FOOD AND DRUG ADMINISTRATION

+ + + + +

CENTER FOR DRUG EVALUATION AND RESEARCH

+ + + + +

ENDOCRINOLOGIC AND METABOLIC DRUGS ADVISORY COMMITTEE

+ + + + +

Thursday, July 26, 2001

+ + + + +

The meeting came to order at 8:00 a.m. in the Versailles Room, Holiday Inn Bethesda, 8120 Wisconsin Ave., Bethesda, MD, Robert A. Kreisberg, M.D., Acting Chair, presiding.

PRESENT:

Robert A. Kreisberg, M.D., Acting Chair

Marie C. Gelato, M.D., Ph.D., Member

Deborah Grady, M.D., Ph.D., Member

Lynne L. Levitsky, M.D., Member

William V. Tamborlane, M.D., Member

Allan R. Sampson, Ph.D., Member

Kathleen Reedy, Executive Secretary

Jose F. Cara, M.D., Consultant

Wendy W. McBrair, R.N.,M.S.,C.H.E.S., Consumer Rep.

Rebecca W. Killion, Patient Representative

Robert F. Misbin, M.D., FDA

Dragos G. Roman, M.D., FDA

John Jenkins, FDA

David Hobberman, FDA

David D. Orloff, M.D., Director, Division of Metabolic and Endocrine Drug Products

 

 

 

 

A-G-E-N-D-A

Call to Order and Introductions: Robert A. Kreisberg, M.D., Acting Chair 3

Meeting Statement: Kathleen Reedy, Executive Secretary 5

Welcome and Introduction: David G. Orloff, M.D., Director, Division of Metabolic and Endocrine Drug Products 6

Amylin Pharmaceuticals, Inc. Presentation

Overview: Joann L. Data, M.D., Ph.D., Sr. VP

Regulatory Affairs and Quality Assurance 8

Unmet Medical Need: Kenneth Polonsky, M.D.,

Adolphus Busch Professor of Medicine, Chairman,

Department of Medicine, Washington University

School of Medicine 11

Pharmacology of Pramlintide: Andrew Young, M.D., Ph.D., VP, Research at Amylin 21

Clinical Program for Pramlintide: Orville Kolterman, M.D., Sr. VP Clinical Affairs at Amylin 29

Risk/Benefit/Summary: Alain Baron, VP Clinical Research at Amylin 78

FDA Presentation: Division of Metabolic and Endocrine Drug Products

Efficacy: Robert I. Misbin, M.D. 106

Safety: Dragos G. Roman, M.D. 129

David Hobberman 150

Open Public Hearing 172

Charge to the Committee: David G. Orloff, M.D., Director, Division of Metabolic and Endocrine Drug Products 204

Discussion and Questions 209

Summary and Review 310

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

8:08 a.m.

DR. KREISBERG: Good morning. This is the Endocrinologic and Metabolic Drugs Advisory Committee to discuss the new drug application Symlin submitted by Amylin Pharmaceuticals. I'm Bob Kreisberg and I am the Acting Chair of the meeting for today.

I would like to ask that each of the members of the advisory panel introduce themselves. Let me tell you that you have to click on the microphone and talk directly into it. The red light will come on and when you're done, I would appreciate it if you would turn it off. Do not speak tangentially into the mic because it will not record the proceedings accurately.

Let me start all the way down on my right.

MR. JENKINS: Good morning. I'm John Jenkins. I'm the Director of the Office of Drug Evaluation II and the Center for Drug Evaluation Research.

DR. ORLOFF: Good morning. I'm Dr. David Orloff, Director of the Division of Metabolic and Endocrine Drug Products, Center for Drug Evaluation and Research.

DR. MISBIN: Robert Misbin, Medical Officer.

DR. ROMAN: Dragos Roman, Medical Officer.

DR. SAMPSON: Allan Sampson, Professor of Statistics, University of Pittsburgh.

DR. GELATO: Marie Gelato, Professor of Medicine at SUNY Stonybrook.

MS. REEDY: Kathleen Reedy, Executive Secretary of this committee, Food and Drug Administration.

DR. TAMBORLANE: Bill Tamborlane, Chief of Pediatric Endocrinology and Director of the Pediatric Pharmacology Research Unit at Yale.

DR. LEVITSKY: Lynne Levitsky, Chief of the Pediatric Endocrine Unit at Mass General.

DR. CARA: I'm Jose Cara, Section Head of Pediatric and Endocrinology and Diabetes at Henry Ford Health Systems in Detroit.

MS. McBRAIR: Wendy McBrair, Director of the Arthritis and Osteoporosis Center at Virtual Health.

DR. GRADY: Hi. I'm Deborah Grady. I'm a Professor of Medicine and Epidemiology at the University of California and San Francisco.

DR. KREISBERG: At this point I would like to introduce Dr. Kathleen Reedy, the Executive Secretary, who will read the meeting statement.

MS. REEDY: For the Endocrinologic and Metabolic Drugs Advisory Committee on July 26, 2001, the following announcement addresses the issue of conflict of interest with regard to this meeting and is made a part of the record to preclude any the appearance of such at this meeting.

Based on the submitted agenda for the meeting and all financial interest reported by committee participants, it has been determined that all interest 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 when evaluated against the agenda with the following exceptions.

Dr. Mark Molitch and Dr. Thomas Aoki are excluded from participating in the discussions and vote concerning Symlin. 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.

DR. KREISBERG: Next on the agenda is Dr. David Orloff.

Welcome and Introduction for us, David.

DR. ORLOFF: Thank you and good morning again. I want to extend my own welcome to the committee and thank them in advance for this service to the agency and to the drug review and approval process.

As Dr. Kreisberg has said and we'll hear as the day proceeds, we are convened to discuss the information on the safety and efficacy of pramlintide, an analog of endogenous peptide of pancreatic origin as an adjunct to insulin therapy in the treatment of Type I and Type II diabetes mellitus.

I have a few remarks I would like to make by way of introduction. The advisory committee process is an important aspect of FDA's review and regulatory oversight function in its decision making with regard to new drugs affording an opportunity for us, the FDA, to hear from experts in the field, from members of the public, as well as from the sponsor on the subject application.

At the outset it should be understood by all that we the agency enter into this meeting without an established course of regulatory action. We are here to engage in a discussion between the committee and the FDA and the sponsor on the scientific merits of the investigations, clinical and otherwise, of this drug and of the ramifications of the resulting data for a decision regarding marketing of the product for the proposed indications.

Let me remind everyone that the tone and outcomes of the deliberations today and the opinions expressed by the committee, as well as those expressed by the presenters for FDA do not represent final agency stance on the application.

Final regulatory action awaits further review, internal discussion, and potentially further discussions with the sponsor and, in this case, will not come for several months likely.

Again, I thank the committee for being here and I'll have further remarks in my charge to the committee early this afternoon following the FDA and the company presentations and prior to the committee discussion in response to our questions.

With that, I'll turn it back over to Dr. Kreisberg. Thank you.

DR. KREISBERG: Next on the agenda is the presentation by Amylin Pharmaceuticals. I would like to turn it over to Dr. Data who will introduce her associates in sequence. This is scheduled to go approximately an hour and a half and I would like to ask the panel to restrict any questions to clarification and not for discussion. I think there should be time at the end of the presentation to further question the presentation.

DR. DATA: Thank you, Dr. Kreisberg.

Good morning, ladies and gentlemen. I am Joann Data and I head Regulatory Affairs and Quality Assurance at Amylin Pharmaceuticals.

On behalf of Amylin I would like to particularly thank the Endocrine and Metabolism Division for their guidance through the drug development process of pramlintide and for their care and rapid review of our application.

Symlin, or pramlintide acetate, as you have heard, is the synthetic analog of the peptide hormone amylin and, as such, can be used as replacement therapy for amylin, the hormone that is deficient in insulin using patients.

Symlin is an injectable product. Our indication, or proposed indication, is as adjunctive therapy to insulin to improve glycemic and metabolic control in Type 1 and Type 2 diabetes. It is to be administered subcutaneously approximately 15 minutes prior to a meal. It will be made available in both vials and cartridges.

Our presentation today includes a presentation by Dr. Kenneth Polonsky. Dr. Polonsky is Chair of the Department of Medicine at Washington University. He will set the stage for the unmet need of insulin using patients.

Following that Dr. Andrew Young of our Research Division will present pramlintide's pharmacology. The clinical program will be provided by Dr. Orville Kolterman, Senior VP for Clinical Affairs. The Risk/Benefit/Summary will be prepared and presented by Dr. Alain Baron. Dr. Polonsky, Dr. Kolterman, and Dr. Baron are diabetologists.

In addition to our presentation today, this afternoon during the question and answer period, in addition to Amylin employees answering questions, we have brought along the following consultants to provide some additional support to the questions that you might have.

Those people are Dr. Huge Black representing toxicology; Dr. Wayne Colburn providing some additional input on pharmacokinetics; Dr. Richard Dickey to provide the clinical perspective from a general practitioner's point of view; Dr. Kerry Hafner talking about the statistics; and Kenneth Polonsky again will be joining us on the platform.

This morning we look forward to presenting our clinical findings and to show to you what we feel that we have that we've demonstrated safety and efficacy for pramlintide for the use in insulin requiring patients with diabetes. There is a role for Symlin. It has a role in terms of both glycemic and metabolic control.

Before moving on to the company's presentation, I would like to introduce Dr. Kenneth Polonsky.

DR. POLONSKY: Thank you. My goal this morning is to give a brief overview of where I feel we are in the therapy of Type 1 and Type 2 diabetes and what the unmet needs are in terms of current therapy and the opportunities for advancement.

The basic message from my perspective is that although there have been significant advances in therapy in recent years, unfortunately we still are unable to achieve goals of therapy in the vast majority of patients.

I think we all know that in Type 1 diabetes the mainstay therapy still remains insulin. Although insulin was discovered in the 1920s and although there have been a number of advances in novel preparations of insulin that have different pharmacokinetics and pharmacodynamics this is still really the only therapy that is available for reducing blood glucose concentrations in Type 1 diabetes.

In Type 2 diabetes, obviously there are a number of oral hypoglycemic agents that are sometimes used alone but frequently together with insulin therapy.

As I mentioned, in Type 1 diabetes insulin treatment is the mainstay of therapy and essentially all patients with Type 1 still have to be treated with insulin.

In Type 2 diabetes where there is a gradual destruction of the B-cell frequently we have to resort to insulin in the end stages to control blood glucose concentrations.

There are a number of important points that have become evidence as a result of research, particularly in the last 10 years. The first is that it's extremely important to achieve tight glucose control.

These are results from two very landmark studies in the treatment of Type 1 diabetes on the left, the DCCAT, and Type 2 diabetes on the right, the UKPDS study. What they both show that as you lower HbA1c an acceptable measure of glucose control from high levels shown here. This is the vertical access HbA1c and this is risk of retinopathy. What you can see is as you lower HbA1c from a high-level down to a lower level, there is a substantial reduction in the risk of retinopathy.

As you can see, the exact reduction of risk depends on where you are starting in the curve, but reductions for even relatively small reductions in HbA1c of .5 percent may lead to reductions in risk of retinopathy of 10 to 20 or even 30 percent. The same is true in Type 2 diabetes from the UKPDS.

The other important point is that we currently believe there is no threshold effect and that even when you get down to low levels of HbA1c additional improvements in glucose control have additional beneficial effects on reducing complications.

The other message that has come out of these studies is that it's not easy to achieve this level of control that is needed to really lower complications. You can see from the DCCT on the left again and in the follow-up study, the EDIC study, the HbA1c was 8.8 at the onset of the study and came down very significantly and then over time tends to go up again. The same is true in the UKPDS study of Type 2 diabetes where the initial benefits in control tend to wear off with time. Long-term maintenance of sustained good glucose control is really quite difficult. This has been shown in a number of studies.

This is a cross-sectional study from the University of Wisconsin that followed patients perspectively over time for about 10 years. You can see that even when patients were taking two or three or more injections of insulin shown over here in the blue bars and the green bars, and even when they were using combinations of insulin with multiple short-acting forms of insulin in addition to long-acting, you can see that the HbA1c is really disappointingly high.

The American Diabetes Association has defined that we would really like to achieve levels of around 7. The normal range is around 6. You can see that the average values in this study, which I think is fairly representative of the literature, shows levels of around 9 percent or so.

Additional studies, these are recent ones as you can see that were just presented at the American Diabetes Association meetings and some other more recent studies show that frequently or usually average HbA1c's in the diabetic population are in the 9 percent range approximately. Again, the ADA recommended target is down here at 7 percent, the upper limit of normal being around 6 percent.

Why do we have this difficulty? Well, there are a number of reasons but among them are hypoglycemia, weight gain, and the difficulty of controlling the elevation and glucose that occurs after meal ingestion or postprandial hyperglycemia. I'll briefly discuss these in the next couple of minutes.

It's been well demonstrated that as you take patients and you treat them aggressively to try and achieve the sort of glucose control that we've outlined, there is a significant increase in the risk of hypoglycemia.

This just shows in patients treated conventionally or with intensified regimens, as you attempt to lower the HbA1c there is a substantial increase in risk of hypoglycemia, particularly in the intensively treated patients but even in those under conventional therapy.

This occurs in patients with Type 1 diabetes that was on the previous slide, but also in Type 2 diabetes. This is data from the UKPDS study. You can see a significant increase in hypoglycemia in intensively treated patients.

Even in Type 2 diabetes when we add oral hypoglycemic agents as an adjunct to insulin with increasing dose of these agents there is a significant increase in the risk of hypoglycemia.

In addition to that, we also know that as we treat patients more aggressively with insulin, they tend to gain weight. These are data from the DCCT which just shows that if you divide patients up into the quartile of weight gain, at each level of weight gain the patient is treated intensively shown here in yellow gain more weight than the patients on the conventional regimen.

In the upper quartile you can see the gain in weight is really quite significant. Very substantial increases in weight gain. The same principles apply in Type 2 diabetes, although the absolute amount overall in the study was not as much as in the highest quartile in the DCCT.

