I would expect a consistency and significance across those results. It's not as though we have 13 independent assessments all of which --

DR. COHN: Oh, no.

DR. FLEMING: -- achieve a p of .001.

DR. COHN: I would agree. I was just commenting on the fact that in other trials, and I've been involved in a lot of other trials, it's been very hard to actually demonstrate any clinical benefit on these kinds of clinical measurements, and we were able to do it in this study on all of them.

But your point is well taken that they are mutually dependent in many respects. So you might expect them to go together.

This is the Minnesota living with heart failure questionnaire. Now, this is filled out by the patient. All the other data are obtained by the physician, the physician's assessment. This is the patient filling out a form. This form is completed when the patient walks in the door before they meet with the health care provider.

So they sit down and they fill out this 21 question form. So it's very independent kind of assessment before they've been influenced by meeting with the nurse or the doctor.

And the primary endpoint was the change from baseline to endpoint, whenever the endpoint occurred, and the patients on placebo exhibited a progressive worsening of their quality of life. A rise in score means quality of life has become worse. The patients on valsartan did not exhibit that worsening. So the overall score was highly significantly favorable for valsartan.

Now, this score is traditionally broken down into two components. I must point out to you that this overall score has been heavily validated in a number of trials. The breakdown into emotional and physical is not as well validated, but nonetheless, there are a group of questions which are defined as the physical score and another group defined as the emotional score, and there was a similarity in the benefit of valsartan, perhaps more dramatically with the physical score which you might have expected.

This is the time course of quality of life over the 30 months of the trial. Assessments were made at four, 12, 18, 24, and 30 months, and you can see that in the first few months there was a tendency for an improvement in quality of life, not really different between the two, and at 12 months not really much different.

But by two years, there was a significant difference. At the endpoint, again, out here, the trend was even greater. None of these were quite independently statistically significant, but the endpoint was, and that was the prescribed endpoint for the trial, was baseline to endpoint, and whenever the patients ended, they were assessed.

Now, ejection fraction was monitored by echo, and these are the ejection fraction data from baseline to endpoint. The valsartan group exhibited a rise of about four units of ejection fraction. The placebo group, a rise of about three units. It's a small difference, but highly significant.

And this is the sequential changes in ejection fraction over time. By four months there was already a difference; 12 month, 18, 24, and 30 months. In all of these time frames, the valsartan group exhibited a greater rise in ejection fraction than the placebo group.

Now, I must tell you that this increase in ejection fraction in the placebo arm is not consistent with previous data. We've monitored in the Val-HeFT trials. The placebo group goes down.

Yeah, Jeff.

ACTING CHAIRMAN BORER: Well, I wondered about this, too, but all of these people are on background therapy.

DR. COHN: Exactly.

ACTING CHAIRMAN BORER: That affects EF over time.

DR. COHN: And, in fact, as you will probably see later on -- I think I have a slide that shows it -- the group on beta blocker, we did not prescribe that they had to be on beta blocker for, say, more than six months before they're randomized. So what you will see in the beta blocker treatment is a tendency for an ejection fraction to go up during the course of the trial from the beta blocker itself. So this is a very well treated group, and I suspect the rise in EF reflects the effect of other drugs.

But that's the way the data came out, and it was certainly a significant benefit of valsartan.

Now, we also did ejection fraction Study 106 just to show congruence because here we have three different doses of valsartan, 40, 80, and 160, and here was the placebo group, and you can see in Study 106, all three doses seem to improve EF more than the placebo.

Once again, it does look like 160 was the more effective dose. So we were reassured that we probably did use the right dose in Val-HeFT.

Now, we also monitored left ventricular chamber dimension by echo, and this is the change from baseline to endpoint of left ventricular internal dimension in diastole. It went down in the valsartan group, less reduction in the placebo group, the difference very highly significant.

And here is the sequential changes in LVIDD. The reduction was apparent by four months and persisted throughout the trial. So clear evidence for benefit on left ventricular dimensions or remodeling, which I would call that, from valsartan, but modest, not gigantic, but highly significant, a tribute to the large numbers of patients that we were able to monitor.

Now, we also measured neural hormones. This was norepinephrine and BNP levels monitored over time, and the endpoint was baseline; the secondary endpoint was baseline to the endpoint assessment, and here you can see valsartan prevented the increase in norepinephrine over time that was observed in the placebo group. This is the valsartan group, highly significant difference.

And with BNP, the placebo group rose. The valsartan group fell, and once again, a highly significant difference.

Here are the sequential changes in norepinephrine. They were measured at four months, 12 months, and 24 months, and you can see at each time frame norepinephrine was rising in the placebo group, not in the valsartan group.

And here is BNP once again, a decline in the valsartan group and a progressive increase in the placebo group, again, highly significant differences.

So what can we say from this overall summary of the Val-HeFT data? We can say that valsartan clearly reduced morbidity in patients receiving prescribed therapy for heart failure by 13.2 percent. This was the p value.

It decreased the risk for first heart failure hospitalization by 27.5 percent, and here the p value gave us four zeros before the one.

It improved signs and symptoms of heart failure. It improved ejection fraction, reduced left ventricular dimension, improved quality of life, and had a favorable effect on norepinephrine and BNP, albeit with similar mortality between the two groups.

Now, let me then just put all of this together with the three placebo controlled preliminary studies, as well as Val-HeFT and review what we have learned about valsartan.

Hemodynamics, I think clear evidence that valsartan is effective on hemodynamics both in the absence of ACE inhibitor and in the presence of ACE inhibitor.

We've determined that valsartan cannot produce further improvement in exercise performance, at least in the modest size studies that we have carried out, in addition to background therapy. In the patients in whom we have monitored signs and symptoms, the Val-HeFT study, they were improved.

Quality of life was improved in the Val-HeFT. Neural hormones were demonstrated to be improved in actually all three of these studies, not determined in 106.

Left ventricular function was monitored in two studies, Val-HeFT and 106, and in both of them the effect was favorable, and of course, the morbidity trial and mortality trial, Val-HeFT showing a favorable effect on morbidity, but no demonstrable benefit on mortality.

So let me stop there in terms of the overall study, and then we're going to get into subgroup analysis in a few minutes, but I'll be delighted to take any questions from the committee at this time.

ACTING CHAIRMAN BORER: Let's keep the questions to clarification of the data at this point because Jay has the risk-benefit discussion later when we can get into philosophical issues if there are any.

I have one question while everybody is sort of gathering their stuff. If you go back to slide EC-14 --

DR. COHN: EC-14.

ACTING CHAIRMAN BORER: -- yes, this is really a secondary issue, and it's just for clarification purposes because you already made the point, I think, quite correctly that short-term exercise studies don't predict long-term benefit.

But my reading of what we were sent was that an imputation of a zero value was made for people who died when exercise time was determined for this study, and that the determination that the zero value should be imputed was made after the study was completed.

I don't know when it was made relative to unblinding. I'm sure it was before unblinding, but we weren't told that, and that if you didn't impute the zero value, in fact, the patients on enalapril did better nominally, not significantly, but nominally, than patients on valsartan.

So although this is not the big, burning issue of the day, I'd like a little clarification about that, if you would.

DR. COHN: Yes. I wasn't involved in one of those two studies and the analysis of the substudy. Who can address that?

Tom. Tom is our biostatistician.

MR. CHIANG: Tom Chiang, Novartis.

Yes, the zero imputation has been defined and decided and documented prior to unblinding for analysis, and real data imputation obviously is done after unblinding.

ACTING CHAIRMAN BORER: Well, don't go away yet.

Why did you do that? I mean, I'm just speaking post hoc here. I mean, you have no evidence of a difference in mortality in these groups. So after the fact one might expect that maybe, you know, you're not affecting death, but we're suggesting that we are affecting heart failure.

Why would you impute a zero value to people who died rather than last value carried forward? I mean, what was the reasoning behind that?

MR. CHIANG: Before unblinding, we don't know mortality would be equal. So we plan a lot of sensitivity analysis, and you know, imputation for patient died, you know, or could not work due to heart failure, you know, was defined.

ACTING CHAIRMAN BORER: Thank you.

Are there any other questions, issues of clarification from the committee? Paul.

DR. ARMSTRONG: Could we look at Slide 37, please?

Jay, I'm trying to understand the sample size here, and it just gives me a repetitive number of the overall study rather than the patients who were hospitalized according to these categories. Can I get the appropriate sample size that lines up with those three categories?

DR. COHN: I'm not sure what you're -- this is the number of patients that were randomized.

DR. ARMSTRONG: Right.

DR. COHN: And this is the mean days in hospital for those 2,511 patients, and just averaged out over the --

DR. ARMSTRONG: But I just want to know how many patients were hospitalized for heart failure.

DR. COHN: Oh, I guess I showed you that on another slide. What was the --

DR. ARMSTRONG: Okay. I couldn't find that.

DR. COHN: Yeah, what's the slide that shows the heart failure hospitalizations?

DR. ARMSTRONG: Thirty-four, EC-34.

DR. COHN: Yeah, this is heart failure hospitalizations, and this is the number of hospitalizations. So it's 1,000 in the valsartan group, and I can't -- you'd have to extrapolate about 1,266.

ACTING CHAIRMAN BORER: You have the exact numbers in EC-34.

DR. COHN: There's 266 more here than here.

ACTING CHAIRMAN BORER: It's 923 against 1,189.

DR. COHN: Yeah, okay.

DR. ARMSTRONG: Nine, twenty-three against?

ACTING CHAIRMAN BORER: Eleven, eighty-nine.

DR. ARMSTRONG: And, Jeff, you may want to --

ACTING CHAIRMAN BORER: No, that won't help any.

DR. ARMSTRONG: You may want to reserve the issue of adjudication in the process. You may want to reserve that for later.

ACTING CHAIRMAN BORER: No, if you want to know how it was done, let's ask now.

DR. NISSEN: I'm a little confused about the process of adjudication. As I understand from our briefing book, these were brought to the endpoint committee if it was perceived that they were heart failure, but if they --

DR. COHN: No, all hospitalizations were brought to the -- every hospitalization was referred to the endpoint committee, and then they determined whether the hospitalization was from heart failure. They didn't even have available to them the investigator's assessment. They just saw the data for every hospitalization.

So it was a monstrous undertaking as you can imagine, and then they identified those that they felt were due to worsening heart failure.

DR. NISSEN: Our briefing book suggests that the sponsor screened all hospitalization endpoints, and those that didn't meet endpoint criteria were not submitted to adjudication so that there was some initial adjudication.

And the second point was that I'm confused about overnight stays in the emergency room included in hospitalization. As I understand it, if a patient was in the ER for 12 hours during the day, they would not be categorized as hospitalization, but if they were in the ER for 12 hours overnight, they would be. Can someone help me with that?

Because that's a troublesome issue we all face.

DR. COHN: Bob, do you want to address the process that you used in Novartis?

DR. GLAZER: Robert Glazer, Novartis.

What happened with the documentation that went to the endpoint committee was that clearly cardiovascular events per a serious adverse event report that came to us, and those events that had any question of being cardiovascular were sent to the endpoint committee, and what was sent to the endpoint committee was the SAE narrative itself, case record forms or case record form printouts, and any hospitalization data that was collected after that, meaning hospital discharge summaries and, if needed, depending on the case, histories and physicals, laboratory data, ECGs, chest X-rays, and progress notes.

For those cases that were clearly non-cardiovascular, for example, an orthopedic problem that the person was being admitted for, a listing was provided with the patient's identifier and the diagnosis from the serious adverse event report form, and that was provided to the endpoint committee chairman.

If at that point in time he requested additional information, that information was provided, and we collected the hospital records.

DR. LINDENFELD: Just to clarify, so a hospitalization recorded by the investigator as hypotension due to over diuresis would not have been reviewed?

DR. GLAZER: Oh, that definitely would have been reviewed because it would have been considered cardiovascular. The key points are things like orthopedic problems, for example. Those were put into listings, and again, the endpoint committee chairman could ask for additional information.

DR. ARMSTRONG: You'll excuse me for persisting, but again, in our briefing book, there's an addendum to the endpoint manual that defines an admission due to over diuresis or drug toxicity as a hospitalization for reason other than heart failure.

DR. GLAZER: That's right.

DR. ARMSTRONG: So I'm a little confused. Did you --

DR. COHN: Well, they adjudicated those. Those --

DR. ARMSTRONG: So if a patient came in with over diuresis and hypotension or hyperkalemia or some complication we would ordinarily associate with heart failure, it was classified as not heart failure admission?

DR. COHN: Well, the definition for heart failure, maybe we can put up that slide for the definition here while Bob is still there.

The issue was worsening heart failure. So over diuresis is not worsening heart failure. It is a cardiovascular hospitalization.

This was the endpoint committee definition of hospitalization for heart failure. It was obviously severe collapse, pulmonary edema, symptoms and signs requiring intermittent or continuous IV therapy.

Hospitalization is defined as an overnight stay even if total duration is less than 24 hours. Remember we also included as our primary endpoint more than four hours in an emergency room. So we captured all of those events, but by definition hospitalization is an overnight stay whether it's in the emergency room or elsewhere, but the four hour criteria was also captured.

DR. ARMSTRONG: If you got an inotrope, right?

DR. COHN: If you got an IV diuretic or an inotrope or a vasodilator and you had to stay there for four hours. And there were very few of those, as you saw on that previous. So almost everything was captured by hospitalization.

But the committee made the judgment that if somebody came in because they were over diuresed, that it is a hospitalization, but it's not a hospitalization for worsening heart failure.

ACTING CHAIRMAN BORER: Steve.

DR. NISSEN: Yeah. You know, it's interesting because we all sort of seem to flag the same points when we reviewed this, and I also was a little bit uncomfortable. Let me make sure I understand this.

A patient that came in at 6:00 a.m. and went home at midnight, came in with symptoms of some kind, was not adjudicated as heart -- it could not have been a heart failure admission.

DR. COHN: No, they were adjudicated, and they were counted as a more than four hour stay out of the hospital.

DR. NISSEN: Well, I'm just reading what the FDA reviewer said. It said, "Hospitalizations that were clearly less than 24 hours were not submitted as events." That's what the book says. Now, is that right or wrong?

DR. COHN: Bob, can you clarify what happened with those events which were more than four hours in an intensive care unit?

DR. GLAZER: Again, if the information, any information concerned that it was a cardiovascular event, it went to the endpoint committee. How the endpoint committee classified it if they wanted to classify that as a hospitalization for heart failure, they made a definition that it had to be an overnight stay.

DR. NISSEN: So that would not have been a hospitalization for heart failure no matter what?

DR. GLAZER: From my understanding.

DR. NISSEN: Okay. So the FDA reviewer was correct there. I'm puzzled by the rationale for that, that's neither here nor there.

DR. COHN: Well, you have to have a definition for hospitalization.

DR. NISSEN: Sure, but you know, what it means is that a 12 hour hospitalization that occurred from 9:00 p.m. to 9:00 a.m. was a heart failure admission, and one that occurred, you know, from 9:00 a.m. to 9:00 p.m. wasn't. I mean to us that seems bizarre.

DR. COHN: But they were captured by the four hour criteria. So they all come out as a primary endpoint. There's no distinction made between primary endpoint from hospitalization and for four hours or more in an intensive care unit.

DR. NISSEN: No, no. But I'm saying somebody didn't come into an intensive care unit. They came into, you know, a hospital ward, you know, for an 18 hour admission. If that admission did not have an overnight stay, it would not have been a heart failure admission.

DR. GLAZER: And again, they would have had to receive from my understanding four hours of an IV inotrope during that, and that met the criteria. So I think, if I recollect, the problem was collecting the times, and the dates were collected, but the times were not clear on many of the cases.

So I think for adjudication purposes, change in date was a criteria for calling it overnight stay.

DR. NISSEN: yeah, I understand where we're coming from, but so we all understanding each other, you could come in at nine o'clock in the morning, and you could spend the day getting large boluses of intravenous furosemide to get you out of heart failure and go home at 9:00 p.m. that night, and that would not have been a heart failure admission.

DR. COHN: But it would have been captured as a four hour or more stay with aggressive intravenous therapy.

ACTING CHAIRMAN BORER: I think you've got to have an inotrope, Jay.

DR. COHN: No, no. It doesn't say inotrope. It says --

ACTING CHAIRMAN BORER: It does in our briefing book. So we've got some confusion here --

DR. COHN: Oh, really?

ACTING CHAIRMAN BORER: -- that we need to sort out.

DR. NISSEN: That patient that I just defined --

DR. COHN: Here. Let's just remind you of the morbidity endpoint.

DR. NISSEN: I mean, my reading of the book, and you're going to have to tell me if I'm wrong, is that a 9:00 a.m. to 9:00 p.m. admission getting a bunch of intravenous furosemide for a patient with pulmonary edema would not have been heart failure by this definition.

DR. GLAZER: That was my understanding.

DR. NISSEN: Okay. Now, you know, it may not have made any difference, but it doesn't seem very logical to me, I must tell you.

ACTING CHAIRMAN BORER: Can you tell us, just so we can put this into context, how many patients would have fallen into the category that Steve is talking about?

I mean, you know, this was a big study. If the number of patients not captured under this particular definition is relatively small, we can probably just, you know, move on, but do we know what the numbers are?

DR. COHN: Well, if we didn't capture the data, the endpoint committee would know. And I'm sorry that Peter is not here yet, and hopefully he will show up because he's the one who processed all of these, and all of those data would have been provided to him.

If a patient was in the hospital from 9:00 a.m. to 9:00 p.m., they would have gotten that data as an event, and then they would have decided what to do with it.

Now, my understanding is that they used the overnight stay as a criteria for hospitalization, and I guess we could ask him in how many instances they reviewed a case that didn't meet the overnight stay, but actually didn't get captured at all because they didn't by chance get nitroglycerine or nitroprusside or something and got IV diuretic.

We'll try to ask him when he comes. My understanding is it was almost zero that didn't meet the criteria.

DR. NISSEN: Can I just tell you? I mean, in our institution patients come in all the time, and they come into the emergency department. They will come in in the morning, and they're in pulmonary edema, and they have known severe heart failure. They're put in a short stay unit. They get IV diuretics to get them out of pulmonary edema, and they go home the same day.

DR. COHN: Well, they probably get some IV nitroglycerine, too, which would make them eligible for the four hours, and we can ask --

DR. NISSEN: I actually wish they did, Jay, but they don't always.

DR. COHN: Yes.

DR. NISSEN: And so just to me it's a hole here that I think -- I mean, I just want to make sure I understand it well.

There's a reason why you're getting a lot of discomfort here, and I'm going to just speak for myself and tell you why we're --

MR. MacNAB: I just want to make it clear that --

ACTING CHAIRMAN BORER: Could you use the microphone?

DR. COHN: Use the microphone, Malcolm.

MR. MacNAB: I wish Peter was here, and we can get any additional information you want, maybe not today, but we can get it for you.

I think the real problem -- I remember discussing this with him -- was we wanted to be consistent, and we wanted to be accurate, and the worst thing would have been to improperly classify people and without the times, which were not consistent. The most consistent thing was the date.

And, again, it was randomized. It was blinded, and I think the decision of the endpoint committee was made to do it right and not make mistakes.

ACTING CHAIRMAN BORER: Can I just make one --excuse me one second, Steve.

Just to clarify this further, my understanding is that the data indicate that valsartan was more effective than placebo on top of background for all morbid events combined. That was driven predominantly by the hospitalization, but it was true for all morbid events combined, which would include the hospitalizations, and the non-hospitalizations.

DR. COHN: That's right.

ACTING CHAIRMAN BORER: Is that correct?

DR. COHN: That's correct. That's correct.

ACTING CHAIRMAN BORER: I mean, that may put this in a different context perhaps.

DR. COHN: I've given you the hospitalizations separately to show the 27.5 percent reduction, but the primary endpoint was all the events combined, and I, frankly, believe -- and I can't -- Peter would have to verify this -- but I believe the number of events that were not captured because of these rules is almost zero because the committee was very attentive to every event, and they reviewed all of these events.

And if they had excluded a patient who was getting boluses of diuretic every hour for 16 hours and came in at nine in the morning and went home at midnight, and it didn't count as a hospitalization, they would have been as disturbed as you are actually, Steve. So I think they --

DR. LINDENFELD: Jeff.

ACTING CHAIRMAN BORER: JoAnn.

DR. LINDENFELD: I think JD-3 -- I think the only things that were included were if you got an inotrope or vasodilator for more than four hours. I don't think the kind of admission that Steve was describing would have been included in more than --

DR. COHN: Well, that's right, but what I'm saying, JoAnn, is I don't think there were many, if any, of those events that would have influenced the result, but we'll check with Peter.

DR. NISSEN: See, we don't know, and the reason we don't know is I'm going to read you what the FDA reviewer says. "Hospitalizations that were clearly less than 24 hours were not submitted as events." Therefore, if they're not submitted --

DR. COHN: No.

DR. NISSEN: -- then they're not adjudicated.

DR. COHN: No, no, that's not true.

DR. GLAZER: That's not correct.

DR. COHN: They were submitted, and then --

DR. GLAZER: And that's what we have in our briefing book.

DR. COHN: But the process that we used was outlined by Dr. Glazer. Essentially most everything, unless it was very obviously, you know, a patient was admitted for plastic surgery or something, and then that would have been listed, and the endpoint chairman could have asked for that if he wanted.

But all of those types of events you're talking about would have been listed for him, and I believe, Dr. Cohn, the number of the types of patients you're talking that came into the ER for what we would call a little tune-up of IVs is not that great, but I can get you the -- I will get you those numbers for you.

DR. NISSEN: I guess the reason that a lot of us are uncomfortable is that we would have preferred an independent adjudication process, and when we read about a process where the company is submitting, selectively submitting events to a committee, as opposed to a committee that reviews everything, it makes me uncomfortable. I guess that's the problem.

DR. GLAZER: Well, I can assure you it was done. Every event or every possible hospitalization, the listing of what it was was available to it, and most of the hospitalizations for this type of patient are obviously cardiovascular. I believe it was pretty independent, and I think when Peter is here, I think he will say the same thing.

DR. COHN: And it was all blinded, Steve. So I mean, there was no --

DR. NISSEN: I understand. I understand.

MS. TARGUM: I just want to point out that the information the agency received was relied upon in the manual.

ACTING CHAIRMAN BORER: Okay. So it sounds as if from the definitions that we have here that if you got inotropes or vasodilating agents for four hours, that would be included. But if you just got diuretics, that wouldn't be included, and that may not be correct in practice.

So we'll have to wait for the endpoint committee to clarify that for us.

JoAnn and then Alan.

DR. LINDENFELD: Just to come back to this adjudication process, I understand that from trial end to trial completion there was a difference of from 906 to 975 deaths. My question is: how many additional heart failure hospitalizations were there in that same period of time?

And our briefing book suggests that the deaths and the hospitalizations between that period of time, trial end and trial completion, were not adjudicated; is that correct?

I guess I wonder how many of them that is.

DR. COHN: No, everything was adjudicated up until the final -- I'm sorry. I didn't quite hear your point, JoAnn.

DR. LINDENFELD: Well, our booklet says that between trial end and trial completion there was a difference of from 906 to 975 deaths. It says there was no adjudication on mortality/morbidity endpoints at trial end.

DR. COHN: No, that's not correct. The 906 was the number of reported deaths. When the DSMB met, identified that it had passed the 906 point, and recommended that the study be terminated. That recommendation goes to the sponsor. The sponsor makes a judgment with the steering committee to terminate the study, sets a date for termination, and then many more deaths are still being reported during that period of time.

All the events that occurred until the end of the trial were adjudicated by the committee, everyone.

ACTING CHAIRMAN BORER: Alan.

DR. GLAZER: Can I just clarify? Robert Glazer from Novartis.

The events that occurred from May 3rd to the end of the trial were not adjudicated.

DR. COHN: But you don't mean by the end of the trial. You mean by -- the trial ended on --

DR. GLAZER: May the 3rd, and that's when the 906 deaths we were made aware of, and that was the day that the trial was considered completed. Subsequent to that, bringing in the last patient for last visit for follow-up to conclude, officially conclude, the trial, there were events that occurred. Those events were not sent to the endpoint committee because --

DR. COHN: And they aren't counted --

DR. GLAZER: And they are not counted.

DR. COHN: -- in our analysis either. The analysis is as of May the 3rd when the trial ended.

DR. LINDENFELD: Is the analysis on 906 or 975 deaths?

DR. COHN: Nine, seventy-five.

DR. LINDENFELD: Well, but then that difference between 975 and 906, those were the number that were then not adjudicated, and I assume there's a similar percentage of --

DR. COHN: No, no.

DR. LINDENFELD: -- hospitalizations.

DR. GLAZER: I'm sorry. There was a certain number of events that occurred from when observed the 906 deaths. When we observed the 906 deaths, obviously when we collect documentation afterwards, there were additional people who had an event, a morbid event or a mortal event that hadn't been reported to us or was in the process of coming to us through the process.

That's what accounts for the additional information. So those events, yes, they were adjudicated and put into the analysis. that's why it doesn't end at 906, because that was a date that we were made aware. We found these extra events as we were doing the --

DR. COHN: I mean, let's make it clear. Every event that occurred before May 3rd, which was the termination of the trial, were adjudicated. Now, other people had events, and then they have to be brought back, and they have to be told about the results of the study. They have to be taken off their study drug.

People don't go off study drug on May 3rd. They have to come back in for a visit, and between May 3rd, which was the official termination of the trial, and the time that they came back and were taken off of their study drug, there were events that took place, but they weren't part of the trial. That was post trial, and they're not counted in any of the analysis that we've shown here.

DR. FLEMING: May 3rd was the date of 906 deaths or 979?

DR. COHN: No, come on. You're waiting for reports to come, and when the number of reports --

DR. FLEMING: We fully understand that between a data monitoring committee's review and when the database is finalized, additional events come in.

DR. COHN: Yeah.

DR. FLEMING: The question is very simple, and I think your answers so far seem to be confusing. The date at which the data monitoring committee met, there were 906 deaths. The reports that we've been provided --

DR. COHN: There were over. There were more than --

DR. FLEMING: -- give us 979 deaths. Presumably there were also emerging during that time frame CHF hospitalizations as well.

Simple question: is the primary analysis that we've been shown for morbidity events, were all of the CHF hospitalizations in that analysis? Were all of them adjudicated?

DR. COHN: Yes.

DR. FLEMING: Thank you.

MR. HAUPTMAN: Let me clarify -- Lawrence Hauptman, Novartis -- the 906 and the 975 number. The 906 was reported as hitting -- that was supposed to be the endpoint, that many deaths, and then it was decided that that was May 3rd, but then in going back to the field and getting all of the paper work in, those extra 70 people were discovered in terms of deaths and also in terms of the morbidity endpoint.

But they all occurred before May 3rd. So anything that occurred before May 3rd is what you see in the data that was submitted in the analyses. Stuff that happened after May 3rd is after the trial ended and is not part of any -- you haven't seen any data on anything that happened after May 3rd.

DR. FLEMING: So, Larry, then the final updated database indicated that by May 3rd there were 979 deaths.

MR. HAUPTMAN: That's true.

DR. COHN: That is correct, and it's always true in trials. When the reports come in, you wait for the reports of the target number and then you terminate the trial. You don't terminate it the day the DSMB meets. The DSMB has to meet and make a recommendation. There has to be a date set when you're terminating the trial, and that is the date.

And at that point, by going back and reviewing every center, there were 900 and whatever it is, 70-some deaths.

ACTING CHAIRMAN BORER: Alan and Paul on this same issue, and then we'll go to Tom for a new issue. Alan, did you have?

DR. HIRSCH: Well, my question is related. Obviously our goal is to make sure that the medication if used by the public for heart failure breeds a clear benefit. So we're sort of adjudicating right here.

What I'm wondering is do we have in the room at this time data that we can look at on emergency room use by the two groups in Val-HeFT.

Sort of opinions. We love to see data.

DR. COHN: You mean by the four hour criteria for emergency room or do you mean just having to go to the ER?

DR. HIRSCH: I'll take either.

DR. COHN: Well, we showed you the -- can we go back to the slide that breaks --

DR. HIRSCH: Not hospitalization.

DR. COHN: -- breaks down the morbidity endpoint?

DR. HIRSCH: The breakdown of the morbidity endpoint.

Yes, ideally, in other words, Jay, without the use of inotropes or vasodilaters. We're looking for raw data.

DR. COHN: No, I guess those data were not captured. These are the primary endpoint data. This is the number of patients who got intravenous therapy that were not hospitalized and met that criteria, and you can see there's only five in each treatment arm.

We really did not capture such things as patients coming to the ER not feeling well and being given an antibiotic or an extra dose of oral lasix and then going home. We did not capture that because we wanted to be rigid and maintain a very high standard for what represented true worsening heart failure or events equivalent to a hospitalization.

ACTING CHAIRMAN BORER: Paul and then Steve.

DR. ARMSTRONG: Again, Jay, I'm not trying to be difficult, but the briefing book has said that the endpoints, the morbid endpoints at least between May 3rd and the completion of last patient, last visit were recorded by the investigators and not adjudicated.

DR. COHN: That's right.

DR. ARMSTRONG: You've said that they are adjudicated.

DR. COHN: No, no. Anything after May 3rd was not adjudicated. I don't know how more clearly to say that. Everything up to May 3rd --

DR. ARMSTRONG: Thank you.

DR. COHN: -- and all the data you've seen is everything that happened in the trial up until May 3rd. The duration in which patients stay on drug after May 3rd varies, of course, depending on when they're able to get back and visit with their health care provider and be taken off of their therapy and a decision made what treatment they're going to go on.

And there were events that took place there, and Novartis is obligated since these patients are in a protocol and they're still on test drug; they're obligated to monitor those events, but there was no purpose in adjudicating them because they were not part of the primary analysis.

ACTING CHAIRMAN BORER: Steve and then Tom.

DR. NISSEN: Yeah. Jay, I agree with you that going to the emergency room should not have been the primary endpoint in the trial. I think that the right endpoints were used. The problem that we're having is that, you know, patients with heart failure make frequent trips to emergency departments. They use health care resources to do so, and collecting the data and reporting it for purposes of further understanding the benefit and risk of the drug would have been greatly helpful to us. This is kind of a message to people who do such trials.

I mean, I can understand why you would not want to adjudicate those as a heart failure event, but what if, you know, there were, you know, more trips to the emergency department by patients taking the active drug versus the control? That would suggest that there was a general safety disadvantage to the therapy.

And so if that's not captured, we have no way of knowing about it, and that's sort of what people are saying. That's more of an editorial comment than a question because I know we don't have that data. We don't know how many patients made a trip to the ED.

The other reason for the discomfort is that if you look -- as I look carefully at the data, the risk ratio for hospitalization was substantially lower for the active treatment arm by the endpoint committee than it was by the investigators.

So in the process of going through the adjudication process, there was a -- if you go on the briefing document on page 99 for the committee members, what you see is there was a 27 and a half percent reduction from the rest --

DR. COHN: Yeah, here's the data actually, Steve.

DR. NISSEN: Yeah.

DR. COHN: These are what were defined as heart failure related hospitalizations. You can see the endpoint committee eliminated a lot in both the placebo and valsartan arm. So they were much more meticulous about the criteria for heart failure hospitalization.

And here was the endpoint committee's adjudication showing a 27.5 percent reduction and a p value with four zeros.

The investigator -- which was not the primary endpoint. Remember that the protocol said this is the primary endpoint, but if we went by the investigator assessment, there was still a significant reduction, but it was 16 percent rather than 27 percent.

DR. NISSEN: Right, and so that's why we're focusing on closely on the adjudication process, and I think you used the right endpoint. I'm not disagreeing with that at all, but I'm trying to understand why there was such a substantial difference.

I mean, the benefit of the agent was nearly twice as great if one looks at the way the endpoint committee looked at it versus the way the investigators looked at it.

DR. COHN: I don't know. I mean, I can't answer the question.

DR. NISSEN: Yeah, I know you can't, and that's why we're being so nitpicky in understanding it.

DR. COHN: You know, unfortunately it was significant in both instances.

DR. NISSEN: Yeah, yeah.

DR. COHN: Obviously, and I've been through this many years, Steve, as you know. Investigator assessment of mechanism of death and of reason for hospitalization is seriously variable from investigator to investigator, and the reason we set up an endpoint committee is to have some uniformity in the way we will adjudicate these things.

And when you do it uniformly, you're right. I have no understanding of why there would have been a preferential effect, except that that's, indeed, what we would have anticipated, that if you use much more stringent, uniformed criteria, we'll find the benefit of valsartan.

If you just looked at all hospitalizations, which we did and I showed you that slide, the difference is not statistically significant. So it was very important to adjudicate and to identify what is identified as worsening heart failure hospitalizations, and that's not what the investigators did.

The investigators, and many of them put patients in the hospital for reasons that weren't related to worsening heart failure, and they just checked the box that said "heart failure." And we did it much more carefully. This is very casual.

DR. NISSEN: You don't have to convince me that adjudication is important. But the trigger for many of us to look more closely at this is this fairly substantial disparity between the investigator report and the adjudicated endpoints.

DR. COHN: Well, I think the message is that it's very important to adjudicate.

ACTING CHAIRMAN BORER: Tom, and then after that we'll move on to the next topic because we're falling a little behind.

DR. FLEMING: Jay, just a quick question. You had mentioned that the trial had been powered targeting a 20 percent reduction in the mortality endpoint. Can you tell us what the targeted reduction was in the --

DR. COHN: What the target what?

DR. FLEMING: Could you tell us what the targeted reduction was in the morbidity endpoint?

DR. COHN: Well, we didn't power it for the morbidity endpoint. We knew there would be many more events, and we knew that we would be well over powered for morbidity. So there was no calculation made.

The monitoring was based upon mortality so that we powered the trial for a number of deaths to identify that mortality reduction, and the DSMB monitored mortality only, not morbidity.

MR. CHIANG: Tom Chiang, Novartis.

Yea, the sample size space on the calculation, but we did assess the potential power for the morbid endpoint also as a primary, and the powers are enlarged. You know, a certain percent reduction would have more than 80 percent power.

DR. FLEMING: Well, I don't want -- that's in retrospect. I was interested in what your prospective targeted interest was, and it was 20 percent reduction and death, and in morbidity it was clearly --

MR. CHIANG: We did not talk, as I say. As I say, sample size is based on the mortality as the Dr. Cohn mentioned. We did not, say, target which percentage reduction for morbidity, but we did calculate barriers reduction to get a feeling to feel comfortable

DR. FLEMING: So there was no clinical sense. This was one of your two primary endpoints. There was no preplanned clinical sense of what magnitude of effect you wanted to get on morbidity, and it was one of your two primary endpoints?

MR. CHIANG: Well, we did, as I say, we did try to calculate there as possible. So we tried to insure --

DR. FLEMING: Good. So what were those various possibilities? What were they?

MR. CHIANG: As I say, beyond ten percent we assess all possible power, and for 13 percent because that just give you an idea what is power calculated for.

DR. FLEMING: So when the study was planned, you had planned a 13 percent reduction in morbidity?

MR. CHIANG: It's not planned for that. We calculated various case to feel comfortable. If this case happen, we do have sufficient power.

ACTING CHAIRMAN BORER: Okay. Maybe we can move on to the safety data, and we'll get back to any other clarifications a little later.

DR. COHN: We've got to go into the subgroup stuff first.

ACTING CHAIRMAN BORER: Oh, sorry.

DR. COHN: That comes next.

PARTICIPANT: Unless you don't care about the subgroup.

ACTING CHAIRMAN BORER: No.

DR. COHN: If you want to disregard it, we'll just disregard it, but --

ACTING CHAIRMAN BORER: It's your presentation.

DR. COHN: Okay. Now, as you're all aware, when one does a large scale trial like this, one often assesses subgroups to convince oneself that there is homogeneity among various groups because we're dealing with a very widely divergent population. So it's one of the dutiful things that we all do to look at this kind of a plot of the primary endpoint which was favorable, the morbidity endpoint, and we look at a number of baseline demographics, for instance.

This was the point estimate and the confidence intervals for the overall study favoring valsartan. This is in younger and older patients. This is in males and females. This is in whites and blacks. And you'll notice the only point estimate that goes to the right of that line is the black population. It's only a modest size population and with very wide confidence intervals, but we certainly were unable to convince ourselves that we had demonstrated efficacy in the black patients in this study.

This was the other racial groups. This is the U.S. and the non-U.S. So pretty close consistency for all of these groups.

Now, what about etiology of disease and severity of disease? Here was ischemic heart disease and those without ischemic heart disease. Here are diabetics and non-diabetics. Here's New York Heart Class II and III and IV. Here's ejection fraction above and below the median of 27. Here's ventricles smaller than and larger than the median ventricular size, you know, and for the most part there seems to be no striking difference among these groups.

Yes, maybe those with less severe heart failure, that is, a higher ejection fraction and smaller hearts, don't exhibit quite as much benefit as those who have more severe disease, but that's not terribly surprising.

So no inconsistencies.

However, there was a clear baseline difference in therapy, and it's really appropriate to look at background therapy using an angiotensin receptor blocker on top of ACE inhibitors and beta blockers, mandated that we look at that, as to whether that's influencing outcome, and of course, that's not a continuum. That's a yes/or.

What drug were you on at baseline? And this is the analysis that we did. Now, this was not preordained, and we did stratify for beta blocker use. I didn't point that out in the methods, but we did stratify for beta blocker use with an intent to make certain that we had an equal distribution of beta blockers in the two treatment arms, not because we expected necessarily any interaction.

We didn't stratify for ACE inhibitor, but 93 percent of the patients were on an ACE inhibitor. But it is a yes/no answer. So what about the patients who were not on an ACE inhibitor? There were 366, and you can see that this favored valsartan. This is mortality now, not morbidity. This is the mortality issue, and I show that for a really distinct reason.

So in mortality there appeared to be a trend here for a benefit of an ACE inhibitor. Those patients getting an ACE inhibitor, okay, close to the line. Those patients not on a beta blocker here, those patients on a beta blocker favored placebo, and that appeared to not overlap neutrality here.

Now, this is mortality. We did not find a benefit on mortality. So looking at these subgroups is perhaps not entirely appropriate, but we felt it important to do it, and we seem to see a clear trend for a benefit on mortality in the patients not on an ACE inhibitor, and a worsening mortality in those receiving a beta blocker.

But then we realized, well we have to look at this in more detail because there are four subgroups. There are those who are neither an ACE inhibitor nor a beta blocker, and here was their point estimate.

There are those who are on an ACE inhibitor, but not on a beta blocker, and that's their point estimate. There's those on a beta blocker but not on an ACE inhibitor, and here is their point estimate favoring valsartan, and there are those who are on both ACE inhibitor and beta blocker, and this is the group that seems to exhibit a worsening mortality with a risk hazard ratio of over 1.4.

Now, the interaction p value for overall interaction p value is .0091, and Tom knows better than I, but it's kind of hard to get interaction p values of that significance. So we thought this was something we really couldn't disregard.

Now, let's look at the morbidity, which was the endpoint favorably affected by valsartan. Now, here you've got the patients not on an ACE inhibitor, highly significant benefit. Even the ones on an ACE inhibitor you can see their confidence interval just touches the neutrality line.

Here are patients not on a beta blocker. here are those on a beta blocker, the trend in the wrong direction, and when we look at the four subgroups now, those on neither neural hormonal inhibitor, risk ratios down close to .5; those on an ACE inhibitor not on a beta blocker, still a highly significant benefit of valsartan; those on a beta blocker but not on an ACE inhibitor, there are only 140 of them, but clearly the point estimate trending favorably toward valsartan; and those on both background drugs, point estimate clearly on the placebo side.

And here the interaction p value was .0011. So can we disregard this? Well, we can't for two reasons: one, that there is this highly significant interaction, and it's not terribly surprising.

And, second of all, there's the safety issue because on mortality, the group taking both an ACE inhibitor and a beta blocker exhibited a statistically significant worsening of mortality.

So we felt we could not at all disregard this subgroup analysis.

Now, this is the actual data, which may help you a little bit in the same four major groups, and then the four subgroups formed by the use of the one or both drugs, and here you can see on morbidity here that the p value favoring valsartan was highly significant in all three groups except the one taking an ACE inhibitor and a beta blocker, where the trend not only went in the other direction. The p value wasn't significant, but it was trending adverse.

So these three groups all exhibit by themselves a significant reduction of morbidity.

Now, these are all data based on administration of drugs at baseline. How close does this correlate with the maintenance of these drugs during the study? Because that's of importance. And this data attempts to show you that.

Of the patients on an ACE inhibitor at baseline, 90 percent were still taking the ACE inhibitor at the end of the study. On those with a beta blocker, 92 percent were still on the beta blocker at the end of the trial. That's true of all the drugs except spironolactone in which there was a reduction of the use of the drug by the end of the study.

In those patients not on the drug at baseline, only 16, 13 percent went on the drug during the trial. More patients who were not on a diuretic started diuretic during the trial, which is expected because heart failure worsens, and they then finally required diuretic.

I find it surprising that 12 or so percent or 15 percent of the patients weren't on a diuretic even at baseline because heart failure almost always requires a diuretic.

So pretty good congruence between what the therapy was at baseline and what the therapy was at the end.

Well, one of the groups we said at the beginning we wanted to look at because there are no data available in the literature prospectively is the group not on an ACE inhibitor. Now, a third of these people were on a beta blocker, but they were not on an ACE inhibitor, and one of the important questions is: is valsartan a substitute for an ACE inhibitor in patients who don't tolerate an ACE inhibitor.

So this was only seven percent of the population, 370 patients. This is the mortality curves which separate pretty early and widen over time. That's a 33 percent reduction in mortality. The p value is .017.

Here is the morbidity in that population. A 44 percent reduction of morbidity, the curves really widen out over time. The p value here is 002. It's only 370 patients, but it's the first demonstration prospectively that I know of that one can use an ARB, specifically valsartan, and exert the kind of favorable effect we've associated with ACE inhibition.

Now, we also wanted to look at all of the secondary endpoints to see if they were congruent with our clinical outcome data, suggesting that that one subgroup didn't do very well. Now, these are independent measurements. They have nothing to do with hospitalization or death. These are completely independently measured secondary endpoints.

This is left ventricular ejection fraction, and here are the four subgroups if you will, people taking neither an ACE nor a beta blocker, those taking an ACE inhibitor, not a beta blocker, those taking a beta blocker, not an ACE inhibitor, those taking both.

You will notice that although this group was very small, there was certainly a trend for much greater benefit of valsartan and placebo on ejection fraction. This group highly significant, this group highly significant.

Ah-ha, here's our little culprit group. No benefit on ejection fraction.

If we look at just the groups leaving that group out, the difference is highly significant, and let me show you that.

Here then is the ejection fraction change in the patients who are -- based upon their use of beta blocker, ACE inhibitors. Here was the overall EF change, .00075, benefitting valsartan -- favoring valsartan over placebo.

Here are the group getting both a beta blocker and an ACE inhibitor, the absence of a benefit. If we take that group out and look at the other subgroups, the benefit goes up to .00002 on ejection fraction.

Here is left ventricular internal dimension by echo. You'll see a benefit here, a benefit here, a benefit here, all three subgroups exhibiting a significantly greater reduction of their ventricular size with valsartan compared to placebo. Here is our culprit group. No difference.

So here was the overall effect on left ventricle. Here is the lack of effect in that subgroup, and actually a greater difference now in this smaller group that excludes this one group over here.

Here is the living with heart failure score, independently measured by the patient. Nobody has intervened to influence them.

Here's the benefit in this group, this group, and this group, all of them exhibit rather striking greater worsening of heart failure in the placebo than the valsartan group.

Here is our culprit group. No difference. If we take that group out, the difference gets even greater, and I'll show that in the next slide. Here it is.

Here was the overall benefit on living with heart failure score. Here is the lack of benefit in that one subgroup, and here is the benefit on the residual patients now with a Z of 0002 p value.

What about New York heart class? The same thing is true. There was overall a benefit of valsartan compared to placebo, more improvement, less worsening.

This group over here, no benefit. When you take them out, the p value goes to 000003.

Now, what about the overall mortality? The mortality overall we said neutral, heart risk ratio of 1.02. Here was the risk ratio in the group taking both an ACE inhibitor and a beta blocker, 1.4, which was statistically significant in that subgroup.

When we take that group out, now we're seeing a trend for a favorable effect on mortality, a risk ratio of .92. So beginning to bring out what might be a favorable effect of valsartan even on mortality, obviously way under powered to pick anything up, but there it is.

And what about morbidity? Well, the morbidity --

DR. FLEMING: Could you clarify, Jay, when you said "way under powered," go back to that slide?

DR. COHN: Well, I said we were under powered to pick up an eight percent difference, which is what that slide showed.

DR. FLEMING: Oh, okay. It's not under powered to pick up a meaningful difference. It's under powered to pick up a really small difference because most --

DR. COHN: Well, I don't know what --

DR. FLEMING: -- you've got 30 --

DR. COHN: I would object to your use of the word "meaningful."

DR. FLEMING: -- three hundred patients in that subgroup.

DR. COHN: I think an eight percent reduction in mortality is probably meaningful.

DR. FLEMING: The study was targeting a 20 percent reduction.

DR. COHN: That's right. Yeah. I'm saying we're under powered to pick up the difference that we found.

DR. FLEMING: Well, let's come back to this. Keep going.

DR. COHN: Okay. And now, this is the morbidity endpoint. Remember that the p value for our primary endpoint was .009. In that one subgroup, the trend was in the other direction, the p of .104, and when we take that group out and look at all the other subgroups now, the p becomes 00003, and the risk ratio is reduced to .785.

So we now have to cope with this subgroup that we can't disregard, and in fact, if we think this is an important subgroup that should not receive the drug, and I do believe that that is the case today, until more data are accumulated from other trials, and we say these patients we're not going to treat, we're left with the rest of the patients in whom the statistical significance of the data is far more dramatic.

So in summary, I believe that we can conclude that the benefit on morbidity that we've observed in the overall trial was seen particularly in patients on neither ACE nor beta blocker or on ACE inhibitors or beta blockers, but not in those patients receiving both drugs.

Thanks.

ACTING CHAIRMAN BORER: Any clarification of fact questions for Jay on this section?

No, sounds like you -- oh, you do. I'm sorry, Tom. Go ahead.

DR. GLAZER: We also do have Dr. Carson who has joined us if there's further questions for him.

DR. FLEMING: Actually I'll wait until the end.

ACTING CHAIRMAN BORER: Do you want to hear from Dr. Carson about the unresolved issue here?

DR. COHN: Perhaps we do. Steve?

DR. NISSEN: Shall we go back or do you want to go forward? Do you want to talk about this part or do you want to go back to --

ACTING CHAIRMAN BORER: Well, why don't we start with this? And then we'll go back and test the other.

DR. NISSEN: Okay. Jay, there was another breakdown that I didn't see in there that I'm actually very interested in, and that was the U.S. versus non-U.S. Clearly there was a much greater benefit almost across the board in the non-U.S. population, and I wonder if you would help us understand that.

DR. COHN: Yeah, let's look at that. This is the morbidity and mortality in the U.S. and non-U.S. populations, and although the hazard ratio was, indeed, a little -- it was different here. Of course, this is mortality, not morbidity. This was our primary endpoint in which the non-U.S. had a slightly lower hazard ratio than the U.S., but the confidence intervals are really almost entirely overlapping.

And for first heart failure hospitalizations, again, there is a difference, but once again, the confidence intervals are really overlapping. The interaction p value is very high.

So I think there is no real geographic -- evidence for a geographic difference here.

DR. LINDENFELD: The percentage of patients on beta blockers in the U.S. and non-U.S., is there --

DR. COHN: Very close to the same. It was a little higher in non-U.S. than in U.S., but they were both within the 30 percent range.

DR. NISSEN: Yeah. Actually the data here are slightly different from the data that we have from Dr. Targum's review, but in the FDA book, CHF hospitalization, the risk ratio in the U.S. was .81 and the risk ratio in the non-U.S. was .67. You know, from .81 to .67.

I asked the question, and I'll tell you why I asked it, Jay. This has got to be the fourth or fifth trial I'm aware of, major, mega trial where the benefits were substantially greater in the non-U.S. population than in the U.S. population, and it's something that has been troubling many of us because obviously this agency regulates the use of drugs in the United States.

And I have my own hypotheses here, which maybe later on we can talk about, but any insight here would be useful because there really does appear to be an across-the-board difference if you look carefully at the data in the U.S. versus non-U.S.

Any insight?

DR. COHN: Well, I mean, we've obviously thought a lot about this. I think it was far more pertinent in a previous trial where this became a major issue, as you know.

There is a somewhat higher incidence of hospitalizations in non-U.S., which probably reflects the health care system and much resistance in the United States for hospitalization. So one possible explanation would be you're more likely to be hospitalized when you're in Europe, and therefore, a benefit of therapy might be more demonstrable in the European population. It might be a more sensitive marker.

Many have raised the issue of African Americans because you can see from this data, and you know that I've had a great interest in possible differences based upon African American and white patients, and one possible explanation is the impact of African Americans, but in this trial I think the number was too small to impact on that.

So we don't have any rational reason for seeing a difference. Genetically there's not a major difference between Europe and the United States. So I must say that I don't have a good answer to that.

ACTING CHAIRMAN BORER: Tom.

DR. FLEMING: Just a couple of questions. Jay, in your introduction, you had given some of the motivation for the interest in valsartan in the context of what you might already be able to expect to achieve with ACE inhibitors, and in the briefing document, the sponsor has indicated that the combination of angiotensin receptor blockers and ACE inhibitors may be synergistic by providing the more complete inhibition of the renin angiotensin system through the blockade of the AT-1 receptor, which was exactly the presentation that you have given.

If we look then, in particular --

DR. COHN: I don't think we used the term "synergistic," but okay.

DR. FLEMING: This is exactly a quote. I'm quoting exactly your briefing document.

DR. COHN: Does it say --

DR. FLEMING: "May be synergistic."

DR. COHN: Well, I guess I'd argue about that.

DR. FLEMING: Looking at your data specifically in the group of patients that were on ACE inhibitors, which was 92 percent of the population, when you look there, what you find is an eight percent reduction in the morbidity, but a seven percent increase in mortality.

Essentially then do these data fairly strongly argue against a synergistic effect in the presence of an ACE inhibitor?

DR. COHN: Well, in the absence of beta blocker, now, you can't talk about ACE inhibitors any longer without bringing the beta blocker in because it's another neural hormonal inhibitor. In the absence of a beta blocker, the efficacy of ACE inhibitor, of valsartan on top of ACE inhibitor was greater than that in terms of the morbidity endpoint, and that represented the largest segment of the population, two thirds of the patients.

So I don't think you any longer can look at a subgroup saying ACE or no ACE when there's a beta blocker in a third of the patients who are on the ACE inhibitor or not on the ACE inhibitor.

Do I think that there's a synergistic effect? I think that's a term that -- that's a pharmacologic term that I'm very hesitant to apply to this, and I guess I had missed the fact that that word was used in the briefing document.

"Additive" would be the word that I would have used, and I believe we have demonstrated an additive effect, and that would be the only word I would have used in describing our proposal.

DR. FLEMING: But you only argue that by subdividing out those people who were also on a beta blocker. That's your answer. Your answer is, yes, you believe it's additive, but only by subdividing or eliminating those people who also received a beta blocker.

DR. COHN: No. The question is: does adding valsartan to an ACE inhibitor have an additive effect?

DR. FLEMING: That's correct.