Why does this matter? Well, obviously weight gain per se is not a desirable outcome, but in addition to that there has been an analysis of these data to demonstrate that the patients who gain the most weight also have a disturbing increase in cardiovascular risk factors.

This just demonstrates that if you take the same patients that I showed you on the previous slide looking at the quartiles of weight gain and you look at the effects on systolic blood pressure, diastolic blood pressure, triglycerides, and total cholesterol. As you can see, as people go into the higher quartiles of weight gain, there are increases in these parameters of blood pressure as well as triglycerides and cholesterol which are obvious risk factors for cardiovascular disease.

The final point that I'd like to just briefly touch on is this issue of postprandial hyperglycemia.

These are data from a study that we did a number of years ago which demonstrate that in Type 2 diabetes if you study patients over a 24-hour period and you give them the three normal meals, breakfast, lunch, and dinner, and you measure the blood sugars after meals, the predominate increase in plasma glucose that occurs over the 24-hour period is a postprandial one.

Available therapies have had a lot of difficulty in really controlling or normalizing postprandial hyperglycemia. This has been a particular challenge.

It occurs in Type 2 diabetes as well. This is just a study from a single patient who actually happens to be an endocrinologist who is treated with a pump who is overall in pretty good control. HbA1c of 7.1. This is a glucose sensor measurement of a plasma glucose profile over a 24-hour period.

What you can see is this patient is treated with multiple -- he's actually on a pump and he's treated with multiple bonuses of insulin in addition to the constant insulin infusion shown here in the blue triangles. There are frequent finger stick measurements of glucose, as you can see.

Despite these pretty intense efforts of trying to control the plasma glucose concentration, what you can see is that when you look very carefully over the 24-hour period at what the profile looks like, it's not completely normal. It's not what we would we would like it to be.

There are times when the blood sugar is higher than ideal and there are times when it is sort of bordering on the low limit of normal. In fact, slightly below where we would like it to be. Although this is a single patient, this is a fairly common experience of both patients and physicians who take care of patients with Type 1 diabetes.

There is a real unmet need and an opportunity to develop therapies that will address these issues. Therapies that might control better postprandial hypoglycemia, hypoglycemia that would achieve the levels of glucose control that we've outlined without excessive weight gain and obviously without increasing hypoglycemia which is a very important risk factor and a very important danger.

Just a couple of words about the hypoglycemia. Obviously, particularly in Type 1 diabetes, the patients have to still be maintained on insulin which is always going to have a risk of causing hypoglycemia.

Any combination of an oral agent together, or an injectable agent together with insulin if one is to avoid hypoglycemia, is going to require additional adjustments to the insulin therapy either perspective reductions in the insulin dose, frequent home glucose monitoring, and other maneuvers to try and reduce the risk of hypoglycemia.

I think I'll stop there and thank you for your attention.

DR. YOUNG: Thank you, Dr. Polonsky. My name is Andrew Young. I'm Vice President of Research at Amylin Pharmaceuticals. My purpose here today is to present to you the pharmacologic rationale for amylin replacement therapy using pramlintide.

Doing that, I'm going to compare the amylin and pramlintide molecules. I'm going to describe abnormalities of amylin secretion. I'm going to describe the role that amylin and pramlintide play in glucose homeostasis. In particular, I'm going to describe the glucose dependence of those effects, that they only occur when plasma glucose is normal or elevated.

I'll preface this talk by the remark that our knowledge of amylin biology is based upon seventeen hundred scientific communications including abstracts, papers, and reviews.

Amylin is a neuroendocrine hormone. It is a 37-amino acid peptide. It is located almost entirely within the pancreatic B-cells where it is co-secreted with insulin and from which it is do-secreted with insulin in response to meals as shown in this panel here.

We can see in the orange that the amylin profile follows the insulin secretory profile and they are, in fact, secreted in a comparatively fixed molar ratio.

Recently the receptor for amylin has been characterized and is present in the central nervous system. In particular, I want to draw your attention to the area postrema which is a site of tense amylin receptors. This part of the brain is implicated in glucoregulatory gut reflexes. It has no blood brain barrier and is accessible to circulating peptide.

Type 1 and insulin-using Type 2 diabetes is characterized not only by insulin deficiency but also by amylin deficiency as exemplified in this graph on the right-hand side. Type 1 diabetic patients have an almost absolute deficiency of circulating amylin and Type 2 diabetic patients, although amylin is present, it does not follow the normal secretory profile.

I will also point out that for the last 30 years diabetes has also been recognized as a condition characterized by excessive glucagon secretion.

Pramlintide is an analog of the human hormone amylin. When one wishes to replace a human hormone that is absent, the initial intent may be to replace it with the human hormone that is missing.

In the case of human amylin, this is not pharmaceutically practical because the molecule aggregated and was insoluble and unstable.

Instead, by the substitution of prolines at these three positions in the human amylin molecule, we produced pramlintide which is non-aggregating, soluble, and stable. In addition, has the full spectrum of activity that amylin has. It is equally potent to it and it has similar pharmacokinetics.

There are three fluxes that control plasma glucose. The efflux of glucose from the plasma is primarily under the control of insulin. There are two influxes of glucose into the plasma. One from endogenous sources, principally the liver, and that is under the control of the hormone glucagon. Also uptake from the gut.

Amylin by directing affecting nutrient assimilation from the gut and also indirectly by affecting glucagon secretion is able to modulate plasma glucose, particularly in the context of a meal.

This is exemplified in this slide where pramlintide administers prior to the ingestion of a test meal in Type 1 diabetic human subjects is shown to have a dose-dependent effect to flatten postprandial glucose excursions.

In describing the glucoregulatory actions of amylin I'm going to exemplify this by describing those effects which are well characterized and well understood in man as well as in animals.

I'm going to start with amylin's effects on the glucagon secretion. The effect of pramlintide to suppress postprandial glucagon secretion is exemplified here in Type 1 and Type 2 diabetic subjects where we can see that in the absence of pramlintide the postprandial secretion of glucagon is excessive.

Here a prior injection of pramlintide 30 micrograms has essentially normalized postprandial glucagon. The same effect is observed in Type 2 diabetic subjects where the excessive postprandial glucagon secretion has been normalized by a pramlintide infusion.

These animal data show an additional aspect of the glucagon suppression effect. First of all, in this part of the experiment where plasma glucose is clamped to glycemic levels, we can see that the co-infusion of amylin, which occurs throughout the experiment, has reduced nutrient stimulated glucagon secretion.

However, at this part of the experiment where glucose infusion was turned off the animal became hypoglycemic, we can see that the hypoglycemia induced glucagon secretion has been unaffected by the presence of high levels of amylin. That is, amylin has selectively inhibited nutrient stimulated glucagon secretion.

In support of amylin's absence of effect during hypoglycemia our some clinical data which show that pramlintide does not suppress the secretion of the counter-regulatory hormones, glucagon, growth hormone, cortisol, epinephrine, and norepinephrine.

Further, pramlintide does not impede the glucose response to exogenous glucagon injection.

I will now turn to the effect of amylin to modulate nutrient uptake from the GI tract. I will focus on gastric emptying

Amylin and pramlintide potently slow gastric emptying as amplified in this study of human Type 1 diabetic subjects. Here gastric emptying has been measured scintigraphically and is expressed as the time to half empty the stomach.

We can see that compared to the occasion when placebo was administered, that the administration of pramlintide has slowed gastric emptying by approximately one hour. Again, we ask the question what is the effect of hypoglycemia on this response. This is an animal study and here gastric emptying has been assessed by measuring the gastric contents 20 minutes after gavage. In a normal rate about half of the gastric contents are present 20 minutes later.

Amylin in this case has almost entirely ceased the emptying of the stomach such that 100 percent of the contents are present 20 minutes later.

However, in the presence of insulin-induced hypoglycemia shown as a progression to the right, we can see that there is an over-ride of the effect of amylin to slow gastric emptying. That is hypoglycemia has opened the emptying of the stomach.

The clinical implication of this animal experiment is that the presence of amylin or pramlintide will not inhibit the oral rescue from hypoglycemia.

How does this occur? We believe we understand why this hypoglycemic over-ride occurs. You will recall that I pointed to the area postrema, the part of the brain where we believe amylin has its effects to modulate gastric emptying.

This is a recording of a neutron from that part of the brain. We can see from this spike that when amylin is applied to that neuron that it is amylin sensitive. But we also see that when the glucose concentration surrounding that neuron is changed, we get an inhibition of its activity.

It would be tempting to suggest that the property of hypoglycemic override resides within the properties of these neurons that control gastric emptying.

To summarize, amylin and pramlintide exert their glucoregulatory actions via two broad mechanisms. Firstly, the inhibition of nutrient stimulated glucagon secretion. Secondly, the regulation of nutrient uptake from the gut. Importantly, both of these mechanisms are over-ridden during hypoglycemia.

It is clear that amylin is meant to be there. Pramlintide replaces amylin that is absent in those subjects who do not possess amylin.

Pramlintide restores control of glucose influx and in this way it complements the action of insulin which controls the efflux of glucose from the plasma.

At this stage, I would like to pass over to Dr. Kolterman who will present the clinical data.

DR. KOLTERMAN: Thank you, Dr. Kreisberg and members of the committee. My name is Orville Kolterman. I service as Senior Vice President of Clinical Affairs at Amylin Pharmaceuticals.

This morning I come before you to provide an overview of the clinical efficacy and safety data supporting the regulatory approval of pramlintide acetate.

Pramlintide is indicated as adjunctive therapy to insulin to improve glycemic and metabolic control in people with Type 1 or Type 2 diabetes.

Pramlintide is intended for patients with significant B-cell dysfunction. Patients with Type 1 diabetes enter the zone of B-cell dysfunction rapidly due to the autoimmune nature of the disease which destroys B-cells. They arrive in this zone soon as the diagnosis of Type 1 diabetes.

On the other hand, patients with Type 2 diabetes follow a slower more progressive course to arrive at the area where they rely upon exogenous insulin injections to allow them to achieve metabolic control.

It is this collection of patients, those that have severe impairment of the B-cell function and are, therefore, deficient in both endogenous insulin and amylin that pramlintide is intended.

These patients essentially have no other therapeutic options. Since the advent of insulin therapy, patients with Type 1 diabetes have had no other choices.

Patients with Type 2 diabetes who have progressed to this region have by in large extracted the therapeutic benefits available to them of the other therapeutic agents available.

As a principal investigator in the diabetes control and complications trial, I learned first hand how difficult it is for patients to achieve the desired level of metabolic control relying upon insulin alone.

In a center which was successful in the diabetes controlling complications trial, it took the significant part of a physician's time, the full-time efforts of a trial coordinator research nurse, part-time dietician, and part of a mental health care professional's time to achieve the target level of glucose control in only 30 to 40 patients pursuing intensive control.

It was that experience that led me to appreciate a paradox in my professional career. Namely, as a clinical endocrinologist I was trained that when we see hormone deficiency, that replacing the deficient hormone restores reasonable -- not perfect but reasonably normal physiology.

Therefore, I was struck by why if insulin was the answer that giving insulin back to these patients, highly motivated intelligent patients, did not do better and easier in terms of restoring normal physiology and became attracted to the possibility that there was something else that was missing.

When I learned about amylin it seemed that the fact that another peptide that came from the B-cell hormone was also deficient could perhaps be involved. That is what leads me to be here today to review the data with you.

The presentation will provide a program overview followed by a pharmacodynamic review. I will then turn to review of the safety and efficacy data for first Type 2 diabetes, and subsequently patients with Type 1 diabetes.

As I give this presentation, I will draw from an extensive database where 4,493 patients with Type 1 and Type 2 diabetes have been exposed to pramlintide. The total duration of pramlintide exposure is 2,727 years. We feel unequivocally that this database is robust and satisfactory for regulatory decision making.

As I progress through the presentation, approximately $1,300 with Type 2 diabetes, approximately $1,200 patients with Type 1 diabetes who participated in the long-term control trials will serve as the basis for many of the points that I make.

The 2,727 of patient exposure is composed as follows. There were 2,109 patients exposed for six months or longer. There were 1,350 patients exposed for one year or longer. And there were 261 patients who are exposed for an excess of two years.

The population demographics of the patients participating in pramlintide clinical program are representative of those of the intended populations. If we look at Type 1 diabetes the mean age of 40 years, the duration of the diabetes of 17 years is representative of what we see in the clinic when we treat patients.

Baseline HbA1c in both Type 1 and Type 2 diabetes is approximately nine years -- I'm sorry, 9 percent. The duration of the diabetes in patients with Type 2 diabetes that participated was approximately 12 years.

With this duration of diabetes, these patients had a representative presence of the various comorbidities which we see in patients with diabetes and, therefore, were exposed to the relevant concomitant medications.

You can see 56 percent of patients with Type 1 diabetes used concomitant medications. And only over 80 percent of patients with Type 2 diabetes use concomitant medications --

(Whereupon, off the record.)

DR. KOLTERMAN: -- are well represented. Of particular interest to our considerations today are patients with Type 2 diabetes where over 20 percent of patients were using an oral hypoglycemia agent in addition to the insulin therapy that they employ.

This use of oral hypoglycemic agents consisted primarily of the biguanide used by approximately 12 percent of patients and various sulfonylureas used by 13 percent of patients.

The data that I will review shows that pramlintide is adjunctive therapy to insulin results in further improvement in glycemic control above that seen with insulin alone. This improvement in glucose control comes primarily from a reduction in postprandial hypoglycemia which with chronic therapy translates into a reduction in HbA1c.

This improvement in glucose control, as you will see, is achieved without an increase in insulin use. Unlike the majority of situations where we improve glucose control in patients with diabetes, the improvement in glucose control with pramlintide is accompanied with weight loss as opposed to weight gain.

Let us know turn to the pharmacodynamic review. To begin that review, look at the pharmacokinetic profiles in patients with Type 1 diabetes shown in the left panel, Type 2 diabetes shown on the right panel.

You can see that there is a dose dependent increase in C-max with increasing doses of pramlintide both in Type 1 and Type 2 patients.

You will also notice that in both patient types that circulating plasma pramlintide concentrations are essentially gone from the circulation by the end of three hours consistent with a short duration of action.