DR. COHN: We know it has an additive hemodynamic effect. That's been studied in 103 and 104. We know it has a neural hormonal effect. That was studied in 104, and it's also been studied in 107.

The question is does it have additive benefit on morbidity and mortality, and the answer --

DR. FLEMING: On your primary endpoints, correct.

DR. COHN: On the primary endpoint.

Well, on mortality we haven't shown an additive benefit. So at the moment we can't say that it has.

Does it have an additive benefit on morbidity? Yes, it clearly did, especially when one takes out the beta blocker group.

DR. FLEMING: Not especially when. Only when.

DR. COHN: Well, the point estimate favored here, I mean, and let's look at those patients on an ACE inhibit. The hazard ratio is .9, and the p value is .096.

Now, you can argue whether that means we did or didn't have an effect, but we have produced a ten percent reduction of morbidity when you add an ACE inhibitor to valsartan regardless of beta blocker use.

Now, ten percent reduction of hospitalization rate is not infinitesimal. It's not small. It's fairly substantial when you think of the number of hospitalizations.

DR. FLEMING: I'll go back to what I said before, and I'm quoting from the FDA briefing document on pages 102 and 103. "If you look at morbidity, the relative risk given there is .92." You're giving it as .90, but an eight to ten percent reduction in morbidity with use of ACE inhibitors, but a seven percent increase in mortality.

So the same magnitude of mortality increases you see in reduction in morbidity when you look at the biggest group of patients in the trial, which are those on ACE inhibitors.

ACTING CHAIRMAN BORER: Jay, can I --

DR. COHN: Well, I can -- can we put up the mortality slide, the comparable --

ACTING CHAIRMAN BORER: Jay, before you answer the question --

DR. COHN: Yeah.

ACTING CHAIRMAN BORER: -- let me just introduce a concept that I'd like you to deal with as you respond to this.

You know, there were background ACE inhibitor therapy, but the doses varied. There certainly is no suggestion that background therapy was titrated to maximally tolerated dose of ACE inhibitor. Different ACE inhibitors were used, one and on and on. You know, you can response, and you should respond because the question was asked, but it seems to me that we don't have a data set that allows us to discuss whether an angiotensin receptor blocker is additive to ACE inhibitor because the trials weren't set up to answer that question.

All we can say is that in this population getting these number of drugs at these number of doses, when you added valsartan and there wasn't a beta blocker on board, that we see this.

I mean, is that a reasonable --

DR. COHN: Yeah, I showed you the mean dose of ACE inhibitors used, which, you know, is like 18 milligrams a day of lisinopril. so that on average it's close to target dose.

Is there a differential response based upon how much ACE inhibitor the patient is getting? You know, you get into very --

DR. FLEMING: Sure.

DR. COHN: -- small subgroups, and this is --

DR. FLEMING: You just can't answer that.

DR. COHN: -- the analysis we plan on doing, but all we're saying is this is the kind of -- this is good therapy. I think if you went to the community at large, the dose of ACE inhibitor would not be that high. This is the best doctors in the world treating patients the best way they know how.

Now, if you add to that valsartan, do you further improve the outcome? The answer is yes, with the provision that there seems to be one subgroup that doesn't get benefit.

Now, the mortality, the issue that Tom raised -- can we go back to that mortality slide? Because you raised this, and I want to show you the data.

This is those same subgroups based on mortality, and what Tom is saying is that there was a 1.055 hazard ratio on mortality when you or on an ACE inhibitor. However, when you go down here and you look at the group that was on an ACE inhibitor and a beta blocker, it's 1.42. When you look at the group that was on an ACE inhibitor without a beta blocker, it's .959.

So the adverse trend here, which is clearly not significant -- it's a five percent increase -- appeared to be entirely related to this group that was also taking a beta blocker.

DR. FLEMING: Yeah, in the briefing document we have on page 103 ACE inhibitor use, all cause mortality, relative risk of 1.07, numbers similar to, but not exactly the same as what you have.

DR. COHN: Yeah. I'm not sure. I guess the --

DR. FLEMING: And so if you're looking at all of the ACE inhibitor patients, there is a seven percent increase. As you point out, if you further subdivide, and that is controversial, those on beta blockers would have a 42 percent increase. Those off would have a four percent or six percent decrease.

DR. COHN: Right, and the numbers to make that -- I suppose we should clarify. Perhaps somebody from the FDA should clarify why their numbers in the briefing document differ from ours.

I believe it's they did not use the covariates in doing the analysis which were prescribed in the protocol to adjust for covariates. Maybe we could hear from the FDA.

MR. HUNG: Jim Hung, FDA statistician.

The number I have is unadjusted as a ratio because the primary test for all these primary endpoints, morbid events, are low rank tests, and so, therefore, I try to be -- try to use the numbers to be consistent with the test.

The sponsor's numbers are adjusted for the covariates. That's the difference.

DR. COHN: This was in the protocol, prespecified adjustments of covariates and the Cox regression --

DR. FLEMING: And this is a peripheral issue. The main answers are the same.

DR. COHN: Yeah, it comes out the same.

DR. FLEMING: The main answers are if you look at those on ACE inhibitors, which is 92 percent of the cohort, you see a very small increase in morbidity and a very small worsening of mortality, a net comparable result.

ACTING CHAIRMAN BORER: Paul.

DR. ARMSTRONG: Jay, although the number of blacks was small, it was comparable to the subgroup analysis in size of the patients not on ACE inhibitors. So I wanted to ask you a series of questions related to that population.

First, from the knowledge that you or the company has, what is the information on the anti-hypertensive effect of valsartan in blacks?

Second, do we have information on the hormonal and the morbidity data in blacks vis-a-vis whether this trend, if you will, which is not significant on mortality, is supported by some of the other ancillary measures that you so eloquently discussed in the other subgroups?

DR. COHN: Malcolm can probably answer the question.

MR. MacNAB: I can answer the first one. The package insert for valsartan states that the efficacy in whites and blacks for hypertension is about the same. That's what the database that we have shows.

ACTING CHAIRMAN BORER: What about the renin angiotensin profile?

DR. COHN: Yeah, let me answer your first question here. Can we get that? Yeah.

This is the breakdown, if you wish, for plasma norepinephrine mean change from baseline, and you'll notice that the blacks appear to exhibit the same difference as do the whites in terms of lowering norepinephrine.

So from a hormonal standpoint there doesn't appear to be a racial difference in response.

DR. HIRSCH: And was there a mortality difference in the subgroup?

DR. COHN: In the blacks there was a trend in the other direction, right. None of it was significant; very wide confidence intervals.

ACTING CHAIRMAN BORER: Do you have -- I mean these are norepinephrine. Do you have renin data on the --

DR. COHN: No, we have norepinephrine and BNP only. That's the only two hormones that we monitored, and the BNP also exhibited, I think, the same.

Well, here. Okay. We can give you this, too. Boy, my team has more slides there.

All right. Here's the BNP data just for completeness. Now, with BNP there did appear to be a difference. That is, all these other groups show a favorable effect of valsartan compared to placebo. In the black patients that was not true. So now we find a differential between BNP and norepinephrine.

I mean, you can chew on this all you want to. I don't know what to make of these data.

And then we also have the mortality data that was addressed. I'll give you that data directly. Can we have that mortality slide and the racial breakdown?

Here we are. This is black patients. This is mortality. You can see the trend was in the wrong direction. Obviously wide confidence intervals. This is morbidity, a trend in the wrong direction, and these were in the non-black patients in terms -- and mortality obviously.

Now, turning down to .997 and morbidity being strikingly better. So we obviously have no answers. This is retrospective analysis.

ACTING CHAIRMAN BORER: Okay. Why don't we hold any other issues for Jay for his next presentation?

And we'll move on now to the safety --

DR. NISSEN: Jeffrey we going to --

ACTING CHAIRMAN BORER: Oh, I'm sorry. Just one moment.

DR. NISSEN: Are we going to come back to the adjudication?

ACTING CHAIRMAN BORER: Yeah, right. What I was just remembering is that we have the adjudication issue that we can deal with. Why don't we take care of that now?

DR. NISSEN: Before you got here, we all had a number of questions, and I want to see if I can focus on some of them. Since endpoints were not independently adjudicated, we want to understand the process, and I guess I'd like to understand who identified the cases for adjudication, what triggers did they use, what source documentation was provided. You know, was it a narrative? Did you review the charts? Did you get excerpts from the charts?x

And then the third question I had was I present a hypothetical case of somebody that came in at nine o'clock in the morning with an exacerbation of heart failure, was admitted, got intravenous furosemide for the day, and then that evening went home; whether that would have been considered a heart failure hospitalization.

DR. CARSON: Let me just say, first, I'm sorry to make you all go back on this. I must say it is probably easier to fly to the NIH than to drive across town to here.

You started off with a comment that they were not independently adjudicated.

DR. NISSEN: Right.

DR. CARSON: I'm not sure I understand.

DR. NISSEN: Well, the sponsor presented cases to the committee for adjudication. The adjudication committee was not independent of the sponsor.

Independent adjudication means that somebody who's independent of any other interests in the trial, you know, gets source documentation, does adjudication. That's not what happened here.

DR. CARSON: Okay. I don't know of any endpoint committee that has been associated with any trial that has been independent in a way that we weren't, and I actually have done more of this than probably anybody else.

DR. COHN: Can you speak up, Peter? I don't think everybody can --

DR. CARSON: Well, I'm just saying I don't know of any way that this committee was not independent than any other committee associated with any large trial has not been independent. We were blinded. None of us were investigators in the trial, and the materials were provided to us, it is true, by the company, but this is as independent as any committee I've ever been associated with.

In terms of the materials that we received, we required that there be primary source documentation. So we asked for some hospitalization record, particularly a discharge summary. We often received a discharge summary and an initial history and physical exam narrative. We sometimes received hospitalization data.

Now, from the U.S., that data was stronger because that data is easier to find, and in some countries, as I understand it, hospitalization records really cannot be obtained, and in those cases then we relied on a letter from the physician taking care of the patient to another physician, referring physician, for example, or a physician involved in enrolling the patient in the trial.

So there was always primary documentation obtained. There was also a narrative from the physician that described the event, whether it was a death or a hospitalization. So there was primary source documentation on virtually every patient unless towards the end of the trial multiple attempts had been made and there simply was no response.

In terms of the specific patient that you posed, we realized, I think, after the first meeting we had put in a barrier that said in order to be a heart failure hospitalization you had to have been in the hospital for longer than 24 hours because we assumed that would make it then a serious event.

And we found relatively early that it was very difficult to get exact times, and so somebody went back and said, "Well, how did you know it was 24 hours?"

Getting the exact times turned out to be quite difficult, and this has been true in other trials also.

So we then put in the barrier of saying that we may not know the exact times, but if the calendar date changed, we would take that as being a surrogate for a 24-hour admission. So if the calendar date changed, that's what we took.

Now, did it happen that there were patients that were admitted at six o'clock in the morning, got an IV dose of diuretic, and went home at nine o'clock at night?

It may have happened. We would not have taken them as being a hospitalization. They not have had a calendar day change. I have to say in my own experience in heart failure that is a very unusual circumstance, and --

DR. NISSEN: You would not have even seen those charts then? They wouldn't have come to you?

DR. CARSON: I'm not sure whether we would have seen them or not. I think early on we may have seen them, and we may have then said it didn't meet 24 hours. So we don't need to see this one.

Now, you did --

ACTING CHAIRMAN BORER: Those require an assay.

DR. CARSON: I would also answer one part of your question, I guess, which was what about the cases that did not come to us. We looked at the work load of the committee, and we adjudicated something on the order of was it 4,100 events? It was just a large number of events.

Initially we were seeing all hospitalizations. Initially we were seeing second hospitalizations after the initial heart failure hospitalization.

But then it became apparent that the work load was extreme and that it was just not going to be possible to gather members together, have meetings, and actually adjudicate cases.

Now, you can sit and do five an 600 and 800 cases in a day, but you're just turning pages, and so in order to actually have discussion of cases and consideration, we felt the work load had to be altered a little bit, and one alteration was that after a patient had a first heart failure hospitalization, we did not adjudicate after that. They had met the primary endpoint.

But secondly, patients who had a clear non-cardiovascular cause for hospitalization, we did not see them. The first couple of meetings we did see them, and we became confident as a committee that the sponsor could recognize a non-cardiovascular hospitalization.

And then each meeting I would get a list of the non-cardiovascular hospitalizations, and if there was anything that had a primary cause from the investigator that looked like it might have involved heart failure, then we said, "Please send us that," and we got it.

ACTING CHAIRMAN BORER: Does that -- Paul?

DR. ARMSTRONG: I just would like to follow up and also make a comment. For some of us, this adjudication process has involved evaluating data triggered by a data organization that is independent from the sponsor. So I think the spirit of Dr. Nissen's question and my comment apropos of your issue, I'm not questioning the integrity of your committee. Our question was what triggered your opportunity to evaluate things independently, and there are different models, and you obviously used one, but there are others.

As I understand it from the briefing booklet, at some point there was an addendum to your committee perhaps driven by the work load in which over diuresis or drug toxicity were perceived as hospitalization for reasons other than heart failure.

And I just want to be 100 percent certain then that someone with syncope due to dehydration, digitalis intoxication or hyperkalemia, all things that we as clinicians care for in our heart failure patients, would be a hospitalization that would be perceived not related to heart failure and you would not have seen.

DR. CARSON: No, we would have. We would have seen that. In fact, we were particularly sensitive to that particular area.

One of the two areas that we had a particular alarm bell about was the group of patients who could be poor perfusion and poor perfusion with hypotension, poor perfusion with renal insufficiency leading to digitoxicity, et cetera, et cetera.

That was a group that we were very concerned about, and so anything that even looked like that, if I saw that on the non-cardiovascular list, if I saw renal failure, if I saw anything cardiovascular we've got, but if I saw renal failure, I said, "I want to see that and make sure it's not poor perfusion."

The other group that we were very concerned about was the abdominal pains. Could that be passive congestion leading to abdominal pain? So that was another group that we really targeted. We had to see all of them.

DR. ARMSTRONG: Thank you.

ACTING CHAIRMAN BORER: Okay. Thank you very much.

If there are no other issues about the adjudication process, why don't we try and move on to safety data now? We'll have a little discussion afterwards.

DR. COHN: All right. Bob Glazer will present the safety data.

Bob.

DR. GLAZER: Safety of valsartan was evaluated in eight clinical trials worldwide that included over 6,000 patients with chronic heart failure. During this presentation I would like to summarize for you the valsartan safety database, including patient exposure, demographics, adverse events, including severe adverse events, and treatment discontinuations, and laboratory evaluations.

Valsartan safety database for heart failure consisted of eight clinical trials. There were four placebo controlled trials, one positive control trial, and three open label trials.

The primary data set includes the four double blind control trials with trial durations of up to four months and also the first four months of safety data from Val-HeFT. There were over 6,000 patients in this primary data set of which 3,289 patients received valsartan. The long-term safety data set is comprised solely of data from Val-HeFT. The focus of this presentation will be on these two data sets, namely, the pooled, short-term, primary data set and the long-term data from Val-HeFT.

There were only 94 total patients in the open label trials, and therefore, these data were not pooled for any analysis.

The five completed double blind trials consisted of two small hemodynamic studies, Studies 103 and 104; two trials whose primary efficacy variable was exercise tolerance, Studies 106 and 110; and the large morbidity and mortality trial, Val-HeFT.

All trials except Study 110 evaluated twice daily dosage regimens of valsartan which ranged from 40 milligrams to 160 milligrams twice daily, depending on the particular study. All trials were parallel design and included patients in New York Heart Association Class II, III, and IV, with the exception of Study 110 that excluded Class IV patients.

Most trials had an inclusion criteria for ejection fraction requiring patients to have ejection fractions less than either 40 or 45 percent.

In the primary data set, which included trials with the maximum trial duration of four months and also the first four months of Val-HeFT, over 3,200 patients were exposed to valsartan for at least one day. Over 2,700 patients were exposed to valsartan for at least three months.

Long-term exposure from Val-HeFT included over 2,100 patients for at least six months, approximately 2,000 patients for at least one year, and over 1,000 patients for two years or more. The majority of patients in each time category were exposed to the highest dose of valsartan, namely, 320 milligrams total daily dose, as these patients were force titrated to that dose in Val-HeFT.

Overall there were no clinically important differences in baseline characteristics between the valsartan and placebo treatment groups in the primary data set. The mean age of the studied population was 63 years of age. The number of patients above and below the age of 65 was similar in each treatment group. Approximately 15 percent of patients were over the age of 75 years.

Ninety percent of patients were white, and approximately eight percent were black. The majority of patients, 80 percent, were male.

The duration of heart failure in these patients was just over four years. The ejection fraction was 27 percent. The mean blood pressure was approximately 124 millimeters of mercury systolic and 75 millimeters of mercury diastolic. The majority of patients, 61 percent, were Class II and 37 percent were Class III.

Investigators were asked to select the one primary etiology of patient's heart failure in all but one trial. Coronary heart disease was chosen as the primary etiology in the majority of patients, 57 percent, followed by idiopathic cardiomyopathy, hypertension, and other miscellaneous causes.

The majority of patients were receiving ACE inhibitors, approximately 90 percent; diuretics, 85 percent; and digoxin, approximately 68 percent. Approximately one third of patients were receiving beta blockers and one third were receiving nitrates.

Patient disposition was collected differently in Val-HeFT than in the other trials. Therefore, these data were not pooled into the primary data set. The patient disposition data from Val-HeFT has already been presented by Dr. Cohn.

In the remaining trials shown here, patient disposition was similar to that observed in Val-HeFT. The most common reason for premature trial termination was an adverse experience occurring in nine percent of valsartan patients and four percent of placebo patients.

In the primary data set the incidence of adverse events regardless of trial drug relationship was 73 percent with valsartan and 69 percent with placebo. Dizziness and hypotension were the two most frequently reported events, and each occurred more frequently in valsartan treated patients than those patients receiving placebo. Dizziness and hypotension were reported in 17 percent and seven percent of valsartan patients, respectively.

The incidence of cough in this population was similar in both treatment groups.

Similar to this short-term primary data set, the two most frequent adverse events in Val-HeFT shown here were dizziness and hypotension, each occurring more frequently in the valsartan group compared to placebo.

The incidence of aggravated heart failure was less in the valsartan group compared to placebo. Diarrhea occurred slightly more frequently with valsartan.

In the primary data set, there were no clinically important differences among subgroups, including age, gender and race in the overall incidence of adverse events compared to the overall population. Slightly more female patients compared to male patients reported adverse events in both treatment groups.

Dose response information can best be assessed using the data from Study 106, which was a fixed dose, placebo controlled, parallel design trial that evaluated three different valsartan doses. In this trial, the overall incidence of adverse events in each of the three valsartan treatment groups was less than or approximately equal to that observed with placebo.

There was no dose related effect for the more frequently reported adverse experiences, including dizziness and hypotension. A suggestion of a dose response was observed for hyperkalemia, which was reported in one, three, and four percent of patients in the 80, 160, and 320 milligram dose groups, respectively.

Summarizing the data from the adverse event case report form in Val-HeFT, permanent treatment discontinuation occurred in 9.9 percent and 7.3 percent of patients receiving valsartan and placebo, respectively. A significant difference in rates of discontinuation occurred for dizziness, renal impairment, elevated serum creatinine, diarrhea, and hyperkalemia.

The overall discontinuation rates for adverse experiences in the other control trials were similar to that observed in Val-HeFT.

Both sitting and standing blood pressure were measured at all visits in Val-HeFT. At all time points post baseline there was a consistent and significant reduction in both sitting systolic blood pressure, approximately four millimeters of mercury placebo subtracted, and diastolic blood pressure approximately two millimeters of mercury placebo subtracted in the valsartan treated patients.

The incidence of serious adverse events in Val-HeFT were similar with valsartan, 51 percent, compared to placebo, 54 percent. The most frequent serious event, aggravated heart failure, and also atrial fibrillation were reported less frequently in patients treated with valsartan. No other significant differences between treatment groups were observed.

In the short-term primary data set, the occurrence of serious adverse events was generally similar in each treatment group. Again, aggravated heart failure occurred less frequently in patients receiving valsartan. Incidence rates for individual adverse events were generally less than one percent.

The most common cause of death in Val-HeFT as reported by the investigators is shown here, was sudden death, which occurred in nine percent of patients in each treatment group.

Pump failure was the next most frequently reported cause of death. In the remaining four control trials, the number of deaths was small, 14 in total.

Noteworthy laboratory changes observed in the valsartan heart failure program relate specifically to the pharmacology of inhibitors of the renin angiotensin system, and include changes in renal function.

Small increases in serum creatinine, potassium, BUN, and uric acid were observed in those patients receiving valsartan compared to placebo as shown here.

Specific criteria were prespecified to define clinically meaningful changes from baseline in laboratory parameters. For creatinine, BUN, and uric acid a 50 percent increase was defined; for potassium, a 20 percent increase or decrease was defined.

In the short-term primary data set, four percent of valsartan patients versus one percent of placebo patients had an increase in serum creatinine. Ten percent versus five percent had an increase in serum potassium, and 17 percent versus six percent had an increase in BUN.

In summary, valsartan in doses of 80 milligrams to 300 milligrams once per day was well tolerated in patients with New York Heart Association Class II, III, and IV heart failure. The adverse events observed were not unexpected and included dizziness, hypotension, and postural dizziness. Dizziness and hypotension were the most common reasons for discontinuation of therapy.

Laboratory changes included increases in creatinine, BUN, potassium, and uric acid.

Valsartan's safety profile in heart failure patients was consistent with the pharmacology of an agent affecting the renin angiotensin system and also the background therapies these patients were receiving.

ACTING CHAIRMAN BORER: Thank you very much.

Are there specific safety questions from the committee? JoAnn.

DR. LINDENFELD: I know there were a small number of patients on spironolactone at the start of the study, but can you give us some idea in that small number of patients what the incidence of adverse events was? Patients on both ACE inhibitors and spironolactone, and then between placebo and valsartan.

And I think this is an important point because aldactone is used substantially more now than I think it was in the baseline group of patients here. So it's something we have to consider.

DR. GLAZER: We have just started looking at this topic. We have some laboratory data which showed that there was an increase in the number of patients who made the pre-specified criteria for creatinine.

DR. LINDENFELD: I guess creatinine and potassium would be the two we'd be most interested in.

DR. GLAZER: Can I have Slide 020?

There was a small number of patients receiving background spironolactone, and as you can see in this slide, the percentage of patients who had an increase in potassium was essentially similar to that that was seen in the overall population. The patients who had an increase in serum creatinine are greater than 50 percent above baseline, was greater in those patients receiving spironolactone.

Similar incidence rates were seen for BUN and uric acid, again, compared to the overall population.

This patient population has to be -- we're further investigating the laboratory data and safety data.

DR. LINDENFELD: I guess it's important. We don't know how many of these were on ACE inhibitors and were not?

DR. GLAZER: I don't have that information.

DR. LINDENFELD: Because we might expect that those on both would have even a greater incidence. The incidence of withdrawal, I think, on Slide No. 7 shows withdrawal was much greater in patients on spironolactone.

DR. GLAZER: Correct.

DR. LINDENFELD: And it just becomes an issue because we're using spironolactone a lot more than the small number of patients that were in this trial.

ACTING CHAIRMAN BORER: What kind of monitoring recommendations is the company making with regard to electrolytes, BUN, creatinine? Any in your proposed labeling?

DR. GLAZER: Yes. As with the care of any patient with congestive heart failure, careful attention should be paid to monitoring BUN, creatinine and potassium in these patients who were concurrently receiving diuretics, beta blockers, as would be expected.

ACTING CHAIRMAN BORER: JoAnn?

DR. LINDENFELD: One other question. Can you tell us what the incidence of the requirement for dialysis was in the two arms?

DR. GLAZER: The number of patients who had dialysis actually was similar in -- this slide shows the number of patients requiring dialysis in the overall congestive heart failure patients, and you can see that there was 12 patients total receiving valsartan and 12 patients receiving placebo. All but one of those patients were in Val-HeFT. One patient was in Protocol 106.

ACTING CHAIRMAN BORER: Paul.

DR. ARMSTRONG: As I understand it, 15 percent of the patients were over the age of 75, and I wonder if you could develop for me any information you have regarding the relationship between some of the serious adverse events and obviously the hyperkalemia and the creatinine and any others that are of interest in that important very elderly population.

Do you have data on that point?

DR. GLAZER: We don't have any laboratory data or adverse experience data cut by the 75 year of age point.

DR. ARMSTRONG: And the second question coming back to the issue of efficacy in blacks was, again, the serious adverse events in the creatinine and the potassium issues in the blacks since we don't have the renin angiotensin system measured, but indirectly we might be able to get at it.

Do you have any information on that subgroup vis-a-vis serious adverse events and some of the metabolic factors?

DR. GLAZER: Can I have Slide 023? Actually it's AEO-17.

This information is adverse events by racial subgroups looking at the most frequent adverse events overall, and then looking at the subcategories by race.

And essentially, with the exception of chest pain, congestive heart failure and cough, which seem to be reported more frequently in African Americans and black patients in this population, there didn't seem to be any marked differences, though the number of black patients is small and the percentages have to be interpreted cautiously.

DR. ARMSTRONG: Thank you.

ACTING CHAIRMAN BORER: Any other questions? Ray.

DR. LIPICKY: It doesn't appear as though there are dose limiting side effects, that is, there aren't any side effects that come out as a function of dose. Is that a correct interpretation?

DR. GLAZER: From Protocol 106, with the exception of hyperkalemia and possibly if you pull out the MEDRA (phonetic) term postural dizziness, there didn't appear to be any dose related effects in that parallel designed trial, which included about a --

DR. LIPICKY: From any other trial that you have that shows dose limiting side effects?