I also point out to you in this slide that for a given dose of pramlintide administered subcutaneously, that the plasma concentration achieved in patients with Type 2 diabetes are somewhat lower than those achieved in patients with Type 1 diabetes.

The addition of pramlintide to regular insulin results in significant improvement in postprandial glucose concentrations. These patients with Type 1 diabetes were studied on two separate days under identical conditions. The same dose of insulin administered preprandial prior to the morning meal. And the same meal, Sustacal meal challenge.

You can see from the data of the patients that received pramlintide, there's a significant reduction in postprandial glucose concentrations.

I call your attention to the plasma concentration profile engendered by this 30 microgram dose of pramlintide. These plasma concentrations are similar to circulating amylin concentrations in healthy individuals following the ingestion of a similar meal.

Thus, it follows that the addition back of this amylin affect which is there in normal individuals, the addition of this effect back in patients with Type 1 diabetes where it had become deficient accounts for the reduction in postprandial glucose concentrations.

If one looks at another study in patients with Type 1 diabetes who were treated for 28 days with Placebo in a randomized manner and then washed out for six weeks and crossed over for treatment with the other agent.

The study ended on the 28th day of therapy with either placebo or pramlintide and had the same meals administered for breakfast and for the mid-day meal and employed the same dose of intermediate and short-acting insulin.

You can see that the addition of pramlintide resulted in the same reduction in postprandial hypoglycemia after both the morning and the mid-day meal. The administration preprandially is important in terms of getting the effect of the drug.

I also call to your attention here the fact that 28 days of treatment with pramlintide did not significantly impact the fasting glucose concentration consistent with pramlintide being a postprandial drug.

Studies of this sort demonstrate a dose response relationship for pramlintide and served as guidance for selection of dose for study in long-term trials.

Plotted here a change in plasma glucose concentrations following the administration of a standardized Sustecal meal challenge. You can see that there is a dose dependent decrease in the increase in postprandial glucose concentrations when we go from 10 micrograms to 300 micrograms.

Shown here is the mean incremental area under the glucose curve from those studies plotted as a function of dose. You can see there is a nice dose response relationship and when an appropriate statistical test is applied, one see statistical significance.

If one superimposes on this, the incidents of the most frequently encountered side effect from pramlintide, namely nausea, one sees nausea increasing in a dose-dependent manner as well. This data defined the range between 30 and 100 micrograms as an appropriate dose range for exploration in long-term studies.

In a similar way in patients with Type 2 diabetes, treatment with pramlintide of doses between 30 and 150 micrograms yielded evidence of a dose response relationship based on evaluation of HbA1c after 13 weeks of treatment.

If one superimposes a side-effect profile here, you again see a dose-dependent increase. I call to your attention that the incidence of the nausea side effect is significantly lower in patients with Type 2 diabetes.

These data allowed us to define the dose range of 30 to 150 micrograms as appropriate for exploration in patients with Type 2 diabetes. Thus, the long-term studies in Type 2 diabetes examine doses of 30 to 150 micrograms and those in Type 1 diabetes examine doses of 30 to 90 micrograms.

In terms of the Phase 3 trials, they served three purposes, need to demonstrate efficacy, assess safety, and provide a basis for some guidance for clinical use.

Study design considerations that were taken into account in the area of 1995 to 1996 when this program was designed and put in place included the following. At that point in time there was no precedent at all for the evaluation of another drug in insulin-treated patients.

In fact, in patients with Type 1 diabetes there has been no successful evaluation of another drug to lower glucose concentrations. At that point in time the results of the diabetes control and complications trial had just become available validating HbA1c as a surrogate endpoint for glycemic control.

There remained, however, an ongoing debate regarding whether or not there was a threshold effect for HbA1c; namely, was there a degree of glucose lowering that was required in order to see definite benefit. Today, as Dr. Polonsky has indicated, there is a consensus that the answer to that question is no, there is no threshold effect. That was not known in 1995 to 1996.

Also, at that point in time the value of the ancillary metabolic effects, such as changes in lipid profiles, changes in body weight, changes in insulin use, were not as fully appreciated as having value as they are today.

The following approaches were taken to the program. All subjects needed to be treated with insulin because that was the target population. Therefore, all studies employed an add-on design meaning that either pramlintide or placebo was added to existing therapies.

Oral hypoglycemic agents if used by patients with Type 2 diabetes who came to the studies were allowed to be continued but the dose was required to remain constant throughout the participation. As I have shown you, this included sulfonylurea compounds and metformin.

Approaches to insulin management deserve special particular consideration because we have a tension here between clinical trial design and clinical practice. In the clinical trial setting insulin should ideally remain constant in order to isolate the magnitude of effect of the add-on drug. Changes in insulin doses during the study period confound data interpretation.

On the other hand, in the clinical practice setting, patients with diabetes change their insulin doses on a frequent basis. This is needed for considerations of patient safety in terms of limiting the risk for hypoglycemia and also to facilitate the patient's pursuit of glycemic targets. Within this program it is necessary to balance these competing demands.

Outlined here are the ways that we chose to do this after assessing the various clinical ramifications. In four of the long-term studies consistent insulin dosing was encouraged but not mandated.

Two studies placed no constraints whatsoever upon insulin dosing. Patients were always allowed to change insulin doses for patient safety purposes. To do otherwise would have been unethical. Very importantly, patients were not discontinued from the study if they changed insulin by more than what was desirable for study purposes.

To address the issue or to provide insight into the confounding nature of changes in insulin, the analysis plans predefined a stable insulin subgroup cohort defined by patients who from baseline to study completion did not change their total daily dose of insulin by more than plus or minus 10 percent. We feel that this isolates the true drug effect and allows it to be quantitated.

The general overview of the Phase 3 studies themselves, they were multicenter, randomized, placebo-controlled studies. The primary endpoint was change in HbA1c from baseline to either 26 or 52 weeks.

Secondary endpoints included changes in body weight, changes in insulin use and, of course, the important safety parameters.

Let us now turn to consideration of the efficacy data for patients with Type 2 diabetes. This data comes from three studies. First study, 137-111 employed a short placebo lead-in period after which time the patients were randomized to either placebo or three-dose regimens of pramlintide.

The body of the slide, the numbers and the arrows, indicate the times at which the endpoint assessments were made. I call to your attention that in this study, which was done earlier in the program, the formulation of pramlintide was at a high pH than that intended for market use.

With that formulation there is lower bioavailability such that 150 microgram dose here yields plasma concentrations similar to 120 micrograms in the subsequent studies.

The other two studies in Type 2 diabetes are identical in design with the exception that study 123 is six months in duration whereas study 122 is 123 months in duration or 52 weeks.

These studies encouraged investigators to identify patients that they felt were appropriate for participation in the clinical trials to make indicated changes in their management regimens. Having made those changes, to have the patients in a period of metabolic stability for two months prior to entering a one-month placebo lead-in period.

These measures were taken to have patients in appropriate states of control for entry into the study and to facilitate the acquiring of stable baseline HbA1c measurements. You can see that in each of the studies patients were randomized to placebo or one of three treatment regimens of pramlintide.

As we turn to evaluate the data, let me come up front and provide you an overview of what I'm going to show you. Shown here is a plot of HbA1c over time for patients completing 52 weeks of observation.

You can see that the HbA1c in the patients receiving insulin alone decreased somewhat put that there are significantly larger reductions in HbA1c in the pramlintide treated patients.

The bottom panels show the two significant side effects that we need to deal with when we consider pramlintide; namely, overall nausea and severe hypoglycemia.

Both of these show a slight increase in the first four weeks of therapy and with chronic therapy after four weeks one sees little difference between pramlintide treated patients and placebo patients.

We look at the HbA1c data from the individual studies. We'll begin with a summary showing you the data for the intent-to-treat analysis following 26 weeks of therapy. The data plotted is changed from baseline.

We begin with data from study 137-111 where data for the 75 and the 150 micrograms achieve statistical significance versus placebo in a predefined analysis.

The data from study 137-123 provides supporting data with the 120 microgram dose given twice a day achieving a nominal p-value. A similar pattern is seen in study 137-122 where the 120 microgram twice a day dose again achieved statistical significance by predefined analysis plan.

We have two studies that confirm in predefined analysis the superiority of pramlintide added to insulin versus insulin alone in terms of HbA1c. We have supportive data for the 120 microgram dose from study 137-117. These data serve as the basis for the recommendation for Type 2 diabetes of a dose of 120 micrograms administers two or three times a day.

We now draw from study 137-122 for data to look at in more detail to better understand the effects of pramlintide. The left panel plots the intent-to-treat analysis over 52 weeks for patients treated with insulin alone shown in tan, versus patients treated with pramlintide plus insulin, 120 micrograms a day of pramlintide, the orange line.

You can see that both groups evidence a reduction in HbA1c across the 52 weeks but that the reduction in pramlintide treated patients is significantly greater than that treated in patients receiving insulin alone. The results are statistically significant at both 26 and 52 weeks.

The right panel shows data from the stable insulin cohort. Again, this is the analysis that we feel best isolates the true pramlintide effect. You can see that both groups have a reduction in HbA1c across the 52-week period of the trial.

The pramlintide treated patients the reduction is similar in magnitude to that seen within the entire group consistent with the insulin resistance that is a component of the Type 2 disease. Therefore, modest changes in insulin do not have much of an impact in their response.

I call your attention to the fact that the isolated effect of pramlintide here is sustained in magnitude across the second six months of the study contrasted with the decay and glucose control observed in the patients treated with insulin alone.

To date we've looked at mean data which is required for statistical analysis, but is not particularly helpful in terms of understanding how individual patients are performed.

I would now like to draw curves for you that show the data for all patients treated with the recommended dose of pramlintide, 120 micrograms given twice a day contrasted with patients receiving insulin alone. We are going to plot the percent of patients achieving a given HbA1c response shown across the X axis here.

The white vertical line at zero divides the field into a left panel which represents improvement in glucose control, versus a right panel representing worsening of glucose control. If we look first at patients treated with insulin alone, they come across the spectrum as such.

Approximately 55 percent of patients have shown improvement in glucose control and 45 percent of patients show some worsening of glucose control.

We now add the pramlintide treated patients. The most important point on this slide is that the blind for the pramlintide treated patients across the spectrum of HbA1c responses lives to the left meaning that for any particular parameter or any particular target that you want to evaluate, pramlintide added to insulin is adding benefit.

To quantify that more specifically, this line allows us to see that approximately 70 percent of patients receiving pramlintide have some improvement in glucose control with only approximately 30 percent having some worsening of glucose control.

If we look at patients who achieved a .5 percent reduction or greater in HbA1c, you can see that is achieved in 35 percent of patients receiving insulin alone compared to 55 percent of patients receiving pramlintide.

If we look for patients who achieved a 1 percent or greater reduction in HbA1c, you see that that happened in 20 percent of patients receiving insulin alone, whereas it occurred in approximately 35 percent of patients treated with pramlintide plus insulin.

Thus, the addition of pramlintide to the patient's regimen has allowed one out of three patients to achieve a reduction in HbA1c of 1 percent or greater, a highly relevant beneficial clinical response.

Another way of looking at this is to look at the percent of patients achieving the ADA targets. If we look at patients achieving 8 percent or less, you can see that 21 percent treated with insulin alone, whereas a larger proportion, 35 percent, receiving insulin plus pramlintide.

If we look at patients achieving the desired target of 7 percent or less, it happened in only 2 percent of patients treated with insulin alone. This number is increased four fold in pramlintide treated patients, 8 percent of patients.

Pramlintide is unique by the data I'm going to show you in this slide in that in conjunction with this improvement in glucose control, instead of increasing body weight, pramlintide allows patients to decrease their body weight.

This is data from study 137-111. You see a reduction in body weight in all treatment arms. This is a highly reproducible effect seen across all treatment arms in all studies so the beneficial effects of reduction of HbA1c without weight gain.

This slide portrays for you the profile of pramlintide. This is looking at all patients treated with the recommended doses. We have approximately 300 patients receiving pramlintide represented here. Baseline HbA1c for patients receiving insulin alone was 9.3 percent compared to 9.1 percent for the pramlintide treated patients.

We are going to look at changes in HbA1c, change in insulin use, change in body weight. You can see on the left panel that there is a significantly greater reduction in HbA1c in the pramlintide treated patients.

This reduction in HbA1c is achieved without an increase in insulin use. Contrast to get this smaller reduction in HbA1c patients treated with insulin alone are progressively increasing their total daily dose of insulin.

This reduction in HbA1c is accompanied again by a reduction in body weight contrasted with the increase of body weight of patients receiving insulin alone. An improvement in HbA1c above and beyond that achieved with insulin alone without an increase in insulin use accompanied by a reduction in body weight.

Having summarized the efficacy data for patients with Type 2 diabetes treated with pramlintide, let us now turn to the safety review.

This total will be pulled from 1,512 patients receiving pramlintide and influenced, as I have pointed out previously, the 1,273 patients participating in the long-term control trials.

Given that pramlintide was a new chemical entity, the first in the class of a new set of therapeutic compounds for patients with diabetes, it was incumbent upon us to do a safe, careful rigorous safety evaluation. That was done at the individual study levels.

As the application to the agency was prepared, the data from the individual studies was rolled up into an integrated safety summary database where additional analyses were done across the whole database in a very systematic programmatic way.

In addition, serious adverse events were collected and reviewed in a contemporaneous manner throughout the conduct of the program. As the Chief Safety Officer for the company, I personally reviewed all serious adverse events including the severe hypoglycemic events.

There was no increase in mortality in the Type 2 population observed with pramlintide. There were 10 deaths which occurred in the approximately 2,200 patients. None of these were classified as drug related. The incidents of death in pramlintide treated patients was lower than that seen in placebo treated patients. The same is true for cardiac related mortality.

If we look at the event profile for treatment-emergent and adverse events defined as those events that had an overall incidence greater than 5 percent excluding hypoglycemia where the incidence was greater in patients receiving pramlintide, the terms listed here come to the table.

The one where there is the largest imbalance is nausea which occurred in 24 percent of patients treated with pramlintide. But importantly only 2 percent of the nausea complaints were categorized by the trial coordinators and the investigators as severe in intensity.

In study 137-111 there was an apparent increase in the appearance of retinal disorders in patients receiving the highest dose, 150 micrograms. Careful review of this at the study level indicated that this appeared to be related to a failure to properly document the presence of existing retinopathy at baseline.

That observation led to some improvements in data capture techniques for subsequent studies. In those studies where 120 micrograms which produced similar plasma concentrations to the 150 microgram dose were evaluated, there was no evidence of increase in retinopathy -- retinal disorders. I'm sorry.

Similarly, the 75 microgram three times a day dose in the same study showed no signal and there was no signal observed in any of the Type 1 studies. Therefore, we conclude that this is not a safety concern for patients treated with pramlintide.

Shown here is the profile of serious treatment-emergent adverse events. As you can see, the profile is similar in pramlintide treated patients compared to those with insulin alone.

Anytime one adds another glucose lowering agent to insulin regimes in patients treated with insulin, there is a risk for an increase in hypoglycemia.

Therefore, we put in place at the outset of the studies a tracking mechanism which employed a definition of severe hypoglycemia to provide an objective quantifiable assessment of this risk. We used a definition similar to the DCCT -- we used the DCCT definition for this purpose.

You can see that in the Type 2 population there is an increase in the incidence of severe events, but when one corrects -- when one evaluates the data in terms of annual event rates which corrects for the appearance of multiple events within the same individual and corrects for the duration of drug exposure, one sees that the event rates are nearly identical.

The events in pramlintide treated patients tend to occur a bit earlier in that there is some increase during the first four weeks of treatment, .5 versus .2. In the later months of follow up, there is no increase in the pramlintide treated patients and perhaps a decrease during the second six months of treatment.

In terms of other safety observations in patients with Type 2 diabetes, there is no evidence of serious events that are unusual in the absence of drug therapy. There is no evidence of cardiac, renal, or hepatic toxicity.

Equally important, there's no increased frequency of clinically significant changes in lipid profiles, electrocardiograms, diastolic or systolic blood pressure, or the various safety laboratory parameters monitored. This leads us to a conclusion that pramlintide is efficacious and safe in patients with Type 2 diabetes.

I've shown you an improvement in glycemic control without an increase in insulin use with weight loss as opposed to weight gain, no safety issues of concern, though there is a slight increase in severe hypoglycemia during the first four weeks of therapy which is manageable as we will talk about in greater detail in the Type 1 review. We have a dosage recommendation of 120 micrograms given two to three times per day before meals.

That concludes the presentation related to Type 2 diabetes. Let us know move on and review similar data for patients with Type 1 diabetes. Again, we begin with the efficacy review.

I will move through this a bit more quickly because the layout of the presentation is identical to that that you've seen for Type 2.

Again, there were three trials. The first trial had a short lead-in period followed by randomization to placebo or 30 micrograms four times a day. This trial was somewhat unique in that at week 20 based upon changes in HbA1c at week 13, patients who had not achieved a reduction of HbA1c of 1 percent or greater were rerandomized to either remain on 30 micrograms or escalate to 60 micrograms.

That imbedded rerandomization lead to no increase in reduction in HbA1c. Thus, for the purpose of the presentation, the patients are treated -- the pramlintide treated patients are handled as a single group.

The other two studies in Type 1 employed the study design similar to that that we described for the Type 2 patients. Here patients were randomized to either placebo or one of three treatment arms, one of three pramlintide treatment regimens. The endpoint assessments again were drawn at the same points as those for the Type 2 program.

Again, allow me to show you the profile before we actually look at the data. Again, you can see that there is a significantly greater reduction in HbA1c that is maintained across the entire 52-week period for patients treated for 12 months.

The two side effect issues of concern, overall nausea and severe hypoglycemia, both increased during the first four weeks of therapy, but between weeks four and 26 and 26 to 52 have rates for both nausea and severe hypoglycemia that are similar to those seen in patients receiving insulin alone.

This slide summarizes the intent-to-treat analysis for changes in HbA1c for baseline six months of treatment. Study 137-112, the 30/60 microgram QUID regimen achieved statistical significance and a predefined analysis.

In study 137-117 there is supporting data coming from 60 micrograms given three times a day compared to placebo. And in study 137-121 both 60 micrograms given both three times a day and four times a day achieved statistical significance in a predefined analysis plan.

Again, we have two studies by predefined analysis plan demonstrating statistical significance and supporting data from study 137-117.

The orange bars, again, highlight the recommended doses for going forward; namely, we propose initiating therapy with 30 micrograms or less given three to four times a day with maintenance therapy of 30 to 60 micrograms given three to four times per day.

Shown here is the intent-to-treat analysis for change in HbA1c for baseline over the entire 52 weeks for patients treated with insulin alone contrasted with pramlintide 60 micrograms three times a day or four times a day this data being drawn from study 137-121.

Again, you see a pattern where all arms have a reduction in HbA1c from baseline throughout the 52 weeks. The reduction in pramlintide treated patients is significantly larger, statistically significant at week 26 and at week 52 and is maintained between week 26 and week 52.

The right panel now becomes relevant because we are dealing with patients with Type 1 diabetes. This is the predefined stable insulin cohort where patients did not change their insulin by more than plus or minus 10 percent during the period of observation.

When one looks at the data, one sees that the patients receiving insulin alone decreased HbA1c initially but during the second six months of the study, there is a deterioration in glucose control.

In contrast, patients receiving pramlintide plus a stable dose of insulin have a significantly larger reduction in HbA1c that is well maintained over the 52 week period of treatment.

The difference between placebo and pramlintide treated patients is .7 percent, exactly what we predicted based upon the changes in the postprandial glucose profiles that I showed you at the outset of this presentation.

We feel that this isolates the pramlintide drug affect and allows one to quanitate the true HbA1c lower properties of the compound.

Again, we look at the response pattern seen in all patients receiving recommended doses. Again, patients receiving insulin alone are shown by the tan line. If we look at that group of patients, just over 55 percent have some reduction in HbA1c. Just under 45, 42, 43 percent show worsening of glucose control.

If we look at the recommended doses for pramlintide, the 30 and 60 microgram treatment arm in the various studies, one sees that over 70 percent of patients -- over 70 percent of patients show improvement in glucose control contrasted with only 25 percent showing a worsening of glucose control.

If we look at patients achieving a HbA1c reduction of .5 percent or greater, one sees that it happens in 25 percent of patients receiving insulin alone versus 45 percent of patients treated with pramlintide.

Look at 1 percent target. There's about 12 percent of patients receiving insulin alone versus 25 percent patients treated with pramlintide. In the Type 1 population the addition of pramlintide yields a 1 percent or greater reduction in HbA1c in approximately one out of four patients.

If we look at the ADA targets for those a achieving less than 8 percent, it's 28 percent for insulin alone, significantly larger number of patients than the pramlintide treated arms.

Those achieving the important ADA target of 7 percent or less occurs in 7 percent of patients receiving insulin alone is doubled in the pramlintide treated patients at 14 percent.

As we saw in the Type 2 population, pramlintide therapy improves glucose control without leading to an increase in body weight. You can see the same reproducible reduction in body weight across all treatment arms, all studies. Very similar and consistent with what I showed you in the Type 2 population.

We now look at the slide that shows the pramlintide profile. Again, the data from all patients treated with the recommended doses of 30 and 60 microgram in the Type 1 program, 716 patients receiving pramlintide.

The same three panels you saw before, change in HbA1c, change in insulin, change in body weight. Again, a significantly greater reduction in HbA1c compared to patients receiving insulin alone.

This larger reduction in HbA1c is achieved with minimal change in total daily insulin use contrasted with patients treated with placebo who are constantly increasing their insulin dose through the period of observation.

This change in HbA1c with this pattern of insulin use is accompanied very importantly by a reduction in body weight contrasted with the increase in body weight patients treated with insulin alone. Thus, pramlintide allows improvement in glucose control without an increase in insulin use and without an increase in body weight.

There is one other important cohort to look at in the Type 1 population; namely, patients who are approaching the optimal target for glycemic control, 7 percent. These patients represent the lower quartile of patients treated with pramlintide.

We took the entire cohort, rank ordered them based on baseline HbA1c and took the lower third of the patients for evaluation. These patients had entry HbA1c's of less than 8.3 percent. The mean for the cohort was approximately 7.7 percent.

In this group of patients who have mean HbA1c's below the target for intervention proposed by the American Diabetes Association, these patients treated with pramlintide show a beneficial effect compared to the patients receiving insulin alone.

There is a larger reduction in HbA1c that is maintained compared to the patients receiving insulin alone across the entire period of observation.

Look at insulin use and we see the same pattern that we've seen elsewhere. This improvement in HbA1c is coming without an increase in insulin use. Importantly for these patients, it is accompanied by a reduction in body weight.

Having presented the efficacy data for patients with Type 1 diabetes, let us now turn to the safety review for patients with Type 1 diabetes. Again, we will use a similar format to that in the Type 2 presentation. Here we are drawing on 1,970 patients, Type 1 diabetes treated with pramlintide. Of those, 1,179 participated in the long-term control trials.

When we look at mortality in the Type 1 patient population, again there is no increase in mortality. There were seven deaths which occurred among 3,477 unique subjects. Again, when we look at the overall incidence of death, there is no increase in pramlintide treated patients and the incidence of cardiac death is somewhat larger in pramlintide treated patients.

There was one death in the pramlintide treated patient that was classified as possibly drug related. This is a male patient approximately 44 years old who appears to have had a hypoglycemic seizure at approximately 4:00 a.m.

During the seizure he entered cardiac arrest and it was not possible to resuscitate the individual. At autopsy a 70 percent lesion was found in the right coronary artery and the cause of death as described by the coroner was coronary arterial sclerosis.

There was a second patient that I believe you have seen mentioned in the briefing materials from the agency, a patient that died in a motor vehicle accident during the first day of therapy.

That patient also underwent autopsy and at autopsy food was found in the stomach and cerebral spinal fluid collected at the time of autopsy had a glucose concentration that argues against that patient being hypoglycemic at the time of death.

We have prepared as backup slides the details of each of the deaths that occurred in the pramlintide program and would be happy to review any of those aspects with you that are of interest during the question and answer period.

Because, as you will see, there is an increase in hypoglycemia in pramlintide treated patients in the Type 1 cohort, concern was raised about the possible increased risk of motor vehicle accidents and injury associated with motor vehicle accidents.

The data for this are summarized on this slide for all motor vehicle accidents in the left panel and hypoglycemia related motor vehicle accidents in the right panel. The data are plotted as annual event rate for patient year of exposure.

Before we look at that data, I want to call to your attention that there are more events that occur in pramlintide treated patients than in placebo patients in both categories here, but there are also significantly more patients exposed to pramlintide than there are exposed to placebo.

Therefore, we felt that evaluating this as an annual event rate where we looked across the entire safety database looking at all such events that occurred was the most appropriate way to address this.

When we do that, we come with the annual event rates that are plotted here with the corresponding confidence intervals. You can see that on both sides the ledger here that the confidence intervals overlap and that the p-values do not approach statistical significance.

A similar evaluation was done of other accident and injury, so-called non-motor vehicle accidents, for all hypoglycemia shown on the left panel and all hypoglycemic related accidents and injuries on the right panel. Again, one comes to a similar conclusion. Rates are similar, confidence intervals overlap, and the p-values do not approach statistical significance.

Moving onto the treatment-emergent adverse event profile, it is similar to that that we saw in patients with Type 2 diabetes. However, the incidence of nausea is higher as I alluded back on the dose response slide, occurs in approximately 50 percent of patients with Type 1 diabetes treated with pramlintide. Only 7 percent of this is rated as severe in tensity by the investigators and the trial coordinators.

To look at this in more detail, 49 percent of the patients during 52 weeks of observation never register a nausea complaint. This nausea is nonspecific so events related to nausea associated with other medications, viral gastroenteritis, etc., are all captured here.

The nausea in the pramlintide treated patients, 44 percent is classified as mild to moderate. Only 7 percent is rated as severe. This nausea side effect is dose dependent, as you saw previously, increasing as one goes from 30 to 60 to 90 microgram doses.

This nausea is also transient in nature. It occurs quickly with the initiation of therapy if it occurs and it dissipates essentially during the first four weeks of therapy. If one looks beyond four weeks of therapy, the rates are similar between pramlintide and patients treated with insulin alone.

Serious treatment-emergent adverse profile again is similar between pramlintide and patients treated with insulin alone.

I call your attention that in the Type 1 population the metabolic and nutritional category body system has the most events having an incidence of 6 percent in placebo contrasted with 10 percent with pramlintide. This captures hypoglycemia and is a signal that there is an increase in hypoglycemia that needs to be evaluated.

As I indicated earlier in the presentation, we were diligent in this evaluation. It was established at the start of the program that this was something that needed to be looked at and we employed a procedure, a process, similar to that that I participated as a member of the safety monitoring board in a diabetes control and complications trial.

The objective endpoints employed were those of the DCCT looking at patients who required the assistance of another individual including aid in the adjustment of oral carbohydrate or requiring the administration of glucagon or the injection of intravenous glucose.

It was the sponsor's intent to have severe hypoglycemia reported as a serious adverse event to ensure the timely and complete collection of data.

When one looks at the entire data set, one sees that there is no overall increase in the annual event rate of severe hypoglycemia.

Let me remind you that I am using annual event rate data because it captures, it compensates, deals with, addresses the issue of multiple events in a single individual and also accounts for differences in exposure of subjects. The overall rate is identical. However, given the importance of this, it needs to be looked at in greater detail.

The first thing to call out is within the placebo treated patients. We have an individual that we refer to as our century man who reported in excess of 100 events. If that patient is excluded as an outlier, you have an annual event rate of .8 versus 1.1 percent in the pramlintide treated patients.

One looks at this over time and one sees that there is a clear -- there is a clear increase in the incidence of severe -- I'm sorry, in the annual event rate for severe hypoglycemia during the first four weeks of treatment.

As one progresses through time, this difference goes away. If one looks beyond four weeks, one does not see it. This is looking at all patients in the Type 1 indication who received pramlintide.