DR. GLAZER: I believe in the hemodynamic trial, with the exception of the laboratory events as you would expect with potassium, there were no adverse events in that small, parallel design, hemodynamic trial that showed any --

DR. LIPICKY: Can you then tell me what the process, the thinking process, was for deciding that the top dose you studied was the best dose to study?

DR. GLAZER: The dose was chosen based on the hemodynamic trials, and we wanted to --

DR. LIPICKY: Yes. So just where does it fit on the dose response curve do you think? The low end, the high end or where since you don't have any reason to -- from adverse effects to not have increased the dose?

MR. MacNAB: Dose selection was based on several factors. One, from the hemodynamic trials that you saw where one clearly had much better dose response than the other, but in the both clearly the 160 b.i.d. gave the most effect.

DR. LIPICKY: But --

MR. MacNAB: There was --

DR. LIPICKY: -- it was also consistent with the dose response still going up.

MR. MacNAB: That's correct.

DR. LIPICKY: Right. So it didn't tell you you got to the top.

MR. MacNAB: I mean you get to a point of some practicability in that, too. I mean, the 160 --

DR. LIPICKY: No, no, no.

MR. MacNAB: -- tablets are pretty big, but I just want to finish.

All right. We did see a sense of the dose response in a little bit of potassium in 106. If you go back to some very, very early studies done of the angiotensin infusion studies in normal volunteers where the pharmacodynamics of hypertension appeared to be at 24 hours, you really need to go up to higher doses to get full suppression of the system, and it was generally believed, theoretically at least, to see optimal effects that you needed the most suppression of the system that you could get.

So I think based upon those angiotensin-1 infusion studies, while in normal volunteers and not perfect, it was one the hemodynamics; a little bit of potassium that we saw; the 160; and just the practical fact that, you know, a 160 milligram capsule of valsartan is not --

DR. LIPICKY: Big.

MR. MacNAB: -- is pretty big as it is; that putting that all together, that seemed the best to do. Now, obviously it would have been great if we had done a, you know, 50, 60,000 patient trial with multiple arms and multiple doses, but again, that's a practical matter.

So it was a sum total of all those reasons is why we picked that dose, and it could have been too high, could have been too low. That's the best we could do with what we had.

ACTING CHAIRMAN BORER: Alan.

DR. HIRSCH: Well, to come back to the same questions, I'm not sure I -- I'm not sure I'm on.

(Laughter.)

DR. HIRSCH: I could raise my voice.

Could we come back to safety Slide 14?

I'm not sure, as Ray said, that I'm seeing a dose dependent adverse effect profile that worries me, but nevertheless, let me just explore the data if that's why we're here.

I just want to see if you can help me understand the sort of negative dose response for dizziness and for hypotension. Does this mean that the physicians stopped the dose titration at 80 because they perceived or someone perceived dizziness and hypotension as being an end effect?

DR. GLAZER: This was a parallel design trial. They were on a fixed dose.

MR. MacNAB: One, oh, six was parallel. There was no -- it was a fixed dose.

DR. HIRSCH: One, oh, six. Sorry.

DR. GLAZER: That's why we couldn't use Val-HeFT, because of the forced titration. This is the only trial you can get pure dose response information.

MR. MacNAB: I mean, overall, if you looked at all of valsartan data and from hypertensions, heart failure, there is not a real dose response for side effects.

DR. HIRSCH: Right. I didn't see it either.

MR. MacNAB: I mean you have to really dig. Now, I imagine we could give a 1,000 milligram capsule, but for what we've seen up through 160 or through these trials, no.

ACTING CHAIRMAN BORER: Okay. If there are no other questions about safety at this point, although we can always revisit that a little bit later, we'll take our break now and come back and have the risk-benefit or the benefit-risk discussion after the lunch break and then go through any additional discussion and the questions we've been presented.

We'll take 45 minutes. So we should be back here ready to start at 12:30.

(Whereupon, at 11:40 a.m., the meeting was recessed for lunch, to reconvene at 12:30 p.m., the same day.)

 

 

 

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

(12:31 p.m.)

ACTING CHAIRMAN BORER: It is now 12:31 and a half. So we're a minute and a half behind, which we cannot be.

Jay, would you give the benefit-risk discussion, please? And then we'll have whatever additional questions we have, and we'll go into the structured review.

DR. COHN: All right. Thanks, Jeff.

I thought it would be appropriate to kind of place Val-HeFT in perspective with what we have already learned about heart failure and then how this should impact upon the management of the syndrome.

So if you'll bear with me for a few minutes, let me first review for you where we have come in the management of heart failure.

The first trial I was involved in, and of course, it was the first trial carried out in heart failure was V-HeFT 1, and at the time that we introduced that trial, the one-year mortality in the placebo group in V-HeFT 1 was about 20 percent, and we were really focused on reducing mortality as a primary endpoint in this disease. Twenty percent of people were dying each year with the disease.

By the time we got to V-HeFT 2, having been introduce to hydralazine and nitrate and did the solve trial, the placebo arm had now exhibited a lower mortality, right, a better cumulative survival. So we were clearly down moderately in mortality.

In Val-HeFT, the one year mortality was nine percent. So we've made actually remarkable impact on this syndrome. Even in the absence of the new therapies, this is the placebo arm in Val-HeFT, and we're getting closer and closer to the epidemiologic data on the predicted survival in nonselected 65 year old white males.

So the room for further benefit is still there, but the magnitude of efficacy that we can hope to achieve with further trials in patients with advanced heart disease is limited, and let's face it. All of these trials enter patients with left ventricular remodeling. They've had their disease for years. They're sick. We're not going to restore these people to a normal life expectancy, but we've already made a rather dramatic impact.

And I think when you look at the prevalence of this disease, five million people in the United States estimated to have the disease, 500,000 new patients per year, we're dealing with a very monstrous public health problem, and if, indeed, 91 percent of these people are going to survive each year, which is what Val-HeFT told us, then is it not time to stop focusing on the nine percent who die and begin focusing on the 91 percent who are alive who suffer from impaired quality of life, recurrent hospitalizations, frequent office visits, and incredible cost of health care?

Now, remember in Val-HeFT there were four days per year per patient hospitalized. If we have five million patients, that's 20 million days in the hospital per year in the United States alone. So producing a decrement in that hospitalization is of potentially important public health, as well as a personal benefit.

So I'm proposing, and this is not new. I've been preaching this way for a long time. So this is merely further evidence for my position, that mortality should not, cannot long serve as the marker for efficacy of treatment in a disease like heart failure.

Now, I would love to be able to monitor the biological process, and Ray has heard me preach this many times, and we may not be there yet, but we used all of the tools at our disposal in Val-HeFT to get a sense of the biological process, and they all turned out favorable.

So I think we're getting closer and closer to being able to understand this disease and deal with it before people die and intervene effectively to alter the quality of life by altering the biological process.

Now, let's look at where we've come now. This is fascinating data analysis from Val-HeFT which to me is the most impressive piece of data in the whole study, even though it's not necessarily randomized therapy. Let's look at the morbidity in Val-HeFT.

This is the morbidity in patients who were for reasons only their physicians know, being treated with neither an ACE inhibitor or a beta blocker. Almost 50 percent of them have had a morbid event during the course of the study.

This is the group of patients who are on an ACE inhibitor, but not on a beta blocker, and their morbidity was quite a bit lower. We've made a fairly important benefit by adding an ACE inhibitor.

And then, of course, we also have a group of patients who are on a beta blocker, but not on an ACE inhibitor, and they also exhibited a benefit, further evidence in a non-randomized way that giving an ACE inhibitor or a beta blocker is good treatment for heart failure.

Now, let's look at the group that were on both drugs. Look at this. Their morbidity has been reduced to about 20 percent. This is the best therapy for heart failure, and we should continue to aggressively attempt to get patients with heart failure on ACE inhibitors and beta blockers, and to me this is kind of the nicest data we've even got because this isn't part of a randomized trial. This is real world use of these drugs.

Now, let's see what Val-HeFT did to these four subgroups. In this group a striking reduction in morbidity, p, .0033. In this group who were on ACE inhibitors, but not on a beta blocker, a significant reduction, p of .0019. In this group that were on beta blockers, but not on an ACE inhibitor, a rather dramatic reduction, statistically significant even though the group is a small sample.

And this is our favorite little whipping boy here, the patients who are already on an ACE inhibitor and a beta blocker, and lo and behold, as we've already pointed out, it went in the other direction.

So these three groups exhibit a striking benefit on morbidity, mostly on hospitalizations, and of course, mostly having a major impact on the 20 million hospital days, I guess, that occur in the United States each year for worsening heart failure.

Well, what about getting everybody on ACE and beta blockers so we don't need to use valsartan? That would be wonderful.

The estimated drug usage in the United States today of these drugs is that ACE inhibitors are employed in somewhere between 50 and 75 percent of the population. Remember in Val-HeFT it was 93 percent, but that's not representative because those on an ARB were excluded, and that was also highly specialized heart failure centers.

So in the real world it's 50 to 75 percent.

Beta blocker use, the estimate is somewhere between ten and 25 percent. Well, what's the matter with the practicing community? Don't they know that the ACE and the beta blocker is effective treatment?

Well, I guess they do and they don't. Why are more patients not being treated with ACE inhibitors and beta blockers?

Well, number one, I guess, is physician education and the pharmaceutical industry is out there promoting these drugs. So it isn't a hidden message. Yet it hasn't had as big an impact as we might have thought.

Patient compliance. Patients are told to take drugs, and they don't have a good idea why because their physicians don't really know why, except that it was shown in a trial to reduce mortality. Patients want to feel better, and these drugs like beta blockers often don't. In fact, they have a period of time when they may feel worse.

There's wide perception about drug intolerance. Cough with ACE inhibitors, renal dysfunction with ACE inhibitors, hypotension, bradycardia, side effects that cause patients not to want to take pills; co-morbidity which excludes drugs, chronic obstructive lung disease,peripheral vascular disease, et cetera.

And with beta blockers particularly, the need for careful drug titration, that many primary care physicians do not want to undertake. So my view is that although we should continue to work on all of these issues and reassure physicians and patients of the importance, we are not likely to markedly improve the compliance with the currently recommended drugs which appear in every consensus document and recommendation. They are still not as widely used as they could be.

So is there room then for addition of a new drug like valsartan, which is perceived to be very well tolerated and is well tolerated, does not need careful -- as careful drug titration, and can reduce some of the burdens that the heart failure patient undergoes?

So here is my algorithm, if you will, for the current management of heart failure. Patients with heart failure are generally treated with diuretic, and they may be treated with digoxin. They also may be treated with nitrates and hydralazine. I'd like them to be, but Ray hasn't seen fit to give us permission to market it for that purpose. Is that?

(Laughter.)

DR. COHN: They may be given spironolactone, as has been suggested, although it's particularly advanced Class III and IV, but they will be on all kinds of background therapy. We think that everybody with heart failure, if tolerated, should be on an ACE inhibitor. If they tolerate the ACE inhibitor, they should be given a beta blocker. And if they tolerate the beta blocker, they should be monitored.

This is optimal therapy for heart failure. Are we going to get 100 percent of our population on these drugs? No way, no way. There's lots of reasons why, as I've already pointed out to you.

If they do not tolerate the beta blocker, I would suggest they be given valsartan based upon the Val-HeFT data, and then they should be monitored on ACE inhibitor and valsartan.

If they don't tolerate the ACE inhibitor, they should be given valsartan based on the very good data in Val-HeFT on valsartan in the absence of ACE inhibitor, and if they tolerate valsartan, they should be given a beta blocker and monitored. Now they're on valsartan and a beta blocker.

If they don't tolerate valsartan, what should you do? You should give them a beta blocker anyway because the beta blocker is an effective form of therapy, and this is not ideal therapy. Beta blocker alone is not as good as beta blocker with either an ACE inhibitor or valsartan, but it's better than neither.

So this is the algorithm then that I would propose based upon Val-HeFT, superimposed on all the data that we've accumulated over the last 15 years, and I believe we can make a major difference now in the outcome in patients with heart failure with this approach to therapy.

And I'll stop there. Thank you.

ACTING CHAIRMAN BORER: Thank you very much, Jay.

Tom you had two overarching questions about efficacy?

DR. FLEMING: I think they can be incorporated in my comments at the end in the general review.

ACTING CHAIRMAN BORER: Okay. Are there any other questions left for Jay or for anybody else who's presented or hasn't presented?

DR. NISSEN: Just one, Jay. You know, one of the problems is, as you've said, we don't know why those patients that didn't get an ACE or a beta blocker didn't, and so in your algorithm, what the algorithm doesn't reflect is the fact that we don't know that the patients that got valsartan in Val-HeFT got it because they couldn't tolerate ACE or couldn't tolerate beta blockers. For some reason or another they didn't get those drugs.

DR. COHN: I agree with you.

DR. NISSEN: And so it's a gap in our knowledge that doesn't reflect it in your algorithm.

DR. COHN: Well, that's true. I can't address the beta blocker story at all because only a third of people were on a beta blocker, and I think that reflects physician judgment to a large extent.

I think the ACE inhibitor story is probably a little clearer. This is a population of physicians who really do believe in ACE inhibitors, and since 93 percent of the patients were on it, I suspect, but I agree with you. I can't say that. We have no data to back that up, that the seven percent who didn't were at least perceived to be ACE intolerant; otherwise they would have been on it.

But I agree. We have a gap in our knowledge.

DR. NISSEN: But it obviously has major implications for how we think about this because because do we think about valsartan as a drug that can be used in ACE intolerant patients? You know, based upon those patients that got it without ACE inhibitors, without really know what their baseline characteristics were. It's a bottom --

DR. COHN: You're going to have data in a couple of years from the CHARM trial in which they specifically have a group that are ACE intolerant, and however, I just forewarn you. Given the number of people in the CHARM trial in the ACE intolerant group, it is clearly much higher than any of us ever observe ACE intolerance.

So I don't even think you're going to be much off with that database. These are not ACE treated patients, and they're not getting ACE inhibitor for various reasons, and intolerance is the easiest one to identify.

ACTING CHAIRMAN BORER: Paul.

DR. ARMSTRONG: Jay, two thirds of the Val-HeFT population was on digoxin. Was the effect on hospitalization equally powerful with digoxin versus non?

DR. COHN: Yeah, I think it was. That's a good question, and I don't know that we have digoxin/no digoxin data, but we've looked at that, and there didn't appear to be any impact of dioxin on the efficacy of valsartan. Good question though, Paul.

ACTING CHAIRMAN BORER: Ray.

DR. LIPICKY: As you carefully pointed out, I'm old fashioned, and --

DR. COHN: I didn't say that.

DR. LIPICKY: It used to be that one made people live longer or feel better, and now there's this morbidity thing that is an index, that is -- or morbidity is counted as hospitalizations, and I don't know where to put that in the spectrum.

For example, if one thinks about just feeling better, usually we're fairly hard on the strength of evidence that's required to accept the fact that people feel better , and morbidity is somewhere in between there.

How should this be viewed?

DR. COHN: Could I have 027? Because I want you to expand your two paradigm concept, Ray.

This is the goals for heart failure therapy: make people feel better. That's short term. Prevent people from feeling worse because that's quite different, and what we've shown in Val-HeFT is that we kept people from getting worse because we slowed the progression of the disease.

Making people feel better, you can make them feel better with a dobutamine infusion or a milrinone infusion short term because you correct hemodynamics. That's short term benefit.

This is the progression of the disease, the second one, that we all recognize is going on under our eyes and in every patient we see unless we're optimally treating them, and that's more of an intermediate term outcome.

And, yes, you'll begin to see a reduction of hospitalizations. We saw it within about three months, but the long-term effects really do relate to prevention of progression, and that's delaying death and reducing morbidity, which is hospitalization.

DR. LIPICKY: Right. It --

DR. COHN: So there's three phases.

DR. LIPICKY: If we could talk about that just for a minute longer, you could make people feel better for a long time. It's not necessarily short term, but --

DR. COHN: But you have to prevent the progression --

DR. LIPICKY: Right.

DR. COHN: -- as well as improve the short term. It's two phases.

DR. LIPICKY: Now, if you take the prevention of hospitalizations as evidence for altering the natural history of the disease, why don't you affect mortality?

DR. COHN: Well, that's a good question. Why don't you affect mortality? We are producing a modest slowing of progression of disease with valsartan.

I think valsartan has two components to its efficacy. Now I'm going way beyond the data, but I'm basing this on all my experience. It does have a favorable hemodynamic effect. The BNP comes down. The wedge pressure falls. There is a hemodynamic benefit.

That hemodynamic benefit may produce some modest improvement in the way people feel, not enough to improve exercise tolerance with the small sample that we've done. So it's a modest benefit, but it's also slowing progression of the disease even on top of existing therapy. The change in ejection fraction is one percent.

You know, with ACE inhibitors it was about three percent. With beta blockers, it was about seven or eight percent. So it's a modest effect, probably not great enough to further reduce mortality on top of existing therapy, ACE inhibitors and beta blockers, but enough to reduce hospitalization, make people feel better, keep them out of the hospital.

Is it a gangbuster's drug? Should it replace an ACE inhibitor or a beta blocker? I say no. I think ACE inhibitors and beta blockers do a little more of both. They are hemodynamic, and they improve the left ventricle.

Now, if we gave valsartan in the absence of an ACE inhibitor and we showed that, we get a benefit on mortality. So the reason we're not seeing a greater benefit may well be that they're all on ACE inhibitors, but we're getting further incremental benefit.

DR. LIPICKY: Right, but I guess I'd just like to press it one more minute. So if you're to take hospitalizations as an index of changing natural history of disease, what you're saying is that the change in natural history of the disease that you can demonstrate may not be enough to affect mortality, but that it should be accepted as changing the natural history of the disease and not as symptomatic care.

DR. COHN: Well, to say not affect mortality, I think, is perhaps an overstatement. As you know, when we eliminated that one subgroup who we really believed should not be treated with the drug now, there was a nine percent reduction in mortality. Now, the point estimate was nine percent reduction.

The confidence intervals don't tell us anything. Tom is shaking his head appropriately. I mean, that's -- but if we had powered the study like some of reperfusion studies, TPA, et cetera, and found a one percent reduction in mortality and went to the bank with it, you would say, "Gee, that's great," but we'd have to put 40,000 people in the trial.

So I'm not sure you wouldn't see -- we can't say that -- but I'm not sure we wouldn't have seen a modest, but small reduction of mortality had we powered this study like the TIMMI trials or the JC trials.

ACTING CHAIRMAN BORER: Any other questions or comments? Is there any other comment the company wants to make? Okay.

MR. MacNAB: Unless you have specific questions from the company's perspective, I think what you've heard this morning is essentially, you know, our perspective of the trial and some of the ways we think the drug should be used based upon the data of the trial.

But I'm here for questions if you have them.

ACTING CHAIRMAN BORER: Tom, did you?

DR. FLEMING: Yeah, I just had one procedural question unrelated to what we've been discussing while we have the company at the microphone.

I was delighted to read in the FDA briefing document that the FDA had requested the minutes of the Data Safety Monitoring Board's meetings, which I think should be an absolute procedure in all settings, in all trials that have data monitoring committees.

I was perplexed then to hear that the data monitoring committee didn't produce minutes, which I don't think has ever been a situation in which I was on a monitoring committee that we didn't produce minutes.

Can you clarify what happened in this setting?

MR. MacNAB: I'd ask Dr. Glazer, the data safety monitoring at endpoint committee. That was, again, an independent committee. So it was probably more independent than you discuss.

Why they didn't keep minutes?

DR. GLAZER: The DSMB chairman is not here. I think it was their decision not to do that. Again, it was an independent committee. I don't know the precise reason as to why they didn't keep written minutes.

DR. FLEMING: Well, I guess in general terms these are very important insights. I would urge that in the future all sponsors insure that their monitoring committees are generating minutes where the open session minutes presumably with recommendations would be disseminated to the sponsor, and the closed session minutes would be archived precisely to allow insights to regulatory authorities and review bodies such as this.

ACTING CHAIRMAN BORER: Okay. Well, we'll move on to the structure portion of the session.

MR. MacNAB: Could I make just one procedural comment? You have in your questions, I think it's Question 3, which is sort of a statistical question. Once you get into that, will we be able to comment on the comments or would you like us --

ACTING CHAIRMAN BORER: Not unless somebody on the committee wants to hear a comment.

MR. MacNAB: I mean, because perhaps Dr. Fisher would have some comments on that.

ACTING CHAIRMAN BORER: If you have something you want to tell us about now, Lloyd.

MR. MacNAB: Yeah, why don't you do that?

He can express these things much more elegantly than I will ever begin to.

DR. FISHER: Well, I wanted to speak about several things, why I feel the way I do, which you'll hear shortly, but I also wanted to discuss this relative to the two study paradigm. I thought that would come up, and the reason I was trying to flag Malcolm was because I think it will come up in the questions, and I think you will hear a discussion of it.

And this is an unusual situation. In most trials where you have a primary endpoint and a secondary endpoint, if you don't meet your primary endpoint, then in a certain sense you're not allowed to look at your secondary endpoints, although there have been some where they were incredibly extreme.

Carvedilol comes to mind, which generates a lot of discussion, but you think of it as an exception. Here we met the primary endpoint. There were two of them, but according to the rules adjusting for multiple comparisons, that endpoint was met, but it was met at a level that's somewhat problematic.

The two trial paradigm, I've heard some other arguments, but a p value somewhere between .00125 and .0056 is kind of the range you would like to get. But nevertheless we met the primary endpoint, which suggests you can look at the secondary endpoint.

Well, the secondary endpoint of hospitalization predefined blows away the p values usually required for the two study paradigm. So this bring up the issue of, well, can you only look at it down to the p value of the primary endpoint.

And as far as I know, there's no precedent for this. At least I've never heard it discussed in cardiorenal, but the point I wanted to make is in any event this is not a clear violation, but it's something that's a very subtle statistical point. It deserves consideration.

Another thing was Tom looked at all of the signs and symptoms data and said, well, gee, this isn't surprising. They're all correlated.

I think I spent more time than Tom thinking about heart failure, and I can assure you I have been in numerous meetings that said can't we find one sign or symptom to show a benefit for the patient. And anybody here that's been involved with heart failure, I'd be amazed if you don't think it's bizarre somehow that these things don't correlate like they should.

So I found this amazing. In fact, I think of those signs and symptoms there were five of them that also met the two study paradigm. They were not prespecified secondary endpoints, but again, there was less than .001, adding to the weight of evidence.

Some of you know, but I will say it in public. I will not come to a meeting with a sponsor unless I think their compound should be approved, and there have been meetings where I have been invited. I usually don't tell them, "Forget it. I don't think it will be approved." I say, "Well, gee, I'd kind of like to, but I'm busy," or you know, if worse comes to worse, I tell them the truth.

And in this particular case, when I was asked, they gave me the data. Nobody talked to me about anything, and I said, "I'm not sure I'll go with you. I want to see how you're going to handle the subgroups."

And the subgroup that bothered me, the thing that bothered me immediately was the high mortality in the people on both ACE inhibitors and beta blockers, and certainly we have a number of experts sitting on this panel who could run rings around me physiologically, but my gut reaction was: I think this is true. Somehow it's too much of a good thing.

That was just my gut reaction, and not only that. There are a lot of interactions we worry about, but here was one that was significant at the .009 level, and we hadn't looked at all this other stuff when that came up. So we started looking at things.

Jay called an ejection fraction independent, and so on. It's not truly independent. It's all part of the same picture, but there are very different parameters, and as you saw, almost everything you looked at went in the same direction.

So I would think it would be ridiculous for this committee if you voted for approval to not say to the agency, "Well, I can't stand here and claim absolutely equivocally it's proven it's harmful," and I doubt we'll ever know because I wouldn't picture somebody betting their money on that being a mistaken idea.

But when I believe that, there's a flip side to that. That means the other data look better. Now, we don't know how to interpret those p values, and I'm sure Tom is going to get up here, my guess is, and tell us, "Well, gee, in every study even if you have no effect, if you search long enough you can find a subgroup that looks good, and then there's a subgroup that looks bad by complementarity."

But this to me was a very different situation. To me, to the extent I understand the biology, it went along with things. There was a question about sample size. I was not involved, but I can almost bet how it went.

They went into cardiorenal. Cardiorenal said, gee, congestive heart failure. All drugs either help or kill, and we don't which, and we're not right enough to know which. Unless you get quite a bit of mortality data, we're going to put a black box in there saying this drug may -- Ray can correct me if this is wrong, but I've heard him say it in another context -- there might be 20 percent excess mortality, and so a relative risk of 1.2 or 20 percent often comes up.

The sponsor is hoping it works, but at least by powering things for that, you would be attacking in a responsible way the mortality issue, which I think the sponsor has done.

So when I put all of this together -- oh, the other thing is the Minnesota living with heart failure questionnaire. I told Jay to push that, but since he helped design it and so on, he thought that was inappropriate, but I really don't think it's the greatest questionnaire in the world for heart failure because it almost never is beneficial. No offense, Jay, but that's --

(Laughter.)

DR. FISHER: But here's a case where it actually did turn out to be statistically significantly positive, and so all of these things kind of added up in my mind.

I don't think you're greatly violating the two study paradigm. If you only focus on the one primary thing and say that's more than .00125, I guess that would be a defensible thing. If Lemm Moyer were here, I would expect to hear it pushed to the nth degree, but I think given that our tradition is if you meet your primary endpoint, you can begin to open things up, and then when you open it up, everything you look at goes in the right direction, and it goes very strongly. It's not just that they happen to hit two signs and symptoms at .02.

They had a whole bunch of them at less than .001, and almost all of them were statistically significant. So when you get in your discussion, we have a number of people here who have been involved with congestive heart failure trials. I'd be interested to hear in the discussion how you inject yourselves into that, whether you think my comments make any sense or if that's your experience.

For the audience, I'm a know nothing biostatistician. So it's a little bit frightening to get up here and start talking about the clinical things where I hadn't heard a number of points today, and I don't think there will be an opportunity for it to come out later, which is why I'm emoting so much.