We focus on what is now defined as a critical period based on the data of zero to four weeks. There is a clear dose response relationship for the hypoglycemia risk. You can see as it is increased from 30 to 90 micrograms the risk rises.

Thirty micrograms is not too dissimilar from that seen in patients treated with insulin alone. This starts to serve as a basis for a recommendation as to how this is addressed.

This presents the risk for hypoglycemia as a hazard function as suggested by Dr. Robert O'Neill with the Biometrics Group of the FDA as a reasonable way of looking at serious adverse event data such as this.

You can see that when the hazard function is evaluated that during the first four to six weeks of therapy there is clearly an increased risk and the patients receiving pramlintide there is no difference. There is no difference in the risk as assessed by the hazard function.

I now call your attention to the rate in the placebo treated patients here, the risk of about .03. This slide is a similar analysis for pramlintide treated patients receiving 30 micrograms four times a day in study 137-112. There is a slight difference in the placebo data here because this is looking for a realistic or fair comparison at only the placebo patients in study 137-112 where the 30 microgram dose was employed.

You can see that while there is a bit more bounce here, that there is little difference in the hazard function for patients treated with pramlintide compared to those treated with insulin alone, again, starting to build an approach to this.

Before we talk about that approach, however, there is an important question. Does pramlintide itself cause hypoglycemia or does this hypoglycemia represent insulin induced hypoglycemia?

We have data to indicate that pramlintide alone does not cause hypoglycemia. In the initial dose rising study, healthy volunteers were dosed with 10,000 micrograms of pramlintide. They had a bit of nausea but there was no signs of hypoglycemia. That is 80 times the maximum recommended dose.

Pramlintide also does not inhibit the counter regulatory response to hypoglycemia. Both the time to counter regulatory hormone release and the time to glucose recovery are unaffected. There is also no impact on hypoglycemia awareness in a controlled setting as evidenced by no change in catecholamine responses and no change in the perception of symptoms.

This provides us the information that we need, I think, to make a prudent, rationale, clinical recommendation for the management of this hypoglycemic risk. We are dealing with insulin-induced hypoglycemia.

Insulin induced hypoglycemia occurs whenever the insulin effect, insulin action exceeds nutrient availability. My diagram here shows food intake going to the plasma glucose moving on out of the compartment into tissues.

I've just reviewed data with you to show that there is increased incidence of nausea in pramlintide treated patients with the initiation of therapy. The body weight data that we have reviewed indicates that pramlintide increases satiety. Both of these lead to a decrease in food intake. Therefore, if one does not alter the insulin dose, you can readily see how an imbalance could occur.

That provides a recommendation. Education for patients, physicians, and health care providers. That education focuses on using self blood glucose monitoring to make rationale changes in insulin dose just as we do routinely in clinical practice.

To provide an additional margin of safety it will be our recommendation that with the initiation of therapy that pramlintide insulin doses be reduced by 10 to 20 percent until therapy has been successfully initiated and insulin can then be titrated based upon self blood glucose monitoring.

Also for the Type 1 population we recommend that the initiating dose of pramlintide be 30 micrograms or less to limit the nausea side effect.

Having reviewed the hypoglycemia data, I would now like to return to the patients that were approaching ideal targets of seven percent here. In proven HbA1c no increase in insulin use and no increase in body weight.

We now look at the hypoglycemia data. You can see there is little increase during the first four weeks and with follow-on therapy four to 26 and 26 to 52 there is actually some reduction in the risk for hypoglycemia. As patients approach target, the risk for hypoglycemia does not increase.

In terms of the other safety observations, they are the same in the Type 1 population as they were in Type 2 and I'll move on in the interest of time.

Pramlintide in Type 1 diabetes as in Type 2 is efficacious in improving glucose control. It's accompanied with weight loss. There is an increase in insulin induced hypoglycemia but only during the initiation of therapy.

There's no increase with long-term therapy. There are no other safety issues that have been identified. The dosage recommendation is 30 micrograms or less three to four times a day for initiation and 30 or 60 micrograms for maintenance therapy.

I would like to conclude with some very brief guidelines for use of this compound in the clinic. The initial dose for Type 2 is 120 micrograms a day. For Type 1 it's 30 micrograms or lower. Dose frequency is determined by the patient's meal pattern and the drug administered within 15 minutes before the meal. Insulin reduction with initiation and 10 to 20 percent reduction in postprandial short-acting insulin dose.

For maintenance therapy, 120 micrograms for Type 2, 30 or 60 micrograms for patients with Type 1 diabetes and the insulin dose adjusted according to standard clinical practice based upon self blood glucose monitoring techniques to allow patients to optimize their glycemic control.

My presentation has provided you with the data that indicates that pramlintide is safe and efficacious for both patients with Type 1 and Type 2 diabetes. I have just reviewed the recommended dose administrations.

With that, I would like to conclude and thank the committee for your kind attention and introduce Dr. Alain Baron who is the Vice President for Clinical Research at Amylin to provide a risk benefit assessment.

DR. BARON: Thank you, Dr. Kolterman.

Dr. Kreisberg, members of the panel, I speak to you today as a physician scientist, as a dimatologist, but much more importantly as the brother of a patient who has had Type 1 diabetes for 39 years.

I would like to begin by reviewing with you some of the comments made by Dr. Polonsky regarding the risks of insulin therapy. These are well known to us.

First and foremost, the most worrisome, hypoglycemia, a very vexing issue. As we approach glycemic goals with insulin therapy, this barrier to insulin therapy rises. Often we encounter hypoglycemia in an attempt to control postprandial hyperglycemia. As Dr. Polonsky indicated, this is a major problem in patients with both Type 1 and Type 2 diabetes.

Typically we increase the insulin dose to normalize the postprandial glucose profile. In doing so we increase the risk of late postprandial hypoglycemia. This sets in motion a series of glycemic swings or occilations which are very uncomfortable for patients with Type 1 and Type 2 diabetes. That's because they are unpredictable and this causes patients much anxiety.

Thirdly, weight gain. We now know that weight gain is part and parcel of insulin therapy and is particularly vexing with intensive insulin therapy. Not only in Type 2 diabetic patients who are overweight to begin with but we now know that is also true in patients with Type 1 diabetes.

We now have novel delivery and monitoring devices and insulin analogs. These have been valuable therapeutic advances. However, they still fall short of overcoming and pushing back those barriers to insulin therapy. We definitely need more tools.

Allow me if you will to show you a schematic here representing schematically, if you will, the risks of current insulin therapy. Shown on the Y axis is reduction in HbA1c. On the X axis is a package of events, namely change in insulin dose, change in hypoglycemia risk, and weight gain.

Let's take a patient who begins here with HbA1c and let's say we want to effect a reduction in HbA1c using insulin alone. This is what happens. We definitely reduce HbA1c but we have to increase the insulin dose, increase the risk of hypoglycemia, and increase the risk of weight gain.

If one wants to effect a further reduction in HbA1c with insulin, clearly the arrow will go lower but further to the right indicating that we increase the risk and meet up with greater barriers. Clearly therapies that can push back on these barriers are extremely valuable to patients who are treated with insulin.

Let's examine then the risk benefit analysis for both Type 2 and Type 1 diabetes in that sequence. With respect to Type 2 diabetes pramlintide offers clear benefits. This was clearly demonstrated by the presentation by Dr. Kolterman. These benefits outweigh expected and well-recognized risks. These risks are also manageable.

Let's examine the risks. They are relatively few with pramlintide therapy. In patients with Type 2 diabetes the most common adverse event was nausea. This was mostly mild, infrequent, and importantly, transient. After the first four weeks, essentially there was very little nausea.

Of concern to any patient treated with insulin, in particular a patient treated with insulin in whom we add another anti-hypoglycemic agent is hypoglycemia.

We saw that in patients with Type 2 diabetes treated with pramlintide there was no overall risk of increased severe adverse event risk of severe hypoglycemia. However, this has to be taken very seriously. As suggested by Dr. Kolterman, judicious adjustment of the insulin dose particularly focusing in prandial insulin is important to avoid hypoglycemia in this population.

Let's examine how pramlintide might push back or overcome, if you will, some of the barriers of insulin therapy. We see that pramlintide reduces postprandial glucose excursions. This is very difficult to do even with the rapid acting analogs that we currently have today.

Pramlintide causes weight loss in the face of a further reduction in HbA1c that one can achieve with insulin alone. We saw no overall risk of hypoglycemia.

In patients with Type 2 diabetes who are insulin resistant, one often needs heroic levels of insulin doses to achieve glycemic if one can achieve glycemic control at all. This creates hyperinsulinemia.

Many epidemiological studies have suggested that hyperinsulinemia is not desirable as it possibly increases the risk of microvascular disease.

Pramlintide in addition to insulin allows further reduction in HbA1c with reduction of the insulin dose.

Let's now turn to Type 1 diabetes. Again, Dr. Kolterman presented compelling data showing that pramlintide offers clear benefits and these benefits also outweigh expected, well recognized, and manageable risks.

What are the risks? Again, they are relatively few but they are serious and we need to pay special attention to them. The most common is nausea. This was mild and moderate for the most part in this patient population but it was also severe in the patient population with Type 1 diabetes treated with pramlintide.

Importantly, however, it is dose dependent and transient so the management for it is very clear. One starts with therapy at a low dose, 30 micrograms or less.

Let's dwell a little bit more on severe hypoglycemia. There was a clear increased risk upon initiation of therapy with pramlintide when added to insulin in patients with Type 1 diabetes in the clinical trial setting which I remind you is double blind.

This increased risk is understandable. It is explicable and it's manageable. Let's explain it. Clearly anytime one adds a anti-hyperglycemic agent such as pramlintide to existing therapy with a hypoglycemic agent such as insulin, one increases the risk of hypoglycemia if one doesn't adjust the insulin dose.

If metformin, for example, worked in Type 1 diabetes, which it doesn't, we would expect the same risk. Moreover, pramlintide causes nausea and some anorexia. If one doesn't eat the usual meals and quantity or in timing, clearly if one is treated with insulin, there's an increased risk of hypoglycemia. Importantly, both nausea and hypoglycemia, as shown by Dr. Kolterman, is a dose dependent related phenomenon. The path forward for management is very clear. We start once again at a low dose, 3 micrograms or less.

Again, as recommended by Dr. Kolterman, one begins initiation of pramlintide after education and the same recommendations that one would have upon initiation of insulin therapy in a patient newly diagnosed who is insulin naive, or in a patient who is currently treated with insulin in whom we wish to intensify insulin therapy.

What barriers does pramlintide push back on when added to insulin in patients with Type 1 diabetes? In fact, there is no overall increased risk of hypoglycemia despite that fact that we achieve further glycemic control beyond that achieved with insulin alone.

Recall that in the first four weeks the event rate of hypoglycemia is higher. However, overall over the 52 weeks, it is not higher. That means that post initiation of therapy, and the data indicate this, there is actually possibly a reduction in risk of hypoglycemia.

With respect to weight gain, we see weight loss in patients with Type 1 diabetes whose HbA1c has improved and this reduction in weight occurs particularly in patients with Type 1 diabetes who are overweight to begin with and we'll be happy to share that data with you in the question and answer period.

Finally, because of the unique mechanism of action of pramlintide, pramlintide is able to reduce postprandial glycemic excursions in a fashion that is not possible with insulin therapy alone.

Let's ask the important question. Is a reduction in HbA1c obtained with pramlintide worthwhile? Well, we saw average reductions in HbA1c in the pramlintide trial program of .3 to .7 percent in the stable insulin population versus placebo or insulin alone, if you will, and .5 to 1 percent versus baseline.

Clearly these reductions in HbA1c are worthwhile. In fact, according to DCCT data a .5 percent reduction in HbA1c as applied to this patient population that was treated with pramlintide would be expected to result in approximately a 30 percent reduction in the risk of microvascular disease, particularly retinopathy.

Why use pramlintide? Well, I think we know by now, to further reduce HbA1c and obtain glycemic goals beyond that achievable with insulin alone. To control postprandial hyperglycemia and limit the associated glycemic swings. And, importantly, to minimize the weight gain which is part and parcel of insulin therapy.

How do we achieve these benefits with pramlintide? This is due to its unique mode of action. Pramlintide limits postprandial glycemic excursions by two unique mechanisms. Suppression of the inappropriate secretion of glucagon in the post-meal period. Parenthetically this is not achievable with excongenous insulin therapy currently in use.

Moreover, it regulates the nutrient delivery rate to the small intestine.

Both of these effects together, suppression of glucagon secretion, modulation of nutrient delivery, are complementary and additive to insulin in controlling postprandial hyperglycemia.

Let's reexamine the schema again. Remember, if we want to reduce HbA1c with insulin alone, one goes in this direction. We increase the insulin dose, increase hypoglycemia risk, and increase risk gain.

With pramlintide the data suggest the following. Because of its unique mechanism added to insulin, we go in a different direction. We achieve the same HbA1c reduction. However, without an increase in insulin dose, without no net increase in overall hypoglycemic risk, and no weight gain, or perhaps weight loss.

No manipulation of insulin therapy alone, not in the types of insulin, the regimens applied can move this arrow to here so pramlintide offers unique benefits.

The complementary actions of insulin and pramlintide form a potent binary therapeutic tool to control postprandial glycemic excursions. You need both insulin and pramlintide as the B-cell intended to leverage the effects of insulin to lower glucose further than can be achieved with insulin alone without the increased risk of weight gain and having to increase the insulin dose.

This is not an either/or proposition. The two together make insulin work better and facilitates the attainment of glycemic goals.

Amylin replacement with pramlintide represents a novel. In fact, in Type 1 diabetes this is the only novel drug in 80 years for patients with Type 1 diabetes. It also represents a unique therapeutic tool. Why? Because it has a unique mode of action. As such, it represents an important therapeutic advance that fulfills a need for patients with diabetes treated with insulin.

I look forward to being able to use pramlintide to better the lives of patients with diabetes. Thank you very much for your attention.

DR. KREISBERG: I'd like to thank Amylin Pharmaceuticals for their on-time presentation. Appreciate that.