Thank you.

ACTING CHAIRMAN BORER: Okay. Any other points before we move on?

It doesn't look like it.

We have the structured questions here. I think I'd like to make just one preliminary comment to put this in a framework that at least has meaning for me.

I think this was a very interesting development program centered around an extraordinary clinical trial, a landmark study really, and the fact that that study is so powerful, I don't think there's much question that it shows that there's a clinical benefit from the use of this drug in someone, is a tribute to -- the fact that it shows these things and that it was so well done is a tribute to the sponsor, Novartis, and to Jay in directing this and in presenting it as effectively as he did.

You know, I wouldn't have guessed that hitting the renin angiotensin system from another angle would have been an easy thing to do and that it would have been highly likely that one could find something even if there was something there to find.

And this study did find something, and I think, again, that that's a tribute to the precision and perseverance of the people who were involved in organizing and directing this.

So, you know, I don't think there's much question here that there is a clinical benefit. I think the question that we're going to have to focus on is exactly what is the benefit and who benefits, and those are laboring issues, and that's what I think we have to grapple with here today.

We've heard several comments about how we might grapple with those things, but I think that the way these questions are written, and they're superbly logical in their sequence of reasoning through an NDA to a final conclusion will lead us to have to conclude how we want to advise the FDA to write a label and if we, indeed, believe Lloyd's argument and everything else and conclude that the drug actually does something.

So with that having been said, we'll begin the comments, and our committee reviewer, Tom Fleming, will lead the discussion for most of these questions, I think. He hasn't given a formal review, but he'll give what would have been his formal review in the context of the questions.

I will not read the preamble here. I think that everybody has a copy, and you can read it and the sponsor knows what the study was, studies were. They're described here, and the committee knows.

So we'll start with the question. Consider the exercise tolerance Studies 110 and 106. In Study 110, all subjects were on ACE inhibitor for at least three months prior to enrollment. Therefore, the subjects randomized to valsartan were withdrawn from ACE inhibitor.

What is known about the time course for the loss of effects of an ACE inhibitor on exercise tolerance?

Would anybody disagree with me if I said nothing?

No. Okay. So the answer, Ray, is nothing, is no.

One, point, two. In Study 110, subjects walked 420 meters in six minutes at baseline. What degree of impairment does this represent?

Does anybody want to speak specifically about that? I think the answer will be not much, but does anybody want to elaborate on that?

No. Okay. That's not an overwhelming impairment it sounds as if.

What is known about effects of valsartan and enalapril on exercise can be summarized as follows. We have the table here, enalapril versus placebo. Treadmill is improved. Six minute walk data not available. Valsartan versus placebo, treadmill unaffected from the data we have. Six minute walk, it says no data. There are some, but really versus placebo there's not much.

Valsartan versus enalapril. So it doesn't look as if this -- from the way that the question is framed that there's a great deal that we can say about exercise tolerance from the available data, and therefore, what is the effect of valsartan on exercise tolerance?

Tom.

DR. FLEMING: Well, let me comment, and let me preface my comments by mentioning that as primary reviewer, I think these questions have really laid out to my way of thinking what the critical issues are that need to guide this committee discussion.

So in the efficiency of time, I would like to imbed or incorporate my primary review just into these as we go question by question, and before I do, let me just generally reaffirm some of what Jeff has said.

I also believe that this study is an extremely informative study. I'm very appreciative of the efforts of the team of investigators and the sponsor in conducting the study, and I would argue the challenge that's facing us is very much getting a sense of not as much about the reliability of results, but the interpretation of results and understanding the populations that may benefit.

Relative to this Question 1.3, what is the effect of valsartan on exercise tolerance, and the question is in the context of Studies 110 and 106, 106 involving 770 patients looked at exercise tolerance time. Page 78 in the briefing document, FDA, and there's no difference. One, oh, seven, also no difference. Six minute walk test on page 112.

Study 110 looked at an active comparison against enalapril with a six minute walk test on page 121 and essentially showed roughly comparable results, although the two placebo controlled comparisons say that there is no difference.

So essentially -- and I think the sponsor's presentation acknowledged this -- the data from these trials suggests that there was no detectable effect on exercise tolerance.

ACTING CHAIRMAN BORER: Okay. Does anybody have any additional comment to make about these?

Okay. If not, let's go on to Question 2. Ignoring Val-HeFT, what role did the two studies of hemodynamics and two studies of exercise ability and quality of life have in the case for approval? Do they contribute to demonstration of clinical benefit?

Does anybody want to take that? Tom.

DR. FLEMING: Yeah, I might comment. Essentially the two studies on hemodynamics, assuming it's 103 and 104, the two studies on exercise tolerance, 106 and 110; the two studies on hemodynamics certainly do provide evidence of effects, for example, on pulmonary capillary wedge pressure. It's my sense that such evidence is relevant to the extent of establishing a biological effect and potentially plausibility of clinical effects, but I think it's very noncontroversial to say that it can't be; it's not a validated surrogate.

The two studies on exercise tolerance and quality of life, and we've already alluded to this in the answer to Question 1, the results in 106 and 110 on exercise tolerance time were essentially negative, as they were in 110, and the Minnesota living with heart failure questionnaire results in 106 were also negative.

So kind of incorporating these issues into the answers to 2.1 and to 2.2, information on quality of life and biologic activity measures are certainly potentially helpful in getting a sense about mechanisms of action and biological plausibility of efficacy.

But in the context of having such an extremely informative study in 107, in a relative sense they don't provide very much more insight. Hence I would say probably are largely irrelevant in the context of what we have already learned from 107.

Would you be better off without them? No. The sponsor certainly might have had a stronger case in the sense that 106 and 110 are negative. One, oh, seven is also negative on six minute walk, but is positive on quality of life and signs and symptoms.

So, in essence, overall I would argue that measures of biologic activity are of relevance in establishing plausibility, but don't provide direct evidence about that, and the information on quality of life was actually in those two studies, was fairly negative.

ACTING CHAIRMAN BORER: Steve.

DR. NISSEN: A very brief comment. I agree with everything Tom said, except the hemodynamic data. I think it's important for us to all remember the paradox of inotropic agents, which make hemodynamics better and probably adversely affect survival, I think.

So I don't see a lot of signal. If one shows me an acute hemodynamic effect of a drug, it doesn't really tell me very much about whether it's going to be useful in chronic heart failure, and so they're noncontributory from my perspective.

I think they help the sponsor more to understand whether there is some additive hemodynamic effect over ACE inhibitors, but other than that, I think they're not really very helpful for this application.

ACTING CHAIRMAN BORER: Alan, did you have a comment? No.

Ray.

DR. LIPICKY: Well, in part it might be worth hearing a comment or two if you think it's worth commenting on. Those other studies did have some aspect of do patients feel better, and it wasn't striking that there was or that they found anybody feeling better, and that is sort of not the case in Val-HeFT.

It rather strikingly seemed to make people feel better even when they were on ACE inhibitors. So part of the question is does that detract from how you're going to interpret what you see in Val-HeFT.

DR. COHN: May I remind Ray?

ACTING CHAIRMAN BORER: Yes.

DR. COHN: What we found is there's benefit for the long term. These other studies were before any of the curves began to separate. So we kept people from feeling worse. We didn't make them feel better. That's why I made the --

DR. LIPICKY: Okay, okay.

ACTING CHAIRMAN BORER: Actually I was going to remind you of that, too.

DR. LIPICKY: You were?

(Laughter.)

ACTING CHAIRMAN BORER: You know, for me, and anybody on the committee, chime in and disagree, I don't find the lack of clinical efficacy in these trials to be particularly compelling against because there were small trials over a short period of time, and we have this overwhelming data set that is compelling in the other direction.

So it's not a big negative for me. Is it for anyone else? No.

Okay. Number 3, consider the components of the morbidity and mortality endpoint of Val-HeFT, and we have this nice table here. What role do each of the components have in the case for approval? And how do we reconcile the large effects on hospitalization with post hoc analyses that show little or no effect on certain of the subcomponents?

Tom, why don't you just go ahead?

DR. FLEMING: All right. This is certainly an important question. The first part, what role do the components play, my sense in interpreting the components is what is apparent as you look at these four components. The results are really driven by all cause mortality and CHF hospitalizations.

And the sponsor's presentation today certainly did focus appropriately on those two elements. They're complementary. All cause mortality, obviously a critically important endpoint to these patients, but certainly CHF hospitalization is a separate measure that is going beyond mortality looking at issues of quality of life, although as was discussed by the committee, one would have possibly hoped that through affecting CHF hospitalization that might have led you to anticipate a mortality benefit as well.

But I look at them as largely complementary and very important elements, and those two elements are the two for which we have very substantial information about benefit.

Ray, if you could put up the first of those two slides.

The 3.2 though gets into the question about reconciling evidence for a favorable result on CHF hospitalization, but with little or no effect on all cause hospitalization.

And a slide that I had generated, and actually I think these results are similar to what the sponsor had presented in their attempt to clarify this issue as well; what we're looking at here on the top of this slide are events, are basically numbers of people in the trial with events where, of course, we have about 2,500 per arm, and we know that the death rate is about 20 percent of the patients in each of these two arms had died, where there is a two percent increase in deaths.

But we see if we -- and it is problematic to censor the deaths, but that's the approach here in analyzing what I underlined in green, which was the result that was referred to as the favorable result in this Question 3.2, and it certainly is.

I think we would all accept the most important positive signal in this study is this 27 percent estimated reduction in CHF hospitalizations, and these together provide the second primary endpoint, the 13 percent reduction in morbid events.

Now, the FDA asks the sponsor to provide in addition to that the numbers of patients that had hospitalization for any cause or death, and when you look at this data, you see no reduction, and so the paradox or the uncertainty here is why is it that we have no reduction in the numbers of patients that have death or all cause hospitalization when we're seeing a reduction from about 18 and a half percent to 14 percent of patients that would have CHF hospitalization? Does that mean that, in essence, there is an increase in non-CHF hospitalization?

Well, I think to really more appropriately address that and to get better insight into that, we have to look in addition to this analysis, which is the numbers of people with an event; we need to look at the numbers of events overall.

And I think a very important -- and these data, by the way, are in the FDA briefing document on pages 96 to 99. What's relevant here is that whereas there are about 1,365, 1,398 patients with at least a hospitalization or death event, the actual number of hospitalizations is two to two and a half fold as high. This is the total number of hospitalizations that are recorded over the duration of follow-up, 2,056 versus 3,106, which separate into those hospitalizations that are CHF and those that are non-CHF.

And in essence, what you see here is not that there's an increase in non-CHF hospitalizations, but there is this overall decrease of about 260 CHF hospitalizations that shows up in all cause.

Ray, I would say, in essence, what's happening here is what's giving the impression that there isn't a reduction in all cause events, is that basically we've having multiple repeated hospitalizations of both types occurring to many of these patients over the two year period of follow-up.

So I found these analyses that looked at the total numbers of events to be very informative, and in fact, the all cause hospitalization days data, I think, is very consistent with all cause hospitalization.

You've got roughly an eight percent reduction in the number of all cause hospitalizations. You've got about a ten percent reduction in the number of days.

What I'd like to do is in a minute go to another slide that basically then puts into context what the survival data, all cause hospitalization and all cause hospitalization days would look like for a hypothetical 100 people.

Ray, could you put that other slide up to basically finish this response to this part of Question 3.2?

If we looked at a typical 100 people then based on data from the Val-HeFT trial, where that data provided approximately two years' follow-up, what we would have in 100 typical people with about two years' follow-up is about 20 deaths on placebo and about 20 deaths on valsartan. There would be no change in the numbers of deaths.

Essentially, you would reduce for these 100 people the total number of all cause hospitalizations from 124 to 114. That's entirely -- those ten are entirely CHF hospitalizations.

So the total number of non-CHF hospitalizations would be 77 in both groups, but would be reduced by ten the number of CHF hospitalizations.

And correspondingly, with this ten percent reduction we would reduce by ten percent the number of hospitalization days.

So, in essence, what we're looking at here in a typical 100 people that would receive valsartan versus 100 that would receive placebo, no difference in deaths; no difference in non-CHF hospitalizations; but ten less CHF hospitalizations per 100 people, translating into about a day less.

So the number of days alive and non-hospitalized is about 690 per arm, and the number of hospital days is reduced by about a day. So one way of looking at all of these data is that these results suggest no effect on mortality, an average of one-tenth of a CHF hospitalization per patient eliminated. That translates into one less hospitalization day.

And in essence, then that really addresses all three of these elements of 3.2.

ACTING CHAIRMAN BORER: Does anybody have any additional comment about this question? Steve.

DR. NISSEN: Yeah. Well, I also want to complement the sponsor on it. I think this is an incredibly important and informative study, and if I seemed nitpicky about it, it's because that's our job, and I'm going to nitpick a little bit more here before we finish and say that, you know, while these differences are impressive, I would have been happier if the adjudication had been done differently. I think it would have made my confidence of the data a lot higher in terms of looking at these differences because the differences, as pointed out by Tom's analysis, are relatively modest.

I can't accept that coming into the hospital at nine o'clock in the morning, getting furosemide intravenously all day long and going home that evening is not an event, but doing the same thing at nine o'clock at night and being overnight is.

Now, the excuse that the data on the time were not available is a flaw in the design. It's easy to put in the case report form. What time did the patient get admitted to the hospital and what time do they get discharged? You can collect that data.

And, you know, suggesting that the time of day at which you get admitted would make such a big difference on whether it was counted as an event or not is a weakness that could have been overcome by more meticulous, you know, design.

And, secondly, I think that to me independent adjudication means the sponsor is not present. In many, almost all of the trials I've been involved with that's how it's been done, and I think there's good reasons for that because if the committee is doing the selection of the cases along with an independent, you know, group or CRO, whatever, it protects the study.

I think that it didn't make a difference here. Please don't get me wrong, but I think it helps our confidence more if it looks like the adjudication for a trial where these events were the endpoint of the trial, where you're spending I don't know how many tens of millions of dollars on a trial. It's worth a little extra effort to adjudicate these endpoints in the cleanest fashion possible and trying to understand that right up front.

Having said all of that, you know, I do think that Tom's comments are really along my line. I agree with his comments.

ACTING CHAIRMAN BORER: Ray.

DR. LIPICKY: I guess I'm setting myself up to be hollered at, but the way you presented the results, Tom, I was more impressed before. A tenth of a hospitalization? Is that worth adding the seventh drug to a treatment regime, et cetera, et cetera?

You made it seem like the effect is very small.

ACTING CHAIRMAN BORER: That's in 100 patients.

DR. LIPICKY: A tenth of a hospitalization per patient.

DR. FLEMING: Yeah, basically if we put it in the context of per 100 patients followed over two years, then in those 100 patients treated, you would prevent ten hospitalizations. That is a tenth of a hospitalization per patient, but that's basically ten hospitalizations prevented for 100 patients treated and followed for two years.

DR. LIPICKY: So you do think it's small?

DR. FLEMING: Well, that's what we --

DR. LIPICKY: You're not going to holler at me?

DR. FLEMING: I mean, my assumption is we'll be coming back to that specific discussion later on.

DR. LIPICKY: Okay, fine.

DR. FLEMING: One of the issues that you were raising here was specifically how is it that we could see reductions in the number of patients experiencing CHF hospitalizations and not see any reduction in the total number of all cause hospitalizations.

And the answer for that really reflects the fact that these patients are being followed over a time frame where there are really repeated events occurring, and as those repeated events occur, the numbers of people then that have at least one event becomes less informative about what the overall effect is.

But you think step back and look at the analyses that I was showing at the bottom of that first transparency. Then you do see that, in fact, even though the numbers of patients that have at least one hospitalization are the same, there really is not a negative effect in these data on non-CHF hospitalizations.

But the overall level of effect is, therefore, none. In the typical hundred patients, I don't have that slide in front of me now, but I think it was 77 versus 77 non-CHF hospitalizations and CHF hospitalizations were what was it? Thirty-seven against 47?

DR. LIPICKY: That's okay. So you will make a value judgment later. So we --

DR. FLEMING: I definitely think we should.

DR. NISSEN: Just let me holler at you a little bit and say that I would agree with you if it weren't for the fact that the study design was adding this agent to what was already very good state-of-the-art therapy, and so this degree of reduction on top of people who were very well treated is much more impressive than it would have been in another setting.

DR. LIPICKY: Agreed, but in a public health sense it's not too important.

DR. NISSEN: Again, I wouldn't necessarily agree with you about that. I mean, if you have to treat ten patients to prevent one hospitalization, hospitalizations use a lot of resources. They cost a lot, and they're not very pleasant for patients that undergo them.

I don't have a problem with giving a drug to ten people to prevent one of them from being hospitalized. That to me is a fairly impressive effect in this setting.

DR. LINDENFELD: Don't you think, Steve, though that we should consider it 20? Because we usually think about number needed to treat per year. So this is ten for two years. So that would mean you'd have to treat 20 to save one hospitalization.

DR. NISSEN: JoAnn, and this is in the same range --

DR. LINDENFELD: So that makes it a more modest --

ACTING CHAIRMAN BORER: Can I just add? Let's hold this discussion, if I may, because --

DR. LIPICKY: I withdraw the question.

ACTING CHAIRMAN BORER: -- you know, the question we're actually going to have to answer at the end of the day will incorporate elements of all of this, and we can get to it.

And with that in mind, can I ask, Ray, that we be allowed to skip number four? Because unless, Tom, you have something very much to say about it, because it's sort of a speculative value judgment that we're going to get to.

DR. LIPICKY: No, no, no.

ACTING CHAIRMAN BORER: I mean, we're going to get to it.

DR. LIPICKY: No. It gets at the very thing that Lloyd got up to talk about.

ACTING CHAIRMAN BORER: Well --

DR. LIPICKY: What is the level of significance you want from a single trial before you start paying attention to it if you only deal with the primary endpoint?

DR. FLEMING: I think I can comment.

DR. LIPICKY: Or endpoints.

ACTING CHAIRMAN BORER: Okay.

DR. LIPICKY: And then are you going to modify that based on what you see later?

ACTING CHAIRMAN BORER: I withdraw my request, Ray, and I'm going to ask Tom to give the definitive answer.

DR. FLEMING: Well, I think I can give at least a brief comment in response to this, and I think it does -- it really is. The way this is phrased is very specifically asking us to just look at the co-primary endpoints, not look at secondary. The questions will lead us then through five and six to bring in secondary endpoints and how does that influence our perspective?

As I see it, there are in Question 4, there are two aspects to address in this question. Basically what it's asking us is if we focus on the primary endpoints, how strong do these results have to be for us to view that these are significant or established benefits of clinical importance, and I think there are two elements.

There's statistical significance below the line, and there's clinical significance above the line, and this trial is a very important, very good trial. It's 5,000 patients. It's a trial that was adequately powered for mortality, which was a negative outcome, and by all likelihood, I was probing this and didn't really get a clear answer, but in all likelihood, it's more than adequately powered to address the other primary endpoint of morbidity, given that there are a lot more morbidity events than mortality events.

That being the case, it's a study with considerable potential to achieve a very strong level of significance if there's a strong signal of effect.

Traditionally, it's only a guideline, but statistical procedures are useful as a guide. We have considered results at the 025 level, two sided, as the strength of evidence that we would need to see to call a study positive. In this trial with co-primary endpoints and adjusting for multiple testing, that two-sided p value guide is not 05, but 02.

If we looked at this as a trial that provided, in essence, comparable strength of evidence to two independent studies, each of which would be positive, that's where the 00125 has arisen, and technically here I'd remind you it's not .00125 in this trial,b ut that divided by two or even more than two, because there was interim monitoring and there were two primary endpoints. So --

DR. FISHER: Could I say --

DR. FLEMING: -- essentially -- essentially what we are dealing with here is a trial from a strength of evidence if we focus only on, as we're asked to in this question, the primary endpoints that would meet the standard for what would be called a positive study on the morbidity endpoint.

But it doesn't really come close to what we would consider as essentially comparable to the evidence of two independent positive studies each of which just exactly hit strength of evidence that we would need individually.

The issue above the line of clinical relevance is certainly also very important. What we're estimating is a 13 percent reduction in the morbidity events, and I think that is an issue that we need to revisit, as we've said, toward the end in talking about what level of effect is, in fact, clinically important.

ACTING CHAIRMAN BORER: Tom, Lloyd raised a question. In general we wouldn't have more comments here, but specifically about your calculations.

DR. FISHER: I wanted to give a point of information. This is just a numerical fact according to -- the number .1 is wrong. There's actually a gain of one day per patient per year.

So in other words, if you're treating somebody for a year on the average, that person is prevented for a day of hospitalization.

Of course, as it turns out, it's very skewed. Some people are not hospitalized at all, and some more. So but on the average, each person each year gains a day.

DR. LIPICKY: That makes it a big number, does it? No, you don't have to answer that.

DR. FISHER: Well, no.

(Laughter.)

DR. FISHER: I've been in the hospital, and all I can tell you it wasn't a particularly wonderful experience given the -- for me it would be worth avoiding it, but --

DR. LIPICKY: Yeah, I agree. I don't disagree.

ACTING CHAIRMAN BORER: Okay. Ray, can we postpone further elaboration on four until we move a little further?

DR. LIPICKY: Yes.

ACTING CHAIRMAN BORER: Thank you.

Okay.

DR. LIPICKY: You're welcome.

ACTING CHAIRMAN BORER: Let's move on to number five. Consider other endpoints that were not individual components of the morbidity/mortality primary endpoint of Val-HeFT, and we have a table here with a number of them with some p values.

What role do each of these secondary endpoints have in the case for approval?

Tom, why don't you go through that?

DR. FLEMING: Okay. In any trial, it certainly is very important to focus, first and foremost, on the primary endpoints, in this case two primary endpoints, mortality and the morbidity hospitalization/mortality endpoints.

But in any trial it's also clearly important to focus beyond or to take into account, as well, the supported information. It becomes more subjective in interpreting those data. I tend to put less emphasis on p values and more emphasis on what is the magnitude of effect and how reliably is it estimated.

It is certainly relevant though to be considering how the safety profile, as well as secondary quality of life measures influence one's overall perspective of benefit to risk. What we have here, as has been summarized in the question, there are an array of secondary measures.

In particular, I might focus on the overall signs and symptoms results, as well as the Minnesota living with heart failure questionnaire results, and these data are presented in the sponsor's presentation on pages 51 and 52.

There is important positive reinforcement with these results. It is important though or relevant, at least as I look at these results, to note in the signs and symptoms that what we're seeing is on the order of a one to three percent increase in the number of patients that have improvement in fatigue, dyspnea, rales, et cetera, and one to three percent decrease in the numbers who have worsening, which in these sample sizes are showing up as statistically significant.

In the Minnesota living with heart failure where the overall scale at baseline was about 32, we're seeing a relative benefit of about one and a half points, preventing a one and a half point decline, in essence, and in these sample sizes, these results are showing up as significant at the 01 to 001 level in many cases.

The key question that I would throw out though is more than from a statistical perspective, from a clinical perspective. How important is this? It is a benefit. It is of some importance, but how important, for example is a change of one and a half points in the Minnesota living with heart failure questionnaire?

ACTING CHAIRMAN BORER: Steve?

DR. NISSEN: Yeah, this actually has a fair amount of weight with me, and again, when you only have one really good trial, big trial to look at is what we really have here, and I would point out something to you, Tom. You know, for those of us who see patients with heart failure, it's not a pleasant disease. Waking up in the middle of the night with paroxysmal nocturnal dyspnea is not any fun, and the fact that, you know, highly significant differences, particularly the symptoms here of heart failure, because ultimately making these folks feel better is important.

I know, Ray, you've emphasized in a lot of development programs, you know, the feel good aspects of this, and this is very supportive, and I would have been much less enthusiastic here if these things had all been neutral.

But you know, dyspnea on effort, you know, walking out to get your mail and doing the things you do, if you have less shortness of breath doing that, that's a good thing.

And so to me it's very supportive and it's very important, and I'm glad it was here and I'm glad it was done in this trial.

DR. FLEMING: Steve, let me be specific though. The issue that I'm trying to raise is not whether these symptoms are important. The question is: how much of an influence on these symptoms does one have to generate in order to say this is an important effect?

So if we look at paroxysmal nocturnal dyspnea, whereas in the intervention arm 6.6 percent improve, in the placebo 5.7 percent improve. Are you impressed that there is that extra .9 percent improvement?

DR. NISSEN: I am impressed.

DR. LIPICKY: Well, let me stake out a position so that you can argue with it.

DR. NISSEN: Okay.

DR. LIPICKY: Okay? With respect to the primary endpoint, this is one trial that has some other data that doesn't really tell you much. I would say the primary endpoint had to have a p value of .00125, and if it didn't, I wouldn't want to look much farther.

But in fact, you say, "Well, I'm going to look a little farther." Okay, and so you look at the secondary endpoints, and as Tom said, you don't know how to evaluate the strength of evidence there, but you have sort of magnitude effects to look at.

This is a five percent change, ten percent change, 20 percent change. So does that mean then that you're going to say based on the secondary endpoints I modify my stand? I'm satisfied with the p of .01 for the primaries because this really means something to me. Then we'd have to start talking about, well, when do you modify the way in which you make your first decision on the basis of looking at other data when you don't have much ability to be able to decide whether the other data is real or not in any usual sense.

That's what's going on here, and I contracted it and took extreme positions so that you'd recognize that.

ACTING CHAIRMAN BORER: Paul, do you have any thoughts about this that you want to share, the importance of these secondary endpoints?

DR. ARMSTRONG: I am impressed by their consistency, and so I think in that -- but also the fact that I don't put as much weight on them individually as a package in terms of clinical signs and symptoms and their concordance with the other measures. So to the extent that they support that, I think they're useful.

DR. FLEMING: Paul, this is a really tough question to ask because there's obviously a lot of subjectivity to this, but is there any literature? We talk about the Minnesota living with heart failure questionnaire as being validated. Is there a literature that says essentially something about magnitude of change on that scale for it really to be clinically relevant?

I guess what I'm really getting at is a one and a half point change, something that clinically is profound, moderate, but important, really modest, or close to irrelevant.

DR. ARMSTRONG: Well, as you would know, a one and a half point change on a denominator of six in a common problem might be quite relevant. And rales and third heart sounds track morbidity and mortality pretty well in heart failure.

DR. FLEMING: Well, I'm looking specifically at the Minnesota where the baseline average was 32, and based on having that baseline average of 32, which we were informed basically was moderate in the range, what do we know about a one and a half point change and how important that is clinically?

DR. COHN: Well, I guess I should probably address that because this is our form.

DR. FLEMING: Although to be candid here, this is -- I'd really like to get an opinion independent of the sponsor here, Jay, if you would. I would prefer not to get -- I think --

DR. COHN: I mean, I can give you the background on the form.

DR. FLEMING: But at this point --

DR. COHN: Okay.

DR. FLEMING: -- I would prefer to get my colleagues' opinions on the committee.

DR. HIRSCH: Maybe we should all speak up. I don't think we know the answer to that question. I think it's just another signal like the change in signs and symptoms that can't be quantified or qualified even though there isn't development information.

Going down the panel, as we always do?

DR. LINDENFELD: I agree it's a signal, but I think it's very much like the signs and symptoms. It's a modest --

DR. HIRSCH: Okay.

ACTING CHAIRMAN BORER: I must say for myself I'm very impressed with the consistency of the data over so many areas that might be evaluated, and personally I think that tools for measuring symptomatic benefit are inherently -- inherently provide information that's extraordinarily imprecise, and the result is that you should see anything at all is impressive to me, whether it's one point, two points, or three points on a scale that's fundamentally subjectively defined.

We see many trials with many tools, not only the Minnesota scale, but New York Heart Association functional class for a lot of cardiovascular problems. You know, it's often very difficult to see anything despite overall benefit as measured by certain objective endpoints.

So I'm impressed that there's anything positive here with the Minnesota scale.

DR. FLEMING: But, Jeff, you're saying what impresses you is that in many other trials of other agents there are no effects. So the fact that there is anything here is impressive, and yet I'd say, "Well, look at other endpoints. Look at mortality."

ACTING CHAIRMAN BORER: Right.

DR. FLEMING: We have other agents that show big effect on mortality. There's no effect here.

ACTING CHAIRMAN BORER: That's true, but what I see -- I don't want to jump ahead here, but what I see here is a drug that in this population overall didn't have much of an effect on duration of life, but it did seem to have an effect on how patients felt in terms of the various measures that were made.

If you look at the co-primary endpoint, which is one of the feel good endpoints, and you look at collateral data that might support that that endpoint should be accepted as meaningful, when I see the Minnesota quality of life scale coming out positive, that makes me more confident that the co-primary endpoint is a meaningful one, that the result on the co-primary endpoint is a meaningful one.

If it were the only positive and all of these others sort of fell out 50-50, I'd be much less impressed, but that as a component of 12 or 15 other parameters that were looked at here, all coming out the same way, impresses me a great deal.

DR. LIPICKY: Well, but --

ACTING CHAIRMAN BORER: Knowing as I do how difficult it is to get a history of symptoms that is positive when you apply a drug. So, you know, I think the package has to be taken together. This is one component of the package, and it wouldn't matter to me if it was one point or five points. If it were the only component of the package that went the right way, I'd be concerned.

But if it goes the right way and it's one point and the other data are consistent with it, I find it impressive.

DR. LIPICKY: Jeff, are you talking about the point estimates going in the right way or how do you know that that's real?

ACTING CHAIRMAN BORER: Well, I don't know that it's real, but I know --

DR. LIPICKY: Are you looking at the p values here?

ACTING CHAIRMAN BORER: I know that it's consistent --

DR. LIPICKY: How can you interpret those p values? Tell me how you interpret. What does a p of .004 mean?

ACTING CHAIRMAN BORER: I interpret it as indicating to me that the result was consistent across the population. That's the only interpretation I make of it.

DR. LIPICKY: But on the basis of the point estimate or the p value?

ACTING CHAIRMAN BORER: No, on the base of the p value I say that the result for that point estimate, the result that that is the point estimate of is consistent through the population.

DR. LIPICKY: I see.

ACTING CHAIRMAN BORER: And now we have consistent through the population for many, many point estimates. The magnitude of the point estimate really doesn't mean all that much to me.

DR. FLEMING: But here, Jeff, the challenge is really getting a sense of whether the changes that we see in dyspnea at rest and dyspnea on effort and fatigues, all of these measures showing two or three percent more people showing improvement. Are these -- I mean, we keep making it a point. It's really impressive that there are all of these endpoints. Is it maybe very much the same two or three people that are showing this level of improvement on measures that are highly correlated with each other?

ACTING CHAIRMAN BORER: Yeah, it may well be. It may well be.

DR. FLEMING: So it doesn't take away from the argument. It's favorable that we showed an effect rather than not showing an effect. I'm not debating that, although I'm worried about how much of an effect it is, but I guess I'm concerned about the conclusion that there's a page here of results, and most of them are significant, and this is overwhelmingly strengthening the conclusions because they're all significant when many of them must be heavily correlated.

ACTING CHAIRMAN BORER: Yeah, I think they are, but even with relatively heavily correlated parameters, when you look at a heart failure database, they generally don't all go the same way, you know.

DR. ARMSTRONG: Tom, maybe it would be helpful to say that the New York heart class, which is common to many other studies, does show a 2.4 percent improvement, and I think we could collect all of the symptoms in the New York heart class and say there's not a bad bed rock based on the other studies that looks like it moves, and then we could look at the signs, edema, the third sound, and others, and say there's probably two clusters in this data. One is symptoms and the other is signs, and they go in the right direction.

And for me as a clinician, that seems to make sense.

DR. LIPICKY: Well, a third heart sound doesn't do that.

ACTING CHAIRMAN BORER: Yeah, the other point --

DR. LIPICKY: Orthotomy (phonetic), it doesn't do that.

DR. ARMSTRONG: That's the toughest one to get.

ACTING CHAIRMAN BORER: Before you speak, Steve, the other point here is a point that you made, that you, Steve, made just a few minutes ago. You know, when I try to interpret these data, one of the confounders in the interpretation is the fact that there were so many other drugs on board at the outset of therapy. So that my expectations may be less than they would be if those confounding drugs weren't present.

Now, it's difficult to make the interpretation of what the confounding drug -- of how the confounding drugs affect the data. We don't know. The studies weren't designed to tell us.

If you go back then and look at the sub-analyses that were done, which I don't want to make too much of yet because we're going to get back to them later in a way that perhaps is a little different from the way that the sponsor presented them, but if you do look at them, if you look at the sub-analyses and you look at the people who weren't on ACE inhibitors and weren't on beta blockers, why, my goodness, the results come out the way I would have expected them to if this is a drug that works and does good things.

So, you know, when you take the totality of the little pieces, they seem all to point in a reasonable direction if one were to conclude that this drug has a clinical benefit in terms of relief of symptoms and prevention of hospitalizations.

Steve?

DR. NISSEN: Yeah, two quick points that may make Tom a little more comfortable. To really look at this fairly, you have to kind of look at this to some extent as a risk ration. Not everybody has PND, and so you know, if you reduce from seven percent incidence of PND to five percent incidence, then as a risk ratio, you know, that's a pretty big decline.

Looking at it in terms of absolute numbers for any individual marker is really not fair to that marker.

And the second point I wanted to make is that what to me really struck my eye when I reviewed this is that these are not all markers that are markers of symptoms. Some are markers of symptoms. Some, like left ventricular diastolic diameter, are markers of anatomy suggesting that there's effect on the anatomic progression of ventricular dilation or ejection fraction is another measure like that.

And some were measures of physiology, like jugular venous distension and rales. And so to me it is comforting when I see symptomatic changes, physiological changes and anatomic changes all going in the same direction, and so it's just reinforcing.

And so, Ray, it does have an impact. Now, how big an impact should --

DR. LIPICKY: Well, that's what we wanted to find out, and since we failed in answering Question 4 --

ACTING CHAIRMAN BORER: Yeah.

DR. LIPICKY: -- because you were supposed to name a number, will you go back to four and name a number? That is, you would expect that you would act favorably just based on the primary endpoint on --

ACTING CHAIRMAN BORER: And you want a p value or a --

DR. LIPICKY: -- something that was in the first --

ACTING CHAIRMAN BORER: -- hazard ratio or both?

DR. LIPICKY: -- block, second block, third block or fourth block of that line. It's a visual analogue scale. You can draw the line anywhere you want to.

DR. HIRSCH: The fact that we're spending so much time trying to avoid the question, I think, gives you part of the answer. Bill you want to try?

DR. LIPICKY: Well, just, you know, there's one, two, three, four -- four possible places you can put the line. Pick one.

ACTING CHAIRMAN BORER: Alan, would you or, Tom, go ahead.

DR. LIPICKY: Let's number them from left to right, one, two, three, four.

DR. FISHER: Can I make one quick technical comment?

Jeff is right. Hazard ratio is a biological parameter of the population. P value depends upon the study design and sample size, and so you might want to separate your lines. I mean this is the values that may go together with this study, but there's no necessary linkage between the two.

DR. LIPICKY: Yeah, that's true, but it makes it -- we can't answer even this question, let alone making it a little more complicated.

DR. HIRSCH: And these values pertain to populations, and we're talking about individuals here at risk with quality of life questionnaires. So it really is a mental segue I'm not really comfortable going down.

DR. LIPICKY: So all we want to know, and it's just name and number, one, two, three, four starting from the left --

DR. FLEMING: Well, there are two issues here, and in fact, --

DR. LIPICKY: Well, lower --

DR. FLEMING: -- Ray, I tried to -- I tried to answer this --

DR. LIPICKY: Yeah, you're right.

DR. FLEMING: -- the first time through when we got to Question 4, and I stated there are two issues. There's the below the line issue, which is strength of evidence, and there is the above the line issue, which is magnitude of effect.

DR. LIPICKY: Fine, fine.

DR. FLEMING: An they are both --

DR. LIPICKY: Fine.

DR. FLEMING: -- important issues.

DR. LIPICKY: So then if --

DR. FLEMING: The below the line issue, if we would follow what would be a traditional guide, and the critical word there is "guide" --

DR. LIPICKY: Right.

DR. FLEMING: -- any interpretation of a study requires careful consideration of totality of results, but should be heavily influenced by strength of evidence on the primary endpoint.

A tradition here would be to call this a single positive study or the evidence consistent with that, we would typically be using two sided .05, but remembering here that it's really two sided about .02, reflecting multiple testing and multiple primary endpoints.

Correspondingly, if you said, "Gee, what would it take in statistical strength of evidence to equate with two trials, each of which would hit the standard that I just mentioned?" it wouldn't in this case be .00125. It would be about less than half that because .00125 doesn't adjust for the two primary endpoints.

DR. LIPICKY: So it would be --

DR. FLEMING: It would be about .0005, and so --

DR. LIPICKY: So it would be line segment one.

DR. FLEMING: so the simple answer to your question purely from a statistical perspective is that this study does hit the strength of evidence just by a bit, not by a lot, standard for being called a positive trial, but really doesn't come close to what we would have typically stated was equivalent to the strength of evidence that you would have from two such positive trials.

DR. LIPICKY: Right. Okay.

DR. FLEMING: Now, the clinical issues are --

DR. LIPICKY: So that's a very clear answer, right? So --

DR. FLEMING: So now you want a number.

DR. LIPICKY: Well, if anyone disagrees with that answer, that would be important to say, and if no one wants to disagree, that's fine, and then we can go on because that same question comes up. Now you've got secondary endpoints. What are you going to do?

DR. FLEMING: Okay.

DR. LIPICKY: Okay?

DR. FLEMING: Okay.

DR. LIPICKY: But that will give us a feeling for what you think. Otherwise you're leaving us up in the air.

ACTING CHAIRMAN BORER: Can I just add one comment to Tom's comment? You've asked about the hazard ratio, as well.

DR. LIPICKY: Yes.

ACTING CHAIRMAN BORER: Which is the above-the-line part that Tom spoke about before, and I would say that it would be very difficult for me to provide a hazard ratio which is appropriate except in the context of this trial not to be extrapolated elsewhere because of the issues that we've already discussed --

DR. LIPICKY: Fine.

ACTING CHAIRMAN BORER: -- confounders that we can't interpret.

If you want a number, I'll tell you a number.

DR. LIPICKY: No, that's fine. I think you can ignore, I'm sorry to say, what Lloyd said because these hazard ratios were calculated on the basis of this data. So you could look at this sort of as the same scale, and what that was really asking for was do you have to have a 20 percent effect, a ten percent effect, and it gave a feeling for or gave you the opportunity to say, "Well, geez, if I saw a ten percent effect and a p of .05, that would be good enough for me," or that you'd want to vary that around.

So you can, indeed -- it's too complicated.

ACTING CHAIRMAN BORER: Okay. Then since it's too complicated --

DR. LIPICKY: Yes.

ACTING CHAIRMAN BORER: -- we'll avoid it. Let's go on to --

DR. LIPICKY: Let's leave the hazard ratios alone.

ACTING CHAIRMAN BORER: Okay. We'll go on to number six, which is the same question, but about the secondary endpoints.

DR. LIPICKY: But now you have more information. So are you going to change your mind with respect to how much evidence you need for the primary strength of evidence? Because that's the only guy you can really calculate a p value for.

ACTING CHAIRMAN BORER: Okay. Tom, do you have any additional comments about that?

DR. FLEMING: Well, this one is one that I think is intrinsically difficult to judge. If you have a level of evidence in your primary endpoint, how much influence do you allow secondary endpoints to play or to have?

It matters a lot on the relative importance. Lloyd was giving an example earlier of a secondary endpoint that was mortality and had very strong mortality trends or mortality influence.

Certainly mortality is a secondary endpoint; if it showed a very striking result, would appropriately influence your assessment of strength of evidence overall, and it, of course, could go in either direction.

A study that looked good on the primary endpoint with every unfavorable secondary measures could readily move you away from a judgment of positivity, and the reverse could happen where you didn't hit the primary and you had reinforcement on a secondary measure that was profoundly established and highly clinically relevant, and that should, in fact, substantially influence your judgment in a positive direction.

Here I guess what I would throw out is that certainly whereas I had argued that from a traditional perspective we would call it a positive study if we had a significance level less than .02 and it was .01, but not judged as the same strength of evidence as two such studies because that should have been more on the order of .0005 in this setting, and it wasn't close to that.

The question is: should the results be influenced or should your interpretation be influenced by secondary measures?

I think we're hearing pretty consistently that all of us would say yes. Where it might be more difficult though is to say how much should it influence, and where I've been probing with my colleagues is to try to get a clear sense from them as to whether these secondary measures on symptomatic effects represent profoundly important changes, moderately important changes, modest changes, or relatively trivial changes because that would influence.

I wouldn't move greatly on this pathway. I would still be heavily, most heavily influenced, Ray, by the primary endpoint

DR. LIPICKY: Right.

DR. FLEMING: But I would be somewhat influenced, and how much would depend on these clinical judgments.

DR. LIPICKY: Well, in part, to take it out of p values -- let's say that you're selling a watch, and you're asking $1 million for the watch, and someone comes up and says, "Well, I've got 100,000. Isn't that good enough?"

Roughly how the primary endpoints line up here, and you look at them and say, "Well, you're very nice looking and you talk to me nice and you're sweet and so on and so forth, and you have all these other things that I like about you. Yeah, that's enough."

And that's what we're trying to get a feeling for. How are you putting this together?

ACTING CHAIRMAN BORER: I don't think that analogy is apt here, Ray.

DR. LIPICKY: You don't think that's --

(Laughter.)

ACTING CHAIRMAN BORER: Because -- because I think what the $100,000 person may be offering is a little bit more, and I would infer those -- or in some cases a little bit less -- and I would infer that from the sub-analyses that were done, which we haven't gotten up to here yet.

So, you know, I think that once again you have to accept this as package. It's a very complicated package. We've not been asked to review an application like this before, to the best of my knowledge.

DR. LIPICKY: I understand.

ACTING CHAIRMAN BORER: So we're sort of breaking new ground. I think you have to see it as a package.

Can I ask? We haven't heard from --

DR. LIPICKY: We want -- we want you to clearly outline how you're thinking.

ACTING CHAIRMAN BORER: And we will.

DR. LIPICKY: Okay.

ACTING CHAIRMAN BORER: We haven't heard from Mike Artman down there. Mike, did you have any thinking about this?

DR. ARTMAN: Just going on with this package issue, I was very impressed, and Steve alluded to the anatomical and function and physiological changes, but there were also, I think, some very important biochemical changes that we really didn't focus on that provide me with a great deal of additional comfort.

The fact that norepinephrine and BNP levels went down in the treatment group and went the other way in the placebo group I think is a very positive finding that if I were the sponsor I would have stressed a lot more.

So, again, taking all of this together, drawing on Ray's analogy, yeah, it's 100,000 for the watch, but I'll give you another 250,000 for the band, and you give me a case. There's another 300,000, and now we're getting pretty close to the million that you were asking.

ACTING CHAIRMAN BORER: Well, at least we'll give it to you in kind.

Dr. Anderson, do you have any thinking about Ray's analogy or anything else with regard to this particular issue?

DR. ANDERSON: Mr. Chair, I read these documents, and I had a great deal of difficulty trying to sort things out, and particularly in terms of magnitude, you know, how big should it be and all of that.

In reading the FDA reviews, I notice that there were several statements which said we could not draw a conclusion or something to that effect, and I had hoped that in hearing my distinguished colleagues during the discussion I would be able to sort these out and decide which direction to move in.

But it seems to me that we're all somewhat perplexed by whatever it is that I can't figure out. So right now I don't really have anything to add to the discussion.

ACTING CHAIRMAN BORER: Okay. JoAnn, do you want to weigh in on this before Steve does?

DR. LINDENFELD: You know, I don't think -- first of all, I'm not buying any watches from Ray.

(Laughter.)

DR. LINDENFELD: I think that the strength, as Michael said, too, I think that all of the things are going in the right directions. The neural hormones provides several different areas, neural hormonal findings, echo findings. Those just give me additional; they add to the strength of that. Each one I think separately does when you have those three separate areas.

So I think it adds at least a modest amount to the value of the primary endpoint.

ACTING CHAIRMAN BORER: Steve.

DR. NISSEN: Ray, I think different secondary endpoints should have different value, and you know, for me endpoints that relate to how patients feel are much more important, much more weighty in my view than, you know, some very contrived line.

And so I think somebody, you know, that can, in fact, walk without dyspnea, that that ought to have a lot more weight than some finding of JVD on physical examination.

And so for me, while I was very impressed by the fact that this constellation was all going in the right direction, the ones that I think as we consider other applications that ought to have, you know, some weight in our minds are ones that relate to how patients feel because that's part of our goal as physicians: make people feel better.

And so that did have a lot of weight. Now, how much I think is very hard for me to tell you out of the context.

And I'm going to say something equally heretical and say that I view an active control trial a little bit differently than a placebo controlled trial in terms of what I'm looking for in the weight of evidence. You know --

DR. LIPICKY: Well, this is placebo controlled. We're talking about Val-HeFT here.

DR. NISSEN: Yeah.

DR. LIPICKY: And we're talking about whether you are willing to accept the primary endpoint --

DR. NISSEN: Sure.

DR. LIPICKY: -- and make decisions or whether you're going to let the secondary endpoints be the major things that influence your decision. That's what we're talking about.

DR. NISSEN: Yeah, I understand, Ray. I guess what I'm also trying to say is I think that the judgment that I might offer for a trial like this, I might not have the same opinion, you know, in another setting.

DR. LIPICKY: Of course.

DR. NISSEN: I can't take this out of context very easily.

DR. LIPICKY: Of course.

DR. NISSEN: Yeah.

DR. LIPICKY: But I just want you to say whether you're throwing the primary endpoint away, although it was okay. This was a single trial and you're basing your opinion primarily on the second endpoints which you admit you don't quite know how to evaluate except they look impressive.

DR. NISSEN: Well, I didn't say that I don't know how to evaluate them. I mean, I think some people said that, but I guess what I would say is that we fell somewhere between, you know, what we'd like to see for two single trials and we'd like to see for a single trial, and so --

DR. LIPICKY: Well, how far between?

DR. NISSEN: Well, we were closer to -- we were pretty close to what you'd want to see for, you know, a single trial.

DR. LIPICKY: No, no, no. No, no, no, no, no.

ACTING CHAIRMAN BORER: That wouldn't make it on a single trial.

DR. LIPICKY: Don't misunderstand what went on. It was .09. That's nominal p value, but you could double that. So that's .02. So what you want for two trials --

DR. NISSEN: Okay.

DR. LIPICKY: -- is half of .00125. So it's .0006. So the .09 is one and a half line segments off.

DR. NISSEN: Okay.

DR. LIPICKY: Okay? So it's not close in any sense of the word to "close."

Yeah, .009.

DR. NISSEN: Oh, oh, nine, yeah.

DR. LIPICKY: It's still orders of magnitude off, not a little bit, a lot.

DR. NISSEN: It's off some.

DR. LIPICKY: Some? Ten times.

DR. NISSEN: You know, we could quibble about what word to apply to it. I guess --

DR. LIPICKY: It's $100,000 to a million. That's how far off it is.

DR. NISSEN: Well, all I can tell you is I guess that I do think that the secondary endpoints have a considerable weight in my view, and in this case I think they help me a great deal.

DR. LIPICKY: Okay.

ACTING CHAIRMAN BORER: Okay. Did you have another comment, Ray?

DR. LIPICKY: No, but where are you in the questions?

ACTING CHAIRMAN BORER: We're about to hit the meat of the issue here, but, Paul, did you want to make another point here first?

DR. ARMSTRONG: I just wanted -- I mean, in my view, the overall effects on secondary endpoints are modest, and I'm guided by sort of five questions. One is the severity of the disease and its impact, mortality/morbidity in terms of what we're dealing with, and there's no argument about this disease.

The second is what other proven therapy exists for that disease, and there's no argument about that, and there's also no argument that the patients that we are looking at were treated arguably suboptimally, and we can argue why.

The third is the commonality of the problem and its public health implications, and there's no argument about that.

The fourth is the consistency of the effect across subgroups, and there are some concerns that we've heard here in terms of coming to a judgment.

And the final issue is the hazard of the therapy and its implications across subgroups, and we've had some discussion about that.

And so coming back to Ray's question, I put my line between ten and 15 percent relative in this disorder, and how close I am to ten as opposed to how close I am to 15 is modulated by trying to work within that --

DR. LIPICKY: Is that line segment one, two, three, or four?

DR. ARMSTRONG: It's between about .87 and .9.

DR. LIPICKY: I see. Okay.

ACTING CHAIRMAN BORER: We're about to enter the phase of dealing with those issues that Paul just talked about, and I think maybe we can come back to or maybe even obviate how important or how significant we think these data are, how important and significant once we go through the next couple of questions.

Because here's where the confounders really hit us. Consider the effects on mortality and morbidity endpoints by non-randomized use of ACE inhibitors and beta blockers, and here we have a table. We've seen these data.

Which of the following hypotheses are these -- with which of these are the following hypotheses are these data most consistent?

Number one, valsartan is an effective treatment added to ACE inhibitor and beta blocker. Can we say that? Can we not?

Valsartan is an effective treatment as an alternative to ACE inhibitor or beta blocker. Can we say that?

Tom, why don't you lead off and then this is something I think we need some discussion around the table with.

DR. FLEMING: All right. Could I ask the sponsor to put up two slides that I'd like to run back and forth very frequently? They're the ES-5 and ES-4 slides that I thought provided kind of a good overview of these issues.

And actually I might fold in a little bit of the Question 8 at the same time as addressing this.

The first part of seven relates to trying to interpret whether valsartan is an effective treatment when added to an ACE inhibitor and a beta blocker, and the second bullet point on eight says -- refers to the very small apparent effect in patients taking ACE inhibitors.

This is looking at morbidity. I'm going to jump back to ES-4 in a moment, but if we look in those taking ACE inhibitors, which is 92 percent, I think, of the study, we have approximately an eight to ten percent reduction in morbidity.

Go to E-4, please.

And approximately a seven percent increase in mortality. I look at that in a very subjective way as in Question 8 where it said the very small apparent effect. I look at it as actually data suggesting little to any effect, slight decrease in morbidity, slight increase in mortality.

Go back then to E-5, and now we're asking the question about addition to beta blocker, addition to beta blockers. Here's the cohort here of 1,750 people where there is for morbidity actually an estimated increase in morbidity not significant.

Go to mortality, and it is significant here in mortality, although I use the word "significant" with great caution in subgroups like this, but at least I wouldn't use the words at the end of Question 8 "the apparent treatment effect in patients." I would actually suggest if anything there is a slight suggestion of an adverse effect.

Let me go back though to E-5 because I'd like to really step back and look at these results more globally. It is relevant. Certainly I think the sponsor has made a very good case that it's relevant for us to give careful consideration about what is the evidence about effects in the presence or absence of ACE inhibitors and the presence and absence of beta blockers.

But it's really important to look at these numbers and realize that we already have to be very cautious about interpreting subgroups. I think there is -- we have three to five percent of the patients in the no/no category and in the no/yes category. In other words, what we're saying is 92 percent of these people are on ACE inhibitors. This study tells us in my view when we start subdividing into subgroups and then focusing on subgroups with three to five percent of the study, to my way of thinking that is really stretching past what these data are able to tell us.

So when I look at these data, I would focus as we have here first on the global results, but if we're going to then start looking at subgroups, it seems as though the subgroups that we can say something about are those on ACE inhibitors. It's 92 percent, and then subdividing that 92 percent into those that are on ACE inhibitors without beta blockers and ACE inhibitors -- excuse me -- those that are not on beta blockers or on beta blockers, and these two are essentially the same results. These are incredibly unreliable.