We're scheduled in a couple of minutes to take a break for 15 minutes, but I would like to ask the panel if they would like to ask some short questions to clarify any of the material that was presented, again reserving questions requiring extensive discussion for later.

DR. TAMBORLANE: I had a question for Dr. Kolterman. You made a comment that the 30 microgram dose in the Type 1 simulated the increase in amylin that you would see with a meal in a nondiabetic.

DR. KOLTERMAN: That is correct.

DR. TAMBORLANE: How does the 120 microgram dose in the Type 2 compare to the nondiabetic excursion?

DR. KOLTERMAN: The peak concentrations are somewhat higher but by maybe only 15 per ml. It's not out of the ballpark. I mean, we're not doubling or tripling the concentration as seen in patients with Type 2 diabetes.

What do you compare that to? Do you compare it to a normal nondiabetic individual or do you compare it to that individual as they have evolved into the state where we actually implement treatment? Those individuals clearly go through a period of time where they are hyperamylinemic just like they are hyperinsulinemic.

DR. TAMBORLANE: I mean, you could compare it to age and weight match control, older more overweight individuals. You made a very persuasive and eloquent argument about replacing normal physiology and that was the issue that I was curious about.

DR. KOLTERMAN: If you compare the concentration to those individuals you just described, age matched and weight matched, the 120 microgram dose is right in the ballpark.

DR. TAMBORLANE: I had one more just informational thing. Those change that you presented was percent change in dose?

DR. KOLTERMAN: You're referring to insulin dose?

DR. TAMBORLANE: Yes, insulin dose.

DR. KOLTERMAN: That's correct.

DR. TAMBORLANE: We're talking about 4 or 5 percent which for most Type 1's, I assume in absolute terms, was one or two units. Is that probably correct?

DR. KOLTERMAN: It's more on the order of probably four to five or six units of insulin.

DR. TAMBORLANE: They were receiving 100 units to start?

DR. KOLTERMAN: No, the mean insulin dose at baseline, I believe if my memory serves me correctly, was about 60.

DR. TAMBORLANE: So it was 2.5 to 5 percent so it's like 3. Okay.

DR. LEVITSKY: This is a question on physiology. I think I'm being terminally naive about this but diminished gastric emptying has always been felt by me to be a disorder which occurs as a result of long-term poor diabetes control.

It is usually something which interferes with our ability to control people with diabetes because you never quite know when you're going to get your hypoglycemia or your peaks. I'm not exactly sure how a drug which leads to decreased gastric emptying is causing a physiologic change. I'm mystified by this.

DR. KOLTERMAN: Okay. Let me make an attempt to shed some light on this. As we typically treat hyperglycemia in diabetic patients with insulin, we are working only on the output side of the system to stimulate the output or we are working predominately on the output side to stimulate the removal of glucose from the circulation.

Something that affected the rate of delivery of nutrients to the small intestine for absorption will work on the input side so there is a more gradual delivery or a lower rate of delivery of the nutrients to the circulation.

In terms of the issue I think you're struggling with, to state it a different way, are we creating gastroparesis in these patients. Pramlintide does not induce gastroparesis. What pramlintide does is slows or retards the rate of gastric emptying.

We have some gastric emptying data that we'll be happy to share with you during the question and answer period this afternoon. What you see, and I think Dr. Young showed one slide of that, is that it prolongs the time of emptying by approximately an hour or so. We're not arresting it.

If you remember the PK profiles that I showed, I tried to emphasize that plasma concentrations were gone by the end of three hours. Typical time between meals is on the order of four to five hours, so we actually have gastric emptying data to show that the dose given before the morning meal lowers postprandial glucoses and it delays gastric emptying after the morning meal.

If you look at gastric emptying after the

midday meal, if you don't give another dose of pramlintide. What I'm doing is an experimental paradigm where you are looking at the effects of a morning administration of pramlintide on gastric emptying after the mid-day meal you see absolutely no effect on the gastric emptying after the mid-day meal.

DR. LEVITSKY: A practical question. Would you anticipate then that people taking this drug would take a very short-acting insulin like Lispro and then another dose of pramlintide at the same time with each meal to achieve optimal control, that they would need to take the two injections?

DR. KOLTERMAN: I think to achieve optimal control based on the information that we have now, that would probably be the case. The addition of pramlintide to that regimen would allow patients to use less of the short-acting analog prior to the meal. The addition of pramlintide will also blunt the early postprandial rise in glucose that you still see with most situations with the use of the short-acting analog.

Also by virtue of the mechanism of action that we just talked about by having more of the nutrient load available later in the postprandial period will provide a buffer against the late postprandial hyperglycemia.

I'm sorry. I misspoke there. The late postprandial hypoglycemia.

DR. KREISBERG: Dr. Grady.

DR. GRADY: Do you have any idea what the mechanism of the nausea is?

DR. KOLTERMAN: I have no direct evidence to mechanisms there. As Dr. Young reviewed, there are high density -- there is a high density of amylin binding sites in the area postrema. The area postrema is a region of the brain that the gastrointestinal physiologist tell us is intimately involved in the regulation of gastrointestinal functions.

Other areas that appear to hit the area postrema also sometimes induce nausea. It's consistent with the mechanism of action through the area postrema. Exactly what it is that probably happens in the area postrema that leads to the perception of nausea I don't believe we can tell you today.

DR. GRADY: Can I ask one more question? Do you know what the effect of the drug is on HDL and LDL cholesterol? You presented in your slides only total cholesterol.

DR. KOLTERMAN: We have data from each of the six long-term studies looking at changes in the fasting plasma lipid profile meaning total cholesterol, triglycerides, HDL and LDL cholesterol. The overall pattern is that of no change.

There is one study, the initial study in Type 1 diabetes, study 137-112, where there was a beneficial reduction in LDL and some increase in HDL cholesterol. That has not been clearly observed in the other studies.

DR. KREISBERG: I now have 7 minutes after 10:00 by my watch and we're going to break for 15 minutes and whatever time that is. 22 after is when we're getting back.

(Whereupon, at 10:08 a.m. off the record until 10:27 a.m.)

DR. KREISBERG: Can I ask everybody to please sit down so we can begin. After we completed the brief question and answer session, several other panelist indicated to me that they had some questions that needed further clarification so I would like to ask the Amylin people to respond to these.

Dr. Sampson.

DR. SAMPSON: Dr. Kolterman, I had just two small technical questions. First of all, I was wondering if you could say a little bit more about the difference in formulation in study 111? You indicated lower bioavailability. I'm just wondering how to interpret the results of 111 in the context of your statement.

DR. KOLTERMAN: Sure. The only difference in the formulation was a ph difference. It was ph 4.7 versus 4.0. The relative bioavailability was decreased by approximately 30 percent.

DR. SAMPSON: Was that measured in a common study that compared the bioavailability or were those in separate studies?

DR. KOLTERMAN: The comment about the decreased bioavailability comes from study 137-126 or 125 where the two formulations were compared in the same subjects in the same protocol.

DR. SAMPSON: Thank you for that. The other one is a more technical question. That is, in study 123 there is an indication in the statistics that the inference was going to be done because of the multiple doses. As you are aware, there is a question of the simultaneity of the inference and what is an appropriate critical p-value. I don't have access, unfortunately, to your protocol. Otherwise, I could answer this myself.

The synopsis I have says a step-down procedure was used to assess inference there. I'm wondering what was the -- and the treatments. What were the treatment step-downs and in what order? There 120 bid and 90 tid. I'm wondering which was the first one supposed to be tested in that step-down procedure?

DR. KOLTERMAN: I believe that the initial dose arm that was to be tested was the 90 microgram three times a day arm. That prespecified endpoint was not met. The p-value that was on the slide for the 123 study represents a nominal p-value comparing the 120 bid arm versus placebo.

DR. SAMPSON: Thank you. I understand that.

DR. GELATO: I wanted to go back to the gastroparesis issue for just a moment and just ask you because nausea is certainly a symptom of gastroparesis in patients, I just wondered if in the patients if you saw any correlation between those who may have had gastroparesis and when they were put on the drug, maybe that was the group that had the most severe gastroparesis or whether you even looked at the data in that way.

The concern would be if you had a patient with gastroparesis, would you feel comfortable using this drug? From what I understood before, I'm gathering your answer to that is that, yes, you would. I wonder --

DR. KOLTERMAN: There are several points to be made. First is, is that patients with diabetes who have bona fide dense gastroparesis have very dense vagal neuropathy in terms of intervasion of the gastrointentional tract. We have data from preclinical studies that suggest that an intact vagas nerve is required for the effect upon gastric emptying. Should pramlintide be given to a patient with true gastroparesis, I don't think that anything untoward would happen to the patient which would be our primary concern. Whether there would be benefits in terms of reduction in postprandial glucose is a different question.

We attempted to do a study to look at that directly. The problem that we had is in spite of going to three institutions that publish frequently about the incidence of gastroparesis in diabetes, we failed to identify an adequate number of studies to complete the protocol.

DR. GELATO: If I could ask another question. This may be a point that maybe you can clarify for me. My understanding was that in Type 1 diabetes after about five years or so into the disease, that glucagon is really not a component of a problem with that disease and may not even be relevant.

I wonder then if the mechanism you're looking at is really in Type 1 a glucagon mediated mechanism or one that has solely to do with just the effects on gastric emptying and slowing of the intensity.

DR. KOLTERMAN: Excellent question. I believe that both the glucagon effect and the gastric emptying effect are relevant to patients with Type 1 diabetes. The data that you cite related to "an absent glucagon response after five years of Type 1 diabetes," refers primarily to the glucagon response in insulin-induced hypoglycemia.

There are papers, nicely done studies in the literature that clearly demonstrate that those patients who have lost their glucagon response to hypoglycemia still have postpandrial hyperglucagonemia in response to the ingestion of presumably food stuffs that contain a stimulatory amino acids. Even in those patients there is excess glucagon during the postprandial period.

DR. GRADY: One important thing about just the quality of randomized trials is whether or not they're blinded. I wonder if there are ways in which the participants or the investigators could know that the participant was taking the active drug. Does it taste different? Certainly it produces a lot more nausea. Do you have any information? Did you ask participants if they knew what they were taking?

DR. KOLTERMAN: Okay. There's no indication that participants knew what they were taking.

DR. GELATO: Did you ask them?

DR. KOLTERMAN: At some centers they were asked at the end of the trial. That was an ad hoc thing so it's anecdotal information. This is not information that was assessed in a prospective controlled manner.

DR. KREISBERG: Dr. Kolterman, with regard to the formulation, I'm curious why there was this mid-stream correction in the ph? Was that because it was not formulated correctly when you started or were you just tinkering with the preparation in order to maximize bioavailability?

DR. KOLTERMAN: Okay. The change in formulation, as I understand it, relates to the studies that were done showing an effect of ph upon long-term stability of the product in the formulation.

This is not atypical for what happens with various drug formulations during the development program. It so happens at the time that the study that I identified for you was initiated, the status of the drug supply was ph 4.7 formulation.

DR. KREISBERG: Dr. Levitsky.

DR. LEVITSKY: The question I have is did you in the course of any of these studies examine nutrient intake? I assume that total calories, total energy was diminished or they wouldn't have been losing weight but did you look to see whether the nausea effect introduced some changes in the quality of the nutrients that were taken in so that perhaps carbohydrate intake was diminished? Could that be clarified a little bit to see whether you could determine why the glycemic control remained improved with time?

DR. KOLTERMAN: Again, we do not have controlled data addressing that in the program. There are anecdotal reports that come primarily from long-term open label safety studies where investigators and trial coordinators have reported to us an impression that they have of the patient changing somewhat the composition of the nutrients that they ingest in terms of moving away from fat-containing foods or foods rich in fat to those that contain more complex carbohydrates. Understand that is anecdotal reports at this time.

DR. TAMBORLANE: Orville, have you tried to break down the incidence of severe hypoglycemia as a function of time of day?

DR. KOLTERMAN: We have evaluated that. The data is divided into four segments beginning at 8:00 a.m. The period from 8:00 a.m. until about noon time and the period from noon until the evening are roughly the same.

There is a clustering of events from like 6:00 p.m. or so until 10:00 p.m., 11:00 p.m. Bedtime basically. Then there is certainly no increase of events during the night time and there is actually a trend toward decrease in hypoglycemia over night.

DR. GRADY: Do you know what proportion of patients were taking postprandial short-acting insulin?

DR. KOLTERMAN: I can give you a ballpark figure for Type 1. For Type 1 patients over 80 percent -- I'm sorry, over 90 percent were taking two injections a day. There were very few patients taking a single injection a day.

Of those patients taking two or more injections a day, they almost all were using either a premixed formulation or a self-mixed formulation that gave short-acting and intermediate acting or long-acting insulin in the morning and the evening.

There are, if my memory serves me correctly, approximately 35, 40 percent of patients that were using postprandial short-acting insulin at mealtime and/or using insulin pump therapy.

Type 2 population was the twice daily means of administering insulin was more common. I don't have the Type 2 numbers in my memory bank, as well as the Type 1 numbers.

DR. GRADY: And one more question. There were patients who changed their insulin doses during the trials. Do you have any data concerning whether or not increases in insulin dose were associated with increases of hypoglycemic episodes?

DR. KOLTERMAN: We have looked at that and there does not appear to be an increased risk for hypoglycemia in those patients that increased their insulin.

DR. KREISBERG: I think we'll go on with the program now. We have the FDA scheduled to make a presentation. We have allocated one hour for their presentation. Dr. Robert Misbin will lead off.

DR. MISBIN: Mr. Chairman, ladies and gentlemen. May I have the first slide, please.

The FDA presentation will be in two parts. I will be making the efficacy presentation and then Dr. Dragos Roman will make the safety presentation and will focus primarily on the problem of hypoglycemia.

This is a slide that I took directly from the sponsor's briefing document. It shows an overview of the Phase 3 trials in Type 1 diabetes. What is shown here is the placebo subtracted reduction in HbA1c. There were three Phase 3 trials. One had one arm the other two each had three arms.

This slide here shows there is no difference from placebo and anything below that line would show a reduction in HbA1c or a benefit of pramlintide.