Essentially as I look at it what we have is the results indicate for morbidity that we have a slightly positive result, although use of beta blockers is a qualitative interaction where we would say, yes, there is a modest benefit when you're on ACE inhibitors adding valsartan, but if you're not also on a beta blocker, it's a bit more of a benefit, but it's a harmful effect if you are on beta blockers.

Now go to mortality, and the same basic issue. Here you have a slightly adverse trend when you're on ACE inhibitors, and again you have this qualitative interaction where there's a slightly favorable -- I think it's about a four or five percent decrease and a 40 percent increase, and then, of course, you see the exact same pattern here when you look at those that are exclusively on -- basically those all of whom are on ACE inhibitors.

These comparisons are 92 percent ACE inhibitors. These are 100 percent ACE inhibitors, but basically this is the story here. It's what is the effect for people on ACE inhibitors, and when you subdivide those on ACE inhibitors into non-beta blockers and beta blockers, you see a qualitative interaction.

My interpretation of these data are if you're on ACE inhibitors, which is 92 percent of the population that are on ACE inhibitors, you have a modest positive effect on morbidity, a modest adverse effect on mortality that seems to be a wash.

The data seem to suggest there's an effect modifier here. If you're on ACE inhibitors and you're not on a beta blocker, then it's a slightly or it's a modestly positive effect, but if you're on ACE inhibitors and on beta blockers, it's an adverse effect.

If we want to interpret this modestly positive effect, then we have to accept the negative effect here. I'd like to understand. Do we have a theory? Do we have a theory for why it's actually harmful to add valsartan when you're on ACE inhibitors with a beta blocker, but it might be beneficial when you add valsartan when you're on ACE inhibitors, but not a beta blocker?

It may be true, but in my experience in interpreting subgroup analyses, there has always been an historical caution used in interpreting subgroups and even greater caution with it's a qualitative interaction rather than a simple quantitative interaction, quantitative interaction meaning, yes, there's effect modification, but it's just a matter of how much benefit by subgroup.

Qualitative means harmful in one group, helpful in the other.

So in summary, there is certainly very interesting and important evidence here. I find it though very difficult to interpret why it is that valsartan could be beneficial when you're on ACE inhibitors without beta blockers, but harmful when you're on ACE inhibitors with beta blockers.

ACTING CHAIRMAN BORER: Okay. Why don't we go down the table and get additional comments about this issue? I think this really is going to be the crux of our final conclusion.

Alan, what would you say about -- take seven and eight together just as Tom did if you want to.

DR. HIRSCH: Well, I think Tom said it very well. I also pulled out the same two slides, and I think I'll just take off where he left.

Didn't we fail today to really examine this potential paradox? Because that is most worrisome to me. I'm concerned with how the implications of these data would be applied in real life practice where we ten to maximize treatment, add drug onto drug.

As Ray said earlier, this will be the seventh or eighth drug people are taking. So I don't know. I think to summarize, yes, in addition to an ACE inhibitor alone there is both a potential slight benefit, but potential wash. With both an ACE inhibitor and beta blocker together we have potential net detriment.

ACTING CHAIRMAN BORER: Let me just ask you to comment just a little bit further.

DR. HIRSCH: Okay.

ACTING CHAIRMAN BORER: Question 7, actually the second is not is it any good when you add it onto ACE inhibitor. The question is: is it an effective treatment as an alternative to ACE inhibitor or beta blocker, a slightly different issue for which we've have to make inferences based on those small groups as well as the larger subgroups that Tom pointed out?

And in answering this, you know, I want to mention again that there were eight different ACE inhibitors used, and the doses varied across the board, and the targets that Jay talked about that were sort of -- the on average were approached are targets of consensus, not of scientific database because we really don't know the dose response of ACE inhibitors for the hard endpoints for heart failure.

There are some data that suggest the more you give the better off you are, but we really don't have any dose response data. So we don't even know what doses people are supposed to be on. They certainly were on a whole load of varying doses.

The reason I'm emphasizing this, again, is that if you want to say that we can say something from these data about the additivity of ARBs or of this ARB to an ACE inhibitor, I'd like to know how we can make that statement. My opinion is that we can't say very much.

The issue, however, of alternatives to ACE inhibitors is something we need some comment about because that's not the way the trial was done, but we have some data. So --

DR. HIRSCH: Why don't we start the discussion and go around the table?

DR. FLEMING: Okay. I should have -- what you're pointing out, Jeff, is really there's another step I should have taken in my comments. Basically in my comments what I was referring to were the data that I believe at least answers the first part. Is an effective treatment when added to ACE inhibitor?

In my view it's very modestly effective in morbidity, but you have a comparable opposite effect on mortality. So I would say it's very minimal, and when added to a beta blocker, well, certainly it's unfavorable is the estimate.

ACTING CHAIRMAN BORER: With ACE inhibitor and beta blocker, yeah.

DR. FLEMING: Well, unfortunately, the only data I believe we really have to answer the added to beta blocker with any reliability is in the context of ACE inhibitors because there are very few people without ACE inhibitors and without beta blockers.

Is it effective as an alternative to ACE inhibitor? Well, it's a great study. We've got 5,000 people. Now we're trying to answer a question based on 366 if you're trying to answer that question. I believe you need a -- if you want an answer to that question, you need another study.

DR. HIRSCH: I would agree with that. The sample size is too small despite the point estimated and p value to be really confident.

DR. FLEMING: And then is it an effective treatment as an alternative to a beta blocker? There we have some data. We have the ACE inhibitor, yes, beta blocker, no, cohort, but here is where I want to probe because if we accept the favorable pattern there, ACE inhibitor, no beta blocker, then we should be equally accepting the reliability of the very negative result in the ACE inhibitor with beta blocker, and I and to know. My belief is if you're going to put stock in subgroups, there needs to be more than statistical evidence. There needs to be a strong biological plausibility that backs up the argument for why if you're on an ACE inhibitor and you're going to add valsartan when you're not on a beta blocker, that's a good thing to do.

But if you're going to add valsartan when you're on a beta blocker and an ACE inhibitor, that's a bad thing. I don't understand the biological rationale for that qualitative interaction.

ACTING CHAIRMAN BORER: Okay. Why don't we come back to try to answer that? I mean that gets us into the realm of speculation, and there are a whole load of mechanisms that one might propose to account for a negative interaction when you have multiple neurohumoral systems blocked at the same time and then you add a drug that might cause hypotension.

But rather than get into that, let's just move on for a moment to the specific questions here.

Tom gave a very comprehensive answer to what we have. Alan, you were about to discuss this again.

DR. HIRSCH: If I simply say it at face value to make it simple, let's take both subgroups and take the data as they are without speculating and assume that there is a potential value in the one and a potential adverse effect in the other.

ACTING CHAIRMAN BORER: Okay.

DR. HIRSCH: Let's go down and see if we see that clinical data the same along the group, and then I think we should bore into mechanisms a little bit.

ACTING CHAIRMAN BORER: Okay. Paul.

DR. ARMSTRONG: What's the question? Am I answering the first question here?

ACTING CHAIRMAN BORER: Number seven.

DR. ARMSTRONG: Right.

ACTING CHAIRMAN BORER: Valsartan is an effective treatment added to ACE inhibitor and beta blocker. Valsartan --

DR. ARMSTRONG: I don't know about the alternative, and maybe it looks like a wash to me relative to the first, and I would be interested in knowing if we're into subgroups whether the characteristics of these subgroups are similar or different.

ACTING CHAIRMAN BORER: Okay.

DR. LINDENFELD: I can't add to that, except I agree with Paul. We haven't seen a lot of that data to evaluate these subgroups, but I would be hesitant to do it.

The only data -- we have previous data that suggest that an angiotensin receptor blocker would be valuable in the absence of other drugs, and that strengthens a little bit this subgroup, but the subgroup not on any other therapy is so small that it makes me very hesitant to interpret it.

ACTING CHAIRMAN BORER: Can I just point out, I mean, just so that we keep it in mind because Paul has made a very important point here, and we're going to get to it again in Question 8?

The composition of the subgroups is something that is very important to know. However, the trial is a trial. These are the people we got. This is it. So we're going to have to extrapolate as best we can from, you know, an overwhelming group of Caucasians with a very small group of blacks, and on and on.

We can try and get down to smaller and smaller subgroups, but then we're getting into potential hot water here. I think we're going to have to take the population here as it was given to us and extrapolate as best we can.

Steve.

DR. NISSEN: I think we're trying to tease too much out of the subgroups here, and I'm really concerned that we can easily make a mistake by doing so, and you know, it's hard to interpret data, you know, with a few hundred patients in a subgroup either way, either affirmatively or negatively.

And so my view is that we have to take the trial as its entirety and say that in a group of patients and maybe even describe this; and in a group of patients, 93 percent of whom or 92 percent of whom were receiving ACE inhibitors, valsartan produced the following benefits.

And I am unwilling to make any strong commitment based upon subgroups in a single trial, and I think that this is hypothesis generating for those subgroups, but it is not proof, and therefore, we ought to be very careful here and fairly narrow in our interpretation.

ACTING CHAIRMAN BORER: Mike, what about the safety issue here? I mean, you know, a negative result we generally don't require the same degree of statistical certainty when we see detriment, and we've seen detriment here. Mike, can you talk about that a little bit?

(Pause.)

ACTING CHAIRMAN BORER: Or talk about whatever you'd like to talk about.

(Laughter.)

DR. ARTMAN: No, let's get back to Ray's watch.

No. Yeah, I am concerned about the addition of valsartan to a patient who's on a beta blocker, right. And I agree with everything that's been said coming down the table, and I'm not sure I can shine any more light on this, and to try and make some judgments on these subgroup analyses I agree is very difficult, and we're in treacherous waters.

But I think that the data are sufficient to make me very concerned about using this in a patient who is on a beta blocker, and so I think that that part of the question I feel we can't answer.

ACTING CHAIRMAN BORER: JoAnn?

DR. LINDENFELD: I just think, again, this is a difficult subject, but whether or not we like the subgroups or want to evaluate them, if we choose to recommend this drug for the treatment of heart failure, we're recommending a group of people of whom probably 25 percent it was estimated, maybe a little less than that, but at least 25 percent are taking beta blockers.

And I think we're going to have to give people, practicing physicians, some idea of what we think after reviewing this data. I don't think we can just let that drop and say we can't evaluate it. That has to be part of this process.

ACTING CHAIRMAN BORER: What would your opinion be?

DR. LINDENFELD: Well, I'm very hesitant when I don't know. I'm concerned. There's a very modest benefit to this drug, and there's some potential for harm in at least a quarter of the patients that we're recommending it for, and that makes me feel less positive about the very small benefit that we've seen.

ACTING CHAIRMAN BORER: Steve?

DR. NISSEN: There is an alternative. There's a way around this, and the way around this is in the context of giving an indication for the drug doesn't preclude adding to the label a warning that in a single trial the combination of the three drugs together in a small number of patients showed an adverse outcome, and so you've told the prescribing physician that the drug has an effect in a broad population, but there was a subgroup.

And that protects, I think, the patient, the physician, the agency. It protects everybody until further data can be collected to confirm or refute this notion that triple drug therapy is bad, and that would be to me a logical approach.

ACTING CHAIRMAN BORER: Alan?

DR. HIRSCH: Can I take this? Because I was going to take you to task, my distinguished friend on the left, Dr. Nissen, because your usual carrying forth that we should not look at subgroups I usually respect, but I've been so quiet all day, troubled by the subgroup data.

And this is not a small subgroup. I mean, 1,600 --

DR. FLEMING: Right.

DR. HIRSCH: -- more or less patients --

DR. FLEMING: Steve, I wanted to follow up on that, too.

DR. HIRSCH: -- is significant.

DR. FLEMING: Yeah, it's 1,750 in what you've called a small subgroup. Can we flash up --

DR. HIRSCH: I'm not quite done.

DR. FLEMING: Go ahead. Go ahead.

DR. HIRSCH: The subgroup actually following Dr. Cohn's line of reasoning is actually hopefully going to be growing as we become more aggressive in the treatment of heart failure, and we don't quite know yet what's down the pike in terms of the ability to both educate physicians. Other trial data will have us adding more and more agents.

So when I look at this historically, actually the first signal where we finally hit sort of a wall potentially, we should as an agency or as an advisory panel be concerned.

So just that point has been made. I feel very strongly we have to respect that subgroup at least as a -- more than a hypothesis generation, but as a time right now to educate physicians to be careful.

ACTING CHAIRMAN BORER: Tom, did you have an additional point you wanted to make?

DR. FLEMING: That was well spoken. I was just going to, if we could put up for a moment again E-4 and E-5.

Just to reiterate, here is where the data are. The data are in these three groups, and, Steve, this is a -- I mean, your comments are wise, causing us -- encouraging us to be cautious about subgroups. I consider that a very wise comment.

At the same time, there is something here that's not easy to walk away from, and I would concur with the sponsor raising this issue as one that has to be seriously considered. I look at the analysis in these three groups as where the essence of the information is, and if we look at this as essentially what we know the most about, basically, Steve, it's the global analysis. It's a very slight harmful effect in ACE inhibitors.

If we went to E-5, it would be a very slight positive effect on morbidity as you see right there.

When we subdivide into these subgroups, there is a fair amount of data in these subgroups. It leads me to state or leads me to conclude that adding it to an ACE inhibitor is a very, very negligible situation. Adding it to a beta blocker, if I believe the qualitative interaction, is a bad thing to do.

And I keep putting forward to my colleagues here if we look at this and go to E-4 as well, go to E-4 as well, we see this exact same qualitative interaction.

And if we're going to believe that we should market it in these people who are on ACE inhibitors without beta blockers, I want to understand why we think that when you're adding it to the ACE inhibitor and the beta blocker it's bad. I need to understand the biology for why you've got an ACE inhibitor on board.

If you're going to add valsartan without a beta blocker, that's a good thing. With a beta blocker, it's a bad thing. And if I can't believe that, then I really want to go back to Steve's wisdom and say really the only thing I can take home from this is the overall ACE inhibitor result, and I will say I want two studies then. If you really want to understand the answer to the question do you provide it when you're not giving an ACE inhibitor, we need a trial, and those aren't ACE inhibitors.

We only have 300 patients of that type, and if you really want to know if it's an agent that could be given as Jay Cohn suggested, I believe, in his presentation, for people who are not able to take a beta blocker, I don't know that that's this group. I just know that this is a group that wasn't on beta blockers. I don't know that this was a group that couldn't take beta blockers.

DR. HIRSCH: So to accentuate that further, the bet blocker treatment almost looks like a light switch in terms of benefit on and off.

DR. FLEMING: Yes.

DR. HIRSCH: And it is hypothesis generating. So I hate to do this. We usually just talk amongst ourselves, but I have to ask that question, which is: did we see or did I miss in the application something that I'm suspicious as a clinician, which is as I add agents, I often face the wall of hypotension. Did we see blood pressure trends across these beta blocker, ACE inhibitor plus/minus treatment groups to say whether we had finally unloaded the patient so much?

Did we see it or did we ask for it?

DR. COHN: We can show you that.

DR. HIRSCH: Please. Well, then if it doesn't, it will get them to stop worrying, and we can move on to the next subject.

DR. COHN: WE have the blood pressure.

DR. HIRSCH: You must have looked at it.

DR. COHN: Oh, yeah.

DR. HIRSCH: Or some other hormonal paradox.

DR. COHN: All the explanations. Obviously Tom is raising the biological issue, and I think it's wonderful to have a biostatistician want to raise a biological issue.

These are the blood pressure changes in the four main subgroups and then the four combined subgroups, and you'll notice that there's no trend for more blood pressure reduction.

DR. HIRSCH: Thank you.

DR. COHN: Now, these are the last blood pressures recorded in those patients before they died or the trial ended. They aren't measured contiguous with their death, of course, but there didn't seem to be any striking trend that there was a greater blood pressure reduction in the combined therapy.

ACTING CHAIRMAN BORER: There is something there though, Jay. I mean, it's flashing at me too quickly to do a calculation, but the bottom group actually has the lowest blood pressure to start with, and then you're lowering it about the same as everybody else is being lowered.

So, in fact, it may be that the final absolute blood pressure you reach is a little bit lower with yes/yes than with the other groups.

DR. COHN: It could be. It's a pretty small difference I would agree with you in the valsartan group. In the placebo group, the baseline was not lower. In the valsartan group it was on the combined drug.

So, you know, you get into small numbers here. If I could just make a couple of comments on what's going on, and I don't really want to intervene myself, I share with all of you the concern about what to do. The reason we've brought this forward is because of the safety issue on the mortality in the combined treatment group. ACE is yet, beta yeses.

I don't quite share Tom's view that there's a mortality adverse effect. The confidence intervals overlap one, and it could just as well be that there's a mortality benefit.

I also think despite the small numbers that there is some virtue in looking at the combined drugs, not just the first stage because it clearly, both by secondary analyses that Al showed you and by everything else that we have, the quality of life and the ejection fraction, et cetera, there seems to be a difference whether you're on a beta blocker alone or on a beta blocker with an ACE inhibitor.

And I think that that makes a big difference, and from a mechanistic standpoint, I do believe it is multiple drugs, and we now have evidence from other trials that have recently been completed. We've demonstrated that if you lower plasma norepinephrine, pharmacologically with a central inhibitor of the sympathetic nervous system, you get an adverse effect on mortality.

We've demonstrated that when you block endothelia these days in a most recent study or you block cytokines in patients treated with all of these other drugs, you seem to see no benefit and, in fact, a trend for an adverse effect.

So we may be getting to the point where there's too many systems being blocked. So biologically I don't have a lot of trouble with this, Tom, although I can't cite you a mechanism, but it just is intuitive, and I think Alan has sort of said the same thing, and so has Steve.

You know, it's just a little too much blockade, and I don't know whether it's working through blood pressure or through conduction or through something else, but it puts the patient at some risk potentially.

Now, we're going to have more data from the CHARM trial. We're also going to have a lot of data from the VALIANT study post MI on the combination of all three drugs: beta blockers, ACE inhibitors, and ARB.

Those trials are ongoing. The data safety and monitoring boards for those trials are very aware of our data. We keep shipping them updated data. So they're watching it, and they have chosen not to stop the arm with ARBs added to ACEs and beta blockers.

So we're going to have a lot more data, and I think we're at the cutting edge of this now. How do we respond to this subgroup in terms of labeling of the drug?

Now, let me just say one more thing about the p value for the whole study, and this is to help Ray a little bit. One of the reasons -- I always like to help Ray.

DR. LIPICKY: Thank you.

DR. COHN: One of the reasons that the company did the exercise study was that they were told that would be a second trial, and if that had been positive, .05, that would have been the second trial, and all this trial had to do was achieve .025 or .020, whatever it was.

That trial was not positive, as you've seen. It was a wash. Now, I appreciate the committee telling Ray whether a positive exercise test on a 12 week study of .05 would have been more valuable to you than all of the secondary endpoint significance that we've shown on ECHO and LV dimension and quality of life and signs and symptoms and neural hormones. Is that more valuable or less valuable than a 12 week exercise test would have done for the statistics?

Because from the regulatory standpoint, that study, .05 on exercise, would have meant it brought us home free.

ACTING CHAIRMAN BORER: Before we try to answer that question, let's try to answer the questions we have. I'd like to give my opinion before we move on to number eight here about number seven.

It's just a little bit different from what you've heard, and I throw it out to everybody for whatever it's worth. As I suggested earlier, I think safety issues have to be dealt with a little bit differently than efficacy issues. We may demand a great deal of strength of evidence, a great strength of evidence to conclude that efficacy exists, but I don't think we need quite so much evidence to suggest or to conclude that there's real potential for a safety problem if you see some data that suggests that.

And I think we see it, but I don't see it, though the statistical gods may kill me. I don't see it as a beta blocker issue. Virtually everybody who is on beta blocker was on ACE inhibitor. That's more than 1,600 people, and it's the combination of the beta blocker and the ACE inhibitor that was associated with the bad outcomes on all the things we looked at, and that was consistent across the board, even with the secondary endpoints and the tertiary endpoints, you know, everything that was looked at here.

So my concern, and I don't have any problem. You know, I would echo what Jay says. I don't know what the mechanism is, but I could argue in favor of one.

I think that we have to make a comment about the inappropriateness at this moment of adding valsartan to a combination of beta blocker and ACE inhibitor. When I look at the very small subgroups that you couldn't draw any conclusions from on their own and see that the people who were on beta blocker alone, small subgroup though it may be, look different, and that is intuitively not unreasonable to me.

Then I'm less concerned about adding an ARB, valsartan specifically, to people who are on beta blocker alone. When I look at the ACE inhibitor data, however, I fall right on the line with Tom. It looks to me no matter how you slice it that there's a little bit of a benefit when you're on ACE inhibitors alone, but I don't know what dose, and I don't know what drug.

In terms of morbidity and equally a little bit of a detriment in terms of mortality, and I would caution people about the addition of valsartan to an ACE inhibitor. I can't say it's bad. Overall it looks like it's a little bit good, maybe, but I don't know the dose. I don't know the drug. I don't know the combination that's appropriate.

Since I don't know, I'd like to say I don't know, but I wouldn't want to proscribe doing it because I just don't know enough, and when you look at the totality of the data, in fact, it looks like there's a benefit.

And I want to make another editorial comment here. I've been looking at NDAs now off and on for 24 years, and I made this comment at a conference once when I was sitting in the audience, and the response I got I'll tell you in a moment.

But I don't know how any drug works. I know the pharmacological effects that are associated with a lot of drugs. I have no idea what the mechanism of action is, that is, how the drug produces its clinical benefit.

And I can cite chapter and verse of disapprovals based on the lack of a putative mechanism of action for drugs that we now know have exactly the same effects as other drugs that subsequently have been approved and now we think we know the mechanism of action.

So I look at the data first. When I said that at a meeting in which I was sitting at the back of the room, I said that to somebody else sitting back there, and he said, "Well, gee, you'd better start reading the journals and reading the textbooks."

And I said, "You'd better hope that's not true because," I said, "I'm the committee that approved these."

But I think we have to look at the data first, and to me the data show for the population that was studied a clinical benefit, not all of the clinical benefits that we would have liked to have seen. When I look at the subgroups, I see what intuitively I would have expected, that in the subgroups that didn't get the other drugs, the benefits that I would have expected with at least one of the other drugs is there.

So I don't have a problem with concluding that this drug does something good, but I'd sure as heck not want to give it to people for whom I have a strong signal that I'm going to hurt them when I do it, and that to me is the group that's taking this combination of beta blockers and ACE inhibitors.

So that's my opinion. Put it in the hopper, and we'll move on to Question No. 8.

Evaluate the following findings with respect to whether they are considerations related to approval or to labeling, the lack of apparent treatment effect in blacks, the very small apparent treatment effect in patients taking ACE inhibitors, which we just now talked about, lack of apparent treatment effect in patients taking beta blockers we talked about.

I think the one issue to deal with here is the lack of apparent treatment effect in blacks. We've mentioned it. Is there anything more that we want to say about it? Is it something that should be highlighted in some way if we were to give labeling advice to the FDA?

Paul.

DR. ARMSTRONG: Jeff, having picked up on that this morning, I would just say that looking at the data that has been presented that wasn't in our briefing book, there are really four factors, I think. One is the mortality. The other is the morbidity, both of which go the wrong way.

Then there's the BNP, which goes the wrong way, and then we've learned that there's also the safety that goes the wrong way.

So there's a quartet of factors that for me are concerning, notwithstanding the fact that we're talking about seven percent of the population or about 360 patients.

ACTING CHAIRMAN BORER: Do we know at all -- and, again, I don't want to get into sub-sub-sub-sub-analyses, but do we have a gestalt of how what other drugs the black people were taking? I mean, were a lot of them taking the combination of beta blocker and ACE inhibitor, for example? Do we know at all?

MR. MacNAB: No, they were very similar. There were, I think, to some degree fewer beta blockers. The black patients were a little younger, but, again, in a small group with a wide confidence interval it's hard to make definitive conclusions.

ACTING CHAIRMAN BORER: Okay. So Paul has verbalized a concern, a real concern, that maybe has to be highlighted as we move forward. Would anybody disagree with that?

No. Okay. Let's go on to number nine then. Has adequately information --

DR. LIPICKY: Hold it --

ACTING CHAIRMAN BORER: -- to describe instructions --

DR. LIPICKY: Wait, Jeff. You're skipping a couple of things. That said is that -- is you statement that conclusion with respect to provability or labeling? Because --

ACTING CHAIRMAN BORER: Oh.

DR. LIPICKY: -- you made it sound like, of course, it's just a labeling issue, and it doesn't influence my approvability conclusion

ACTING CHAIRMAN BORER: Well, we haven't talked about approval yet.

DR. LIPICKY: Well, it says how do you evaluate this.

ACTING CHAIRMAN BORER: Oh, I'm sorry.

DR. LIPICKY: Approvability or labeling?

ACTING CHAIRMAN BORER: No, you're quite right. You're quite right. Okay.

DR. LIPICKY: Because up until now it's been approvability, and this starts to get mixed now.

ACTING CHAIRMAN BORER: To me it's a labeling issue. I'd like to hear what everybody on the committee has to say.

Go ahead.

DR. HIRSCH: Labeling.

ACTING CHAIRMAN BORER: Paul?

DR. ARMSTRONG: I'm okay with that.

DR. LINDENFELD: I think it's an isolated thing. It's a labeling issue, but with a number of other subgroups we have questions about, it might be an approval issue.

DR. NISSEN: Label.

DR. ARTMAN: It's a labeling issue.

ACTING CHAIRMAN BORER: Glorea?

DR. ANDERSON: I think it's a labeling issue, but I also have some concerns because, one, the population was small and, two, I couldn't find enough information to answer some questions that I had.