As you can see, not all of the individual points are statistically significant. But leaving that aside, there is a reproducible decrease in HbA1c across the board of approximately 0.3 percent units.

What I've added now is something that was not in the sponsor's briefing document and is the HbA1c values at endpoint that correspond to these reductions. Now, as has been discussed earlier, the American Diabetes has said that the goal of treating Type 1 diabetes is to get the HbA1c down to 7 percent. That is the goal of treatment.

As one can see, even despite six months of treatment with pramlintide, all of the HbA1c values here were quite high and would indeed be considered unacceptable by most good clinicians. This, I think, is the major problem that we have with respect to the data that was presented to us.

I'm going to discuss one trial in detail. I'm going to discuss the middle trial here. Here. I'm sorry. These three arms from this one trial, trial 117. Later Dr. Roman will present some data briefly on this trial as well.

This is sponsor's trial 117. It was in patients with Type 1 diabetes. They were all on stable metabolic regimen, as you heard earlier. To get into this, the inclusion criterion was a HbA1c of at least 8 percent at screening.

The approach in this trial was that patients were to remain on their usual diet, type of insulin, insulin regimen, and exercise regimen and exercise regimen as Dr. Kolterman indicated to you earlier.

The mean age of these patients was 38 years. They had a mean duration of diabetes of 16 years. They were not obese. Their mean bmi was 25. HbA1c was 9 percent and they were taking an average of 50 units of insulin per day. This, I think, is very reflective of patients with Type 1 diabetes, indeed, reflective of the patients that Dr. Polonsky said we need additional tools to treat.

Now, let me remind you -- I believe this was discussed earlier -- what are the recommendations of the American Diabetes Association? The goal of treatment is to get the HbA1c level down to 7 percent or less.

At a value of 8 percent or greater the American Diabetes Association says that additional action is suggested, which presumably in patients with Type 1 diabetes would mean further adjustment of their insulin regimen.

This is a repeat of the slide I just showed about the patients with Type 1 diabetes in this trial. They were on a stable metabolic regimen. They had an HbA1c of at least 8 percent and they were told to remain on their usual type of diet, type of insulin regimen, and exercise regimen.

I think it's fair to say that these instructions are exactly the opposite of what the American Diabetes Association believes is the standard of practice for treating patients with Type 1 diabetes. This is really the major problem that the FDA has with accepting the data that we heard earlier.

It's our belief that new drugs should be tested in a way which is consistent with the way which those drugs are going to be used. Unless the sponsor believes that these recommendations are wrong, we do not understand why the drugs were tested in this particular matter and, therefore, have difficulty accepting the data as being clinically relevant.

This is the data that I'm going to show in this trial. I've plotted the HbA1c. This is not the reduction from baseline from this is the actual values which a clinician would see if he was following a patient.

Six percent is the approximate upper limit of the normal range. Seven percent is the goal of treatment by the American Diabetes Association. Eight percent is the value which the American Diabetes Association says something needs to be done. Then 9 percent is the starting value in most of these patients.

As one can see just looking across, all these values are really quite high. They are much higher than would generally be considered acceptable.

I think if one looks at these results, one does see a consistent fall in HbA1c. This is placebo and these are the three treatment arms, 90 micrograms bid, 60 micrograms tid, and 90 micrograms tid.

There is a consistent fall in HbA1c at four weeks. The purple is baseline and the dark blue is four weeks. The light blue is 26 weeks. There is a consistent fall in HbA1c in the treated groups. This is statistically different from the small dip that you see with placebo alone. However, when you go out to 26 weeks, you can see really in every case the HbA1c value is going back toward the baseline.

At this point I would recall a slide, or perhaps two slides, that you saw earlier from the Phase 2 trials showing that pramlintide is very effective in blunting postprandial hyperglycemia. These data, I think, are very impressive. Looking at this, I could understand the sponsor's enthusiasm when they did those trials many years ago why they were enthusiastic about developing pramlintide. Indeed, if you were to stop the trial just at these dark blue lines and look at the fall in four weeks recognizing that HbA1c is a lagging indicator of glycemic control, you would believe that indeed you had a potentially successful treatment.

Unfortunately, diabetes is a long-term disease. The agency requires 26 weeks of efficacy data and control trials and 52 weeks total. One can see here that by 26 weeks one is clearly going back toward the baseline. The efficacy, as we see it, was really not sustained.

I would also recall a statement I think Dr. Baron made earlier. I think he said that the way he saw the data, hypoglycemia was a problem early in the trial but then tended to go away. Nausea was a problem early in the trials and then tended to go away. I think we would largely agree with that.

I would also point out that the efficacy also tended to go away. From our evaluation, the main problem here is that all of the effects of pramlintide both good and bad seem to be very transient.

I need to point out this is an intent-to-treat analysis and this is a very conservative way of looking at data. We recognize there are no other treatments for Type 1 diabetes and it is unreasonable to expect that every patient with Type 1 diabetes would necessarily respond to a new drug in the same way.

We are willing to look at subsets to try to determine if there are any specific patients who might benefit from pramlintide. The subset that I'm going to describe is one that was put forward by the sponsor as a way of identifying patients who based on a four week determination seemed to be responding to the drug.

This is what the sponsor has called the early glycemic responders subgroup. In order to get into this subgroup the way it's defined, it's a reduction in HbA1c of at least 5 percent units at four weeks.

What I'm going to show now is data, mean data, for this subgroup at both four weeks when it is described and at 26 weeks which is the endpoint of the trial.

The first line, I have to orient everybody to this because it might be confusing otherwise. This is placebo and the three arms of the trial. It's not necessary to look at all of these data. Basically the results here are all pretty much the same so I wouldn't concentrate on that.

The point is that expressed as a percent of the total intent-to-treat population, 25 percent of the patients on the placebo were defined as being responders, and roughly 40 to 45 percent of the patients on pramlintide. This, I think, is the efficacy determination in this type of analysis.

When I was listening to Dr. Kolterman, I realized something that I had forgotten. This trial 117 is one of the three trials that was not statistically significant by the initial prespecified statistical evaluation. This data, I think, the way Dr. Kolterman presented it could be considered a negative trial.

Nevertheless, if you look at this subset of population, you would say that the effect of the drug in getting you into a responder group was almost twice as great as placebo. Even though the initial evaluation might be negative, looking at it this way you could see that they might actually be efficacy even in this otherwise negative trial.

But that's not the reason I want to show this data. I mean, we'll take it -- the accept the fact that there were more responders on pramlintide than placebo.

The reason I want to show it is something else. That is, I think looking at this gives you a handle to try to differentiate the effects of lower glucose levels either with pramlintide or with insulin alone.

To illustrate that point, let me show you that in four weeks when the subgroup was defined, the mean HbA1c across these groups was roughly the same, about .8 percent units in the placebo patients as well as in the HbA1c treated patients. At 26 weeks the HbA1c reduction in the placebo patients as well as the pramlintide treated patients was approximately .5 percent units.

Since these values are roughly the same, I think it's reasonable to pose the question given a specific level of HbA1c reduction achieved either with placebo or with pramlintide what is the difference in the way one got to that level.

This is the mean data at 26 weeks. Again, the measure of efficacy here, 25 percent response in placebo versus 40 to 45 percent on pramlintide. The HbA1c reduction in the placebo group, and I say placebo but this really should be insulin alone. We have to remember that all these patients were taking insulin.

Insulin alone reduction across the board, either insulin alone or insulin plus pramlintide you saw the same HbA1c reduction of about .5 percent.

Let's look at the placebo group. This .5 percent reduction in the placebo group, how was it achieved? It was achieved by giving them a little more insulin. Now, the starting value is around 50 units, I think, so this is a very small amount of insulin.

Nevertheless, by giving a little more insulin in this group of patients, you achieve this .5 percent reduction at HbA1c. But they also achieved a mean body weight gain. This is, I think, characteristic of the treatment of diabetes. If you want to lower glucose levels, you give a little insulin and you find that you gain weight.

By contrast, in order to get the same reduction in HbA1c on pramlintide, you could actually give less insulin by in large. It was not totally reproducible but, by in large, you gave less insulin.

A very reproducible finding was that there was a very substantial reduction in weight between .4 and 2 kilograms which is, I think, quite substantial when you consider that the placebo patients gained a kilogram.

If you look at these data, you would say this really isn't bad. You get the same reduction in HbA1c and you lose weight relative to gain weight. If there was nothing else to say about this, I think we would all accept the fact that preventing weight gain is desirable.

Unfortunately, there is, however, a price to be paid. The price to be paid is the annual event rates of severe hypoglycemia presented here as events per patient year. It was very low in the placebo patients, 0.2 percent per year, but in the pramlintide treated groups, there was a reproducible five to eight fold increase in the event rate of severe hypoglycemia.

Really the question is a small reduction in body weight worth a five or seven or eight fold increase of severe hypoglycemia recognizing, as Dr. Roman will discuss later, that some of these events, these were all severe events as described by the sponsor but some of them had very serious sequelae and, indeed, divesting sequelae.

Here I'm showing the rest of the data. This is now at four weeks and, I think, here this is just more of the same. The HbA1c reduction across the board was about .8. Again, you see here a four to 12 fold increase in the event rates of severe hypoglycemia here expressed as events per patient year.

Again, for the same degree of HbA1c reduction, you get an enormous increase in the reporting of severe hypoglycemia.

I do need to point out in the interest of fairness that there is a small flaw in this analysis. The starting value of HbA1c was somewhat higher in the placebo -- in the subgroup anyway was somewhat higher in the placebo patients than in the pramlintide treated patients. To some extent, some of this difference might be accounted for. A small amount might be accounted for by that baseline imbalance.

I would point out that at this high level a small difference in HbA1c would not of itself be expected to cause much of a difference in the hypoglycemia rate and certainly could not account for an eight-fold increase in the event rate expressed per patient year.

In leaving that, I would like to remind everybody why we treat diabetes in the first place. This is from the New England Journal of Medicine published in 1993. This is from the DCCT trial which we've heard a lot about but didn't actually see any raw data.

Here we have HbA1c here expressed as glycosylated hemoglobin, mean values and the year of study. The starting value here was about 8.8 percent which is virtually identical to what we saw the starting value in the pramlintide treated patients that we heard about today.

In the DCCT trial a convention treatment represented one to two injections of insulin per day and intensive treatment with three and more injections of insulin per day or an insulin pump. The point of this slide is that we shouldn't forget that it is possible to lower HbA1c levels very dramatically with insulin alone and to keep them down there for an extended period of time.

Based on this trial, this is the reason that we treat diabetes, the basis for the goals established by the American Diabetes Association. It is also the basis of using glycosylated hemoglobin by the FDA as the surrogate endpoint for approving new drugs to treat diabetes.

But what I would like to -- the reason I'm showing it today is to kind of bear this in mind as to what might be considered a reasonable goal post to how diabetic patients should be treated in the community.

Now, I completely recognize that it's one thing to set a goal and it's quite another thing to achieve that goal. I also agree with what was said earlier, that we do need additional tools.

But in considering today's application, I think the committee should ask itself really given what we know about pramlintide, is that going to help patients achieve a goal or is it going to make it more difficult, or perhaps it's really going to delay even trying to achieve the goal by some kind of futile effort at a product that may not be very effective.

In trying to consider this, I would just pose the following observations. Pramlintide has to be given by three and four -- what is being proposed is it be given by three and four injections per day in addition to insulin.

It cannot be mixed with insulin and has to be given by additional injection. Three or four injections of an additional drug over and above the insulin dose that a patient is taking is really quite a substantial burden on a patient.

Secondly, by our analysis there is much more severe hypoglycemia with pramlintide than reducing HbA1c with insulin alone. This is by far the most important thing that I think we have to bring to your attention. It is in my judgement the single most important reason that would be preventing the approval of pramlintide.

On a more positive note, I would agree with the sponsor that there is weight loss on pramlintide relative to insulin alone. That clearly is the case and of itself would be considered desirable but would have to be balanced to get all of these other risks and burdens that are imposed by using pramlintide.

Now, weight loss is certainly desirable in everyone, or most people anyway, my self included perhaps, but it's more desirable in patients with Type 2 diabetes than Type 1 diabetes. You might expect given a drug like pramlintide which causes weight loss, you might expect that pramlintide would be more effective in patients with Type 2 than in Type 1 diabetes.

The data, however, this is data from Type 2 diabetes, and I don't think one could make the statement that it's more effective in Type 2 than in Type 1. The format of this slide is exactly what I showed earlier with Type 1 diabetes. Again, this is taken directly out of the sponsor's briefing document and it shows the three arms from each of their three Phase 3 trials.

Again, one sees across the board, again, not all these points are statistically significant from zero. There are several of them that are not, but we're not going to make anything of that. We're just going to assume that the point estimates are absolutely right and just deal with the proposition that pramlintide lowers HbA1c levels by about .3 or .4 percent units.

What I'm going to discuss now in detail is, again, the middle trial. Just like I did with Type 1 diabetes, the middle trial here for Type 2 diabetes is what I'm going to show now.

This is the sponsor's trial 123. Mean data at baseline is what I'm showing here. The patients were 58 years old. Again, mean data, 13 years of diabetes. They were mildly or moderately obese, bmi of 30.6 which again is very typical for patients with Type 2 diabetes.

HbA1c of a mean value at baseline was 9.4 percent. They were taking an average of 56 units of insulin per day. Again, these, I think, are very typical of the types of patients that indeed need additional treatment.

The patients had to have an HbA1c of 8 percent or greater at screening. They were to remain on their usual diet and insulin regimen and exercise regimen just like we heard before.

This is a direct quote from the protocol and I'll quote now. "Changes in insulin doses were in encouraged in order to limit the impact of alterations in insulin dosing on glycemic control."

As Dr. Kolterman said, the purpose of doing this is to isolate to the effects of pramlintide and, indeed, maximize the effects of the drug so one could observe that over and above the effects that might be seen by altering the insulin dose in some way.

Despite the efforts to isolate and maximize the effects, I think the effects are really very small. Again, 6 percent is the upper limit of the normal range, 7 percent is the goal set by the American Diabetes Association, 8 percent is the level that the American Diabetes Association says something needs to be done, 9 percent is the starting value.