And incidentally, I had the same question about the size of the population of women who were included in the study, 20 percent, I think it is, about 20 percent.

ACTING CHAIRMAN BORER: Tom?

DR. FLEMING: I have nothing to add.

ACTING CHAIRMAN BORER: Okay. Now, let's go on to number nine. Has adequate information been obtained to describe instructions for the use of valsartan in heart failure?

Would anybody like to give an answer and then we'll see if there's a lot of dissent?

DR. FLEMING: Can I have a clarification? Does this include if in Question 10 it's the perspective of some committee members that one needs to take into consideration whether one is on ACE inhibitors or beta blockers or whether they're contraindicated, is that part of -- for example, what we don't know, I would argue, is what's the level of effect of valsartan in someone on ACE inhibitors where beta blockers are medically contraindicated.

ACTING CHAIRMAN BORER: Yeah. You know, I don't want to answer for the FDA, and I think we'll have an answer from the FDA in a second if I say something incorrect, but at the end of a development program, there are many questions that are left unanswered, and if we have enough information to provide instructions for use, which also can provide instructions about what we don't know so that you ought to be very cautious and maybe not even do it until more information is available. We can do that.

We can provide a very directive or the FDA can provide a very directive label. It can say you should only do this in this situation.

We don't know anything about this. This is a potential show stopper. Don't do it till we have more information.

So I think that the question has adequate information been obtained is a question about how well we believe we could describe to a physician how the drug could be used effectively and acceptably safely today.

That may exclude a lot of groups. It may exclude a lot of drugs. It may do this. It may not.

DR. FLEMING: Then my sense --

DR. LIPICKY: And it includes those.

DR. FLEMING: -- is there are some additional sources of information, but I'd like to clarify my answer to that after I answer Question 10.

ACTING CHAIRMAN BORER: Ray, do you want to add to what I or to --

DR. LIPICKY: No.

ACTING CHAIRMAN BORER: -- refute what I said?

DR. LIPICKY: No, but it includes dose. And I'll take just a minute, and I know you're in a hurry and want to get done in ten minutes.

ACTING CHAIRMAN BORER: No, no. We'll give you a few more minutes.

DR. LIPICKY: But, you know, the sense of these questions as they have evolved up until now was we want to know whether you think a single trial gets approval and whether it gets approval on the basis of its primary endpoints or its secondary endpoints or a combination of the two, and whether you think the subgroups that are here are adequately enough portrayed that they cause concern.

So it's possible that the overall trial result might be weakly positive. Let me put it that way. Okay? And that the subgroup business makes you worried about not knowing who to give it to, in which case you wouldn't care about Question 8 because you don't have enough -- you know, the dose and stuff like that doesn't matter or, conversely, that the principal -- the primary endpoint is so convincing on its own that it absolutely has to be approved for that, and that the rest of this is all just window dressing then.

Okay? So we're -- it's sort of been graded through this whole business of what is most important and what is next most important and trying to get a sense of what you think. I'm not sure I did, but I don't know why I said this.

Forgive me.

ACTING CHAIRMAN BORER: Okay. Let's move on then to Question 10, which we'll have to take in parts. And for this we'll need a vote from everybody, I think.

Should valsartan be approved for use in treatment of patients with chronic congestive heart failure, and if so, what should labeling say about these various things?

Let's start with the global issue because if the answer were no, then we have nothing else to add to talk about.

Should valsartan be ap-proved for use in the treatment of patients with chronic congestive heart failure? Let's start at the far end of the table. Glorea, why don't you go ahead?

DR. ANDERSON: I would disagree at this point based on the fact that I don't think we have enough information. At least I don't.

ACTING CHAIRMAN BORER: That's a no.

DR. ANDERSON: No.

ACTING CHAIRMAN BORER: Mike.

DR. ARTMAN: I would say yes.

ACTING CHAIRMAN BORER: Steve.

DR. NISSEN: Yes.

ACTING CHAIRMAN BORER: JoAnn?

DR. LINDENFELD: I would say no. I think that the endpoint here doesn't meet the level of statistical significance that we want, and it's a modest improvement, and then we have major questions about subgroups and who to treat.

ACTING CHAIRMAN BORER: Paul?

DR. ARMSTRONG: Overall, no, but I think there's a niche.

ACTING CHAIRMAN BORER: I'm sorry?

DR. ARMSTRONG: Overall the answer is no, but I want to come back to potential subgroup.

ACTING CHAIRMAN BORER: Well, okay, but we can't. If the vote is no, then it's --

DR. ARMSTRONG: All right.

ACTING CHAIRMAN BORER: Okay. Alan?

DR. HIRSCH: I was going to say yes.

ACTING CHAIRMAN BORER: tom?

DR. FLEMING: I actually had a similar response to Paul. It's a no, but it's a qualified, and I will make very clear what that qualification is before we finish answering these questions.

ACTING CHAIRMAN BORER: Okay. I'd vote yes.

How does that come out?

Okay. Now we have to get some qualifications because it's four to four. Tom, why don't you start with your qualifications?

DR. FLEMING: Well, let me --

DR. LIPICKY: You've helped us a lot.

DR. FLEMING: -- comment on a couple of things that --

(Laughter.)

ACTING CHAIRMAN BORER: Be could stay until tomorrow.

DR. FLEMING: Let me comment on a couple of things, and actually the qualifications relate to the specifics in 10-1 and 10-2, but there's one or two comments I haven't given yet, and one of them relates to just interpreting these data first on the primary endpoints. The first is mortality.

I believe the study is, in fact, more reliable on its primary endpoint than might have been apparent in the sponsor's presentation of the study. Looking at mortality first, it was pointed out that the anticipated death rate was 12 percent. It was only observed to be nine percent, and that may have left the study under powered for mortality.

And in the presentation it was mentioned there was no demonstrable effect on mortality, which suggests that maybe there is an effect, but we just didn't demonstrate it.

The study was targeted for a 20 percent reduction in mortality. It seems to me in the context of other agents that are out there, such as beta blockers and ACE inhibitors that provide more than that level of effect, I think it was reasonable to have targeted that level as what was clinically relevant.

The study did achieve essentially 1,000 deaths, which by my calculation is a high power for detecting that 20 percent reduction, even with the adjustment for the two primary endpoints. The estimate is a two percent increase in mortality where the lower limit of the confidence limit is .9, meaning it rules out half the level. These data are inconsistent with even as much as half the level of mortality effect that the study was powered to detect, half the level, less than half the level of effect that we would know we can achieve with other agents.

So my sense is this was an excellent study in many ways, and certainly one of those ways was in providing us a very good sense about the effect of mortality. I believe these data are not only not significant. I believe these data are suggestive of no effect and ruling out anything more than a modest effect on mortality.

We've already discussed at greater length the morbidity endpoint. As I see it for 100 percent, what we're doing is we're presenting ten hospitalizations over a two year period per 100 people.

We're also, as was corrected, we're preventing one day or reducing one day hospitalization per year. That is a modest -- that is a moderate, modest, whatever adjective, benefit you want to put on, you want to acknowledge, and the effects on symptoms and the Minnesota living with heart failure are reinforcing, although I'm still struggling with how strongly because I'm still struggling with getting a sense of how strong those data are.

Having said all of that then, if we look at the data where we have information, where do we have information? We have a lot of data in the global analysis, and like Steve says, I look at that first and foremost, and when I look at that, I see one study that does meet standards for strength of evidence for a positive trial. I'm really reluctant to call it though the level of evidence that would be similar to what we would have from two independent studies, each of which would meet that standard for positivity.

It was possible that we could have met that standard if we had had, rather than modest, if we had had moderate effects on morbidity. This was a very large trial that would have been powered to achieve that level of effect.

Now, what adds a lot of complication here is the sponsor's acknowledgement that there is evidence here about potential effect modification. As I look at it, the data on effect modification are specific to refining the question about what it means to add valsartan to ACE inhibitors because we don't have data of any substance in people that aren't on ACE inhibitors.

So in that context, when you're adding to an ACE inhibitor, what we have is, as has been mentioned many times, a modest positive effect on morbidity, but a comparable modest negative effect on mortality. The mortality confidence interval, as Jay points out, includes equality, but so does the morbidity confidence interval include equality.

What we're left with then is this complex issue about whether there is an effect modifier such that it's a good thing to be on an ACE inhibitor; it's a good thing to be on an ACE inhibitor and beta blocker, but in the former case, it's good to add valsartan to the ACE inhibitor. In the latter case it's bad to add valsartan to the ACE inhibitor and beta blocker.

I don't understand that. I don't understand that. If the FDA understands that, then I would argue approval in the context of patients who are on ACE inhibitors and not beta blockers, if the FDA understands the mechanism for that interaction.

I would argue that, in essence, coming back to where I left unanswered in Question 8 and Question 9, what would I like to know that I don't know. What I really would like to know is what is the role of this agent in the setting in which ACE inhibitors are contraindicated. There's too little data to answer that here. It's a subset analysis, much worse.

Secondly, what I don't really know -- I've got clues, but I don't really know -- is what is the effect of adding valsartan to an ACE inhibitor when a beta blocker is medically contraindicated. I don't know. That is also unknown, and that could be addressed in a second supportive trial if, in fact, the FDA remains as uncertain as I am as to what's causing this critical effect modification.

So in summary, my sense is clearly in the answers, in my view, the answer is this an alternative to an ACE inhibitor, is this an alternative to a beta blocker, I think the sponsor answered that question. It's not an alternative we would wish to give beta blockers and we would wish to give ACE inhibitors in settings in which they're not medically contradicted. So the question is: are these agents that would be given in second line?

And my belief is there is additional data that's necessary, but bottom line is if there is a clear understanding or a reasonable understanding of this critical effect modification issue, then I would be more positively persuaded toward an approval for the setting in which somebody is on an ACE inhibitor, but beta blockers are medically contraindicated.

ACTING CHAIRMAN BORER: Okay. Does anybody have any others? Paul.

DR. ARMSTRONG: Well, I guess I should justify my vote, and it's the concern about safety and the uncertainty about questions, questions in 15 percent of the population over 75 in which we have no information, concerns about the spironolactone story, uncertainty about the effect in patients on digoxin, clear concerns about the beta blocker issue.

And so notwithstanding the fact that I believe this drug has an effect, as a clinician trying to inform others as to how to use it with the evidence available, I wouldn't know what to say, Mr. Chairman.

ACTING CHAIRMAN BORER: Steve?

DR. NISSEN: Yeah. I hear everything that, you know, the folks saying no are saying, and I understand your convictions, and I appreciate them very much. Let me just think out loud with you a little bit about this because we're obviously on the fence here.

We all look at a trial, I think. The most compelling data is obviously the data that relates to the primary prespecified endpoint of the trial, and I want to point out to the committee that this sponsor and these trial investigators set an extremely high bar for themselves. They took a bunch of patients that were very well treated with dig., diuretics, 90-plus percent getting ACE inhibitors, a lot getting beta blockers.

These are much better treated patients than the average heart failure patient in America or anywhere else is treated, and they said, "Would adding valsartan to a group of very well treated patients do anything?"

What did it do? Well, for one of the two primary prespecified endpoints at a p value of .009, not .00125 --

DR. LIPICKY: Oh, oh, two.

DR. NISSEN: Yeah, okay. Okay, all right. Again, we can --

DR. LIPICKY: -- oh, two.

DR. NISSEN: Okay, but at a level of significance we can argue about and supported by a whole constellation of symptomatic, functional, structural, and biochemical endpoints.

And so, you know, looking at this on balance and trying to decide, you know, whether there's more harm or good here, you know, I think you have this trial as to live or die by that primary endpoint, and I am influenced by the fact that this endpoint was obtained in a setting of extremely well treated patients.

Now, the big problem is we've got this subgroup. I don't even know if it was a prespecified subgroup. Maybe it was; maybe it wasn't, where something fell out that we didn't like, and I do think we have an ethical duty to make sure people are informed about that.

And I, therefore, think that there is a compromise position here, which is to come up with some labeling that suggests that this agent may be useful because I happen to think it's a good thing to prevent PND, dyspnea on exertion, and hospitalization.

But to provide very clear warnings that triple drug therapy was associated with increased mortality and morbidity, and let the prescribing physician then make a judgment about that. I personally don't want to give this drug in triple drug therapy, but I think I might well add it to patients, particularly those who are still quite symptomatic, who have heart failure and are on ACE inhibitors.

And one more point I want to make is that if you look carefully, the worst the heart failure was, the more the efficacy signal was in this trial, and that to me suggests to me that if I have a patient that has fairly severe heart failure symptoms and is not adequately managed, you know, with current therapy, that I could add valsartan and get additional benefits, and I think that's the take-home message of the trial.

We need the warning in there, but I think the efficacy convinced me.

ACTING CHAIRMAN BORER: Tom and Alan, both have comments.

DR. FLEMING: I just wanted to query Steve about his thoughts. You had mentioned at the beginning of your comments, Steve, that the sponsor basically sat a very high bar, a high standard, and you explained that in the context of having tried to show that there was additional benefit to adding valsartan in the context of patients who are already well treated with ACE inhibitors, beta blockers, et cetera.

I would argue the challenge for any of us as sponsors and investigators is to address the efficacy and safety of our intervention in the real world context of how they would be delivered.

Are you arguing that there was a lot more ACE inhibitor and beta blocker use in this trial than there should be in the real world, and as a result we were assessing this in a setting in which there was too high a goal to hit?

DR. NISSEN: No, I guess, Tom, what I'm suggesting is that it's very challenging to show efficacy on top of good therapy, and therefore, I give some significant sort of weight to the significance of those p values when I understand the context in which the therapy --

DR. FLEMING: But I would agree with you that in many instances in clinical practice it is harder to incrementally improve upon clinical practice when that clinical practice has already reached an effective level of benefit.

But nevertheless, the reality is fortunately we are in a setting now where we have these effective agents, and so the real question is: can we improve on what we already are able to accomplish with those agents?

I thought this was a very good study that was actually answering the question that was in need of being answered.

ACTING CHAIRMAN BORER: Before we go on to Alan and to Ray, let me -- you wanted a biological discussion, Tom, and let me suggest my thinking about this trial, this development program, and this drug.

I must tell you I think it took extraordinary courage for a drug manufacturer to take an angiotensin receptor blocker and study it for this indication with these kinds of a priori projections of effect in people who are being treated with a drug that affects exactly the same neural hormonal system. I never would have expected that an angiotensin receptor blocker would have any particular effect on top of an ACE inhibitor.

It might. You know, you saw Jay's slide with the putative mechanism by which maybe you could get some effect, but I wouldn't have expected much.

You know, I think because of analogy with results with angiotensin receptor blockers and ACE inhibitors in other settings that one might be a replacement for the other. I don't know that for sure in this setting because, of course, that's not the way the trial was set up, although if you look at the subanalyses, they're consistent with the hypothesis I'm suggesting now, that one can be a substitute for the other.

It would have been extraordinary to me that adding a drug of this particular class on top of the drugs that were being used that you would see a tremendous additional beneficial effect.

Nonetheless, I'm impressed that we saw something. We actually saw a reduction in morbidity, and again, I don't want to make too much of small group analyses and all of this kind of stuff, but starting with the hypothesis with which I began, that is, how this drug works pharmacologically, what system it's affecting, I would have expected there wouldn't have been much of an effect in the group as a whole, but if you looked at the subgroup that wasn't getting the other drugs, you would have seen an effect, and, lo and behalf, we did.

So let me just finish. You know, what I saw is what I would have expected to see. It obviously isn't what the sponsor expected to see, but it's what I would have expected.

I think that when you ask how could it be that it's beneficial when added to an ACE inhibitor and not to an ACE inhibitor with a beta blocker or whatever, I would suggest that a lot of people who are on ACE inhibitors were on relatively low doses of ACE inhibitors where addition of the ACE inhibitor could have given you additional benefit just as addition of the ARB might have given you additional benefit.

So I have no trouble with seeing how there could have been some people in the group that drove the group as a whole to show an additional benefit when valsartan was added to ACE inhibitor.

I still don't think, given to maximally tolerable doses, whatever those may be, that you would see such an effect. That's my bias. I don't know if it's true or not. Maybe the data could be plumbed to see if there's a cut point in the doses of drugs that were used to see whether the addition of the ARB was better with the higher dose or the lower dose or if there was any difference at all.

But that would be my bias. When you add the two drugs together, the ACE inhibitor and the beta blocker and, thus, block a great deal of the neural humoral activity, I could then see how adding another drug could cause a problem.

So this doesn't seem intuitively unreasonable to me.

To talk about giving the drug to people who medically can't take a beta blocker for whatever reason, the largest subgroups of those patients are people with pulmonary disease and a smaller subgroup with diabetes who can't be easily controlled on a beta blocker. It's not for cardiac problems.

So you know, that issue of people who medically cannot -- for whom beta blockers are medically contraindicated seems to me to be a side issue. If it were people for whom ACE inhibitors were medically contraindicated, that might be another issue, but even there Steve said it before. The primary reason why people don't get ACE inhibitors when we think that, in general, with their disease pattern they should is that they cough, and it's annoying to them. So they don't get it, and then they're left with nothing except a beta blocker alone if we use the current algorithm for treatment.

I'm looking at these data and saying that valsartan represents a reasonable drug to give to those people. Now, was a study done to test that hypothesis? No, but the data from the study that was done are completely consistent with what I'm suggesting.

Now, that may not be a sufficient basis for approving a drug, but that's the way I would look at these data.

DR. FLEMING: Jeff, I think there's a critical distinction to be made in what you're saying. I'm quite sure I heard you say that you interpreted these data to be suggestive that you could give this agent instead of another neural hormonal inhibitor.

ACTING CHAIRMAN BORER: Well --

DR. FLEMING: And the data are suggestive that you would achieve a comparable effect. I think these data tell us essentially nothing about that question.

What these data are telling us is in the absence of these other neural hormonal inhibitors, there's evidence of some benefit, but there's nothing to say that that level of benefit matches what you would have gotten if you had randomized those patients against the beta blocker or the ACE inhibitor.

In fact, I think there's strong evidence to suggest that if you did randomize these patients to an ACE inhibitor or beta blocker, the ACE inhibitor and beta blockers would substantially improve survival, and valsartan wouldn't affect survival because that's what these data are showing.

These data are though showing that you're affecting --

ACTING CHAIRMAN BORER: I don't think so.

DR. FLEMING: Let me finish. Let me finish. These --

DR. COHN: These drugs were given on top of those drugs. The only group where they weren't on the top showed a benefit on mortality. So you can't say that.

DR. FLEMING: Well, that's correct. These data were given for the most part on top of ACE inhibitors, but specifically what we're seeing here is evidence that is suggesting that the use of this agent on top of an ACE inhibitor is essentially not impacting overall survival.

Now, there's nothing in these data that would argue that if you added the beta blocker on top of this ACE inhibitor that it also wouldn't impact overall survival, and the bottom line point that I'm making is that the evidence that's more favorable here for the effects of valsartan are in individuals who aren't as heavily exposed to the beta blocker or the ACE inhibitor.

But that doesn't tell us anything about whether if we did a randomized head-to-head trial of valsartan against those other agents that we would expect comparable results.

ACTING CHAIRMAN BORER: Well, okay. Ray?

DR. LIPICKY: Were you going to? Go ahead. Finish your thought.

ACTING CHAIRMAN BORER: No, I was going to suggest I know that Ray and Alan both have comments here, but I think we've just discussed a number of these secondary points that you've made. I was going to begin to ask if there were any other -- any change in position or change in vote because if there isn't, I think we've answered the questions, but there are other comments here.

I mean, Alan, did you have something you wanted to --

DR. HIRSCH: I had a long comment, but I think ultimately it comes down to one sentence. There was biologic efficacy that can benefit patients, but labeling is critical.

ACTING CHAIRMAN BORER: Okay. Ray.

DR. LIPICKY: But, see, we're fine. I think you have answered all of the questions in that we know where things are, and we're about as equally divided in the division as you guys were in your conclusions. So that's fine. We understand that.

But there are two things I wanted to say before we quit. One is that if the task is hard, that doesn't mean the standards for whether or not you found something have to be relaxed. I think those two things have to be disconnected and also part and parcel of the same thing, and it's not clear they were; and also part and parcel of the same thing is that a study may, indeed, find a treatment effect. That doesn't mean it has to be approved. Okay?

The level of evidence, how well you believe that the trial results as a whole are applicable to a general population are, indeed, something that is critical, and so something may well say it very well looks like you have a treatment effect, and as a single trial, I'll buy you do.

That's just not good enough, and I don't think you thought that through well enough, but that's okay. All right?

ACTING CHAIRMAN BORER: Can I --

DR. LIPICKY: And then the last thing along those same lines was Jay's comment, and you know, I can't remember the valsartan congestive heart failure discussions, and I never bothered looking up the minutes, but you're probably right that the mistake that was made that you cited was made was stupid, wasn't it?

You know, I think that that was a bad bargain. I think if that's the advice you got, we gave bad advice because to equate the business of exercise tolerance and morbid mortal, and to consider them to be equal with respect to coming up with two positive trials is a stupid bargain retrospectively and over the course of years.

But I know we have done that. Okay? I'm not denying that. I just say that's very bad advice. It got you into the pickle you're in, and I'm sorry.

Pardon? Well, I understand, but we -- you know, I'll just acknowledge if that advice was what we gave and the program was developed on that basis, that's partly our fault.

DR. FLEMING: But, Ray, it's not entirely clear to me why you're as apologetic as you are. Let me see if I understand.

Basically what you're acknowledging is that you have a study here with a very large sample size and duration of follow-up to tell us something extremely important about morbidity and mortality primary endpoints and about secondary measures.

DR. LIPICKY: Right.

DR. FLEMING: And you were looking for, in a sense, some independent, confirmatory evidence, and you chose exercise tolerance.

DR. LIPICKY: Well, that --

DR. FLEMING: Let me go on.

And you're apologizing for having identified exercise tolerance as in any way a relevant supportive measure that should be weighed in this decision.

And yet the question that I would be uncertain about is at the same time what we're saying today is but some secondary measure, such as the Minnesota living with heart failure measure, dyspnea, and fatigue were, in essence, being asked do those things, in fact, elevate this to the same strength of evidence as two positive trials, and I just wonder a little bit in retrospect.

Those were positive and exercise tolerance was negative. Would we be having this discussion if --

DR. LIPICKY: Well --

DR. FLEMING: -- exercise tolerance was positive and those were negative.

DR. LIPICKY: I hear you, except, you know, if it had been a positive trial, there would probably have been -- if it had an effect on exercise tolerance, it would probably have been a somewhat difference discussion, but I guess what I was really saying in shorthand was if a morbid mortal trial were being done, we should have argued as opposed to doing hemodynamics and neural humors and all that and an exercise tolerance trial, which delays things before you get the other trial started because those precede, cost money to do them; we should have argued do a real morbid-mortal trial. It will have another dose in there at least or double the power, and don't give me this p of .05 for a morbid-mortal trial because you get into trouble every time.

MR. MacNAB: I think the discussion we had --

ACTING CHAIRMAN BORER: Well, wait, wait. Let us finish here first.

MR. MacNAB: I'm sorry. I just want to --

ACTING CHAIRMAN BORER: No, it's --

DR. LIPICKY: It's all right. He can argue.

ACTING CHAIRMAN BORER: Well, let's --

MR. MacNAB: I don't want to argue.

ACTING CHAIRMAN BORER: No, he wants to support you.

Can I ask, Ray? I mean, we've come down four to four, and we've answered all of the subissues as best we can, and there's obviously some --

DR. LIPICKY: You're fine.

ACTING CHAIRMAN BORER: -- concern because we lack knowledge here and we lack information, but it seems to me we have some responsibility to provide a statement about what additional information we would expect.

DR. LIPICKY: No.

ACTING CHAIRMAN BORER: Do you want us to say anything about that?

DR. LIPICKY: No, I don't think you have that responsibility.

ACTING CHAIRMAN BORER: Okay.

DR. LIPICKY: I think the way I take what the discussion has said is that as a whole there's a divided bottom line, that on a whole there is a divided way of how you look at this and what you regard as being good stuff and what you regard as being bad stuff, and that that basically will give us in the division a reasonable amount of latitude with respect to what people will be able to say and what they send to Dr. Temple, and will give Dr. Temple all of the ability to exercise his judgment.

ACTING CHAIRMAN BORER: Okay. Okay.

DR. LIPICKY: So it's just fine. I mean, I think you did the right thing, and it's how it came out. It's a tough problem.

ACTING CHAIRMAN BORER: Any other comments?

MR. MacNAB: Really, just to this whole issue about what was agreed to because I want the record to be straight, if you really go back to some time in 1996 the discussion about what had to be done was about as complex as the discussion that we've had today because it talked about many things, not just two trials. Totality of data, mortality, other endpoints; so I think in fairness to everyone we shouldn't have an impression that there was some disagreement or a mistake or you gave us the wrong advice.

DR. LIPICKY: Somebody screwed up.

MR. MacNAB: I think if you go back and look at that, we talked a great deal about totality of data.

DR. NISSEN: I just wanted to say one more thing, Ray. The reason I made the comment about the high bar is that if they had treated these patients the way contemporary --

DR. LIPICKY: They should have changed the standard. That's the only thing.

DR. NISSEN: No, no, no. Just hear me out for a second. If you take a group of patients -- if they had taken a group of patients 50 percent of whom were on ACE inhibitors --

DR. LIPICKY: I understand, but that shouldn't allow you to accept $100,000 for a million dollar watch.

DR. NISSEN: All right.

DR. LIPICKY: Okay?

ACTING CHAIRMAN BORER: Okay, but we don't all agree it's only $100,000, but it doesn't matter.

We've given you the best that we can do, which is a resounding 50 to 50.

Are there any other comments from the committee? If not, we'll conclude the meeting.

(Whereupon, at 3:22 p.m., the Advisory Committee meeting was concluded.)