In most of these 9 to 9.5 percent is where we started in most cases. Again, one can see, as we've seen many times, there is a reduction initially, this time 13 weeks, in HbA1c but by 26 weeks one is going back toward the baseline here almost 9 percent.

Now, Type 2 diabetes is characterized by hyperschlonemia, large insulin dose, obesity, insulin resistance. It is difficult to lower HbA1c levels. One often has to use more insulin and just kind of chase a vicious cycle. I think the sponsor might say, "Maybe this is the best you can do." Furthermore, this is not ordinary practice. This is a clinical trial so we are justified in doing our trials in this way.

Well, I don't really see it that way. To illustrate another way of doing a trial, I would show you this data that was published from Annals of Internal Medicine of testing of metformin in patients with Type 2 diabetes, also insulin treated patients with Type 2 diabetes.

Again, the axes are exactly the same. I haven't changed them. They are exactly the same as what we saw with pramlintide starting at 6 percent and 7 percent is the goal set by the American Diabetes Association.

You can see immediately that the shape of these curves is markedly different from what we've seen before and there is a dramatic reduction in HbA1c during the 24 weeks of the trial. What may not be obvious is that this dramatic reduction was not due to the studied drug metformin. No, this reduction was due to insulin.

This trial was done by Phil Raskin at the University of Texas. He and his colleagues treated these experimental subjects as if they were their own private patients. They treated them in such a way that they were able to bring down the HbA1c very dramatically.

Under these circumstances of good clinical care, the further reduction that one saw with metformin, this .9 percent here, is, I think, clinically important because it's over and above what one could reasonably be expected to achieve with insulin alone.

In other words, if this is good control, then this is better control and metformin really has added something that would not have been present otherwise.

Let's examine something more about these patients. What I showed in the figure is now showed here in the table. Again, they had a starting value of HbA1c of around 9 percent. With insulin alone on placebo the reduction was 1.6 percent. With metformin the reduction was 2.5 percent so the treatment effect due to metformin was a reduction of .9 percent.

These patients were taking a lot of insulin. They were taking more insulin than in the pramlintide trials for Type 2 diabetes that we heard today. They were taking almost 100 units. They were also very obese. They had a mean weight of well over 100 kilograms.

On placebo alone in order to achieve this reduction in HbA1c, they took 23 more units of insulin than what they started with. They gained 3.2 kilos on average. Those things would be considered as undesirable.

However, in the presence of metformin in addition to lowering the HbA1c, there was a mean reduction in the insulin dose. Although they gained weight slightly, most of this weight gain was mitigated and there was a mean 2.7 percent kilogram reduction in body weight.

I'm not showing this data to make a comparison of pramlintide to metformin. That's not the purpose of showing this slide. The reason I'm showing this slide is to make a comparison to how this trial was done to how the pramlintide trials were done.

These data are not just statistically significant. They are clinically meaningful because they were done under circumstances of good medical practice. Therefore, I think we, the agency, as well as clinicians can look at this and be confident of these data.

This is not the same as constructing an artificial design and coming up with the statistically significant reduction in any one of these variables that may or may not have any relevance if applied to real patients.

This brings us then back to pramlintide. Again, I would point out that although there is this brief reduction at 13 weeks -- yes, it does go down a bit -- at the end of 26 weeks we are well back on our way to where we were.

In comparison to the metformin data that I just showed, as well as to the DCC data that we are all familiar with, to me it looks like the real result of being in one of these trials is the perpetuation of the state of hyperglycemia. It is not at all clear to me how these data can be considered relevant to treating real patients with diabetes.

That concludes the efficacy presentation by the FDA. I would like to turn to Dr. Roman for the safety presentation.

DR. ROMAN: Dr. Kreisberg, committee members, the purpose of this presentation is to discuss solely the major safety issues that are associated with the use of pramlintide in the treatment of patients with Type 1 and Type 2 diabetes.

Our review process has covered extensively a vast majority of the safety information submitted with the agency. While a few areas are still under review, the only important safety signal we have come across so far is severe hypoglycemia.

Therefore, the focus of this safety review is severe hypoglycemia as it has been observed during the long-term controlled pramlintide studies.

In this presentation I will cover the following topics. First, I will briefly describe some of the features of the Phase 3 clinical trials in order to provide an understanding of the clinical context during which severe hypoglycemia occurs.

Second, I will discuss aspects of severe hypoglycemia. Assisted hypoglycemia defined as any hypoglycemic event requiring another person's help for treatment and serious adverse events, or SAEs, associated with hypoglycemia in Type 1 diabetes trial.

As an extension of the SAE category, I will discuss motor vehicle accidents associated with hypoglycemia and other types of trauma and injuries associated with hypoglycemia as they occur during the Phase 3 Type 1 diabetes trials.

Both hypoglycemia and SAEs associated with hypoglycemia will be discussed in the context of the long-term controlled Phase 3 clinical trials.

Among the 51 studies which constitute the pramlintide clinical program, the Phase 3 studies allow the most extensive side-by-side comparison between the pramlintide and placebo treatments both in terms of duration up to one year, and number of patients over 2,000.

From the start I would like to make the following clarification. In this presentation pramlintide treatment will always mean pramlintide injection plus insulin injection. Placebo treatment will always mean placebo injection plus insulin injection.

Shown here is a cumulative summary of the Phase 3 Type 1 diabetes trials. There were 1,179 patients enrolled in the pramlintide group and 538 patients in the placebo group. Only 75 of the patients completed the placebo arm and even a lower number, 66 percent, completed the pramlintide arm.

Patients who withdrew for all reasons were 34 percent in the placebo groups and only 25 percent in the -- I'm sorry, in the pramlintide groups and only 25 percent in the placebo groups.

When one looks at withdrawals due to adverse events, they were 18 percent in the pramlintide group and 6 percent in the placebo group. There was a three-fold difference.

Then adverse event with unusually high frequency resulting in early withdrawals in the pramlintide treatment group was nausea.

Shown here is the frequency of nausea related to withdrawals during the first month of treatment in the Type 1 diabetes trials. On the Y axis you have percent of patients who withdrew. The yellow bars represent pramlintide group. The blue bars represent placebo group.

Individual studies 121, 117, and 112 as well as all studies combined are presented. It is quite striking that nausea related withdrawals occur across all studies many times over placebo. On the average, there is a 17-fold ratio between percent of nausea related withdrawals in the pramlintide group and the placebo group respectively.

The net effect of this occurrence is the early preferential loss in the trial of patients sensitive to the effects of pramlintide.

Shown here is the cumulative summary of the Phase 3 Type 2 diabetes trials. A similar number of patients were enrolled in these studies, 1,273 patients in the pramlintide group and 420 patients in the placebo group.

The subject indemnization ratio was the same as in the Type 1 diabetes studies three to one. There was an equal percentage of patients who completed trials and an equal percentage of patients who withdraw from the trials. However, adverse events were slightly higher, 90 percent versus 7 percent in pramlintide compared to placebo.

Nausea was also reason for first-month withdrawals during the Type 2 diabetes trials albeit to a lower extent.

This slide displays the frequency of nausea-related withdrawals during the first month of the Phase 3 Type 2 diabetes trials. The Y axis represents patients who withdrew due to nausea in the first month. It should be noticed that the values on the Y axis are smaller than the previous slides shown for the Type 1 diabetes patients, 1.6 percent versus 6 percent.

The yellow bars represent pramlintide treatment. The blue bars represent placebo treatment. The data are presented for individual studies and all studies combined.

Overall, a two-fold difference between pramlintide and placebo in nausea-related withdrawals is present. First month nausea-related withdrawals were four time less frequently than during the Type 1 diabetes trials.

With this general information in mind about the Phase 3 trials, I would like to move on to discuss one of the aspects of hypoglycemia and assisted hypoglycemia.

Assisted hypoglycemia has been defined as any episode of hypoglycemia requiring the help of another individual for treatment be it oral carbohydrates, glucagon injection, or intravenous glucose.

This definition of hypoglycemia has been applied consistently in all three Type 1 diabetes trials and in two of the three Type 2 diabetes trials. Analyses of the time of occurrence of assisted hypoglycemia during the Phase 3 trials has been presented to the agency stratified by two time intervals. Assisted hypoglycemia occurring during the first month of the trial and assisted hypoglycemia following the end of the first month after the completion of the trial.

Shown here is the incidence of subjects who experienced assisted hypoglycemia during the first month of pramlintide treatment in Type 1 diabetes. Y axis represents patients with at least one episode of hypoglycemia. Yellow bars represent pramlintide patients, blue bars placebo patients, the data presented for individual studies and for all studies combined.

One can observe consistently the two-fold difference between pramlintide and placebo. A different picture emerges after the first month.

This slide depicts the incidence of assisted hypoglycemia after the end of the first month and after the completion of the trial. Again, the Y axis is subject patients with at least one episode of assisted hypoglycemia. The yellow bar represent pramlintide patients. The blue bar represents placebo patients. Data are presented for individual studies and for all studies combined.

It should be noticed that the Y carrier values are different than those shown in the previous slide. Therefore, I included the pramlintide to placebo comparison observed during the first month of the trial in pale colors, pramlintide and placebo.

In contrast to the first month of treatment, the difference in incidence between pramlintide and placebo is less obvious but still present. It should be noticed that the incidence is cumulative for 11 months in the studies 121 and 112 and for five months in study 117.

When we contemplate this particular slide, we need to keep in mind two things. First of all, the patient population observed here is not the patient population which started the trial. As shown previously, the first month of pramlintide treatment is associated with 17-fold high withdrawal ratio in the pramlintide group.

Therefore, subjects sensitive to the drug discontinued early. The second point I would like to make is that the occurrence of hypoglycemia has to be looked at in the context of efficacy.

Shown here are the HbA1c changes during a representative Type 1 diabetes trial study 137-121. Y axis represents mean changes in HbA1c from baseline. X axis represents time within the trial. The top line represents HbA1c in the placebo arms. The bottom line represents HbA1c during the different pramlintide dosage arms.

The two-fold interest in assisted hypoglycemia noticed in the first month of the trial are associated with a drop in HbA1c. The differences in assisted hypoglycemia between the end of the first month of the trial and the end of the trial happen in a context of waning HbA1c reduction.

In summary, pramlintide therapy is associated with a two-fold increase in incidence of assisted hypoglycemia when compared to placebo during the first month of the treatment.

The difference in incidence of assisted hypoglycemia between pramlintide and placebo groups persist following the first month of the treatment, albeit to a lower extent.

The decreasing hypoglycemic events associated with pramlintide takes place in the context of prior nausea related patient withdrawals and waning drug efficacy.

Some similarities to the Type 1 diabetes trials are present within Type 2 diabetes Phase 3 studies. Shown here is the incidence of subjects who experienced assisted hypoglycemia during the first month of pramlintide treatment in Type 2 diabetes.

Y axis represent percent of patients with at least one episode of hypoglycemia. Yellow bars represent pramlintide. Blue bars represent placebo. Data represented for individual studies and for all studies combined.

The first month of treatment is associated with a three-fold increase in assisted hypoglycemia in the pramlintide group over placebo. It should be noticed that the incidence is lower compared to the Type 1 diabetes trials. Only 3 percent compared to 13 percent.

Shown here is the incidence of assisted hypoglycemia in the Type 2 diabetes trials following the first month of treatment. Again, the Y axis are subjects with at least one episode of hypoglycemia. Yellow bar is pramlintide patients. Blue bar is placebo patients.

It should be noticed that the Y scale value is different than the one shown in the previous slide. Therefore, I included the pramlintide to placebo comparison as observed in the first month of the trial in pale colors. This is pramlintide and this is placebo.

When one observes assisted hypoglycemia for the rest of the trial duration in Type 2 diabetes the differences in assisted hypoglycemia incidence are minimal and inconsistent between trials. The one-year study, study 137-122, shows almost no difference. The six-month study 137-123 shows approximately two-fold difference. Overall there is a small difference.

One should keep in mind that the incidence is cumulative for 11 months for study 137-122 and five months for study 123.

We also need to keep in mind that the first month nausea withdrawals were higher for the pramlintide group and also interpret the hypoglycemia in the context of decreased efficacy. With time it is shown previously by Dr. Misbin.

In summary, pramlintide therapy is associated with a three-fold increase in incidence of assisted hypoglycemia when compared to placebo during the first month of treatment.

The difference in incidence of assisted hypoglycemia between pramlintide and placebo groups wanes following the first month of treatment. This decrease takes place in the context of prior nausea related patient withdrawals, although to a lower extent than in Type 1 diabetes.

Finally, we should remember that the incidence of assisted hypoglycemia was higher than the Type 1 and Type 2 diabetes trials when compared to controls during both time intervals analyzed. Hypoglycemia associated with pramlintide was more common in Type 1 diabetes patients.

I would like to move on and discuss serious adverse events, or SAEs, associated with hypoglycemia during Type 1 diabetes. SAEs are defined as adverse events that result in death or life-threatening, result in hospitalization, or disability.

We believe there were two deaths which may have been due to hypoglycemia. The first one, which was presented by Dr. Kolterman also, is a 48-year-old male with a 12-year history of Type 1 diabetes mellitus with a prior history of diabetes related seizures who died during a hypoglycemia seizure. The patient was receiving pramlintide 30 micrograms qid.

The second patient was a 35-year-old male with a six-year history of Type 1 diabetes mellitus who died in a motor vehicle accident within 24 hours from the beginning of the trial. Food was present in the stomach at the post-mortem examination indicating that the subject had eaten lunch prior to the event. The patient was receiving pramlintide 90 micrograms TID.

We feel that the presence of food in the stomach, and Dr. Levitsky has suggested also previously, is a fact of significant concern because it is one of the major effects of pramlintide such as delayed gastric emptying.

In addition to these two patients, there was a 31-year-old patient with not other past medical history except for diabetes who died and who was also in the pramlintide group. Although this death was hypothesized to have been due to an alcohol problem, the evidence was inconsistent and circumstantial.

During the review process the observation was made that SAEs associated with hypoglycemia occurred two to three times more frequently during the pramlintide treatment over placebo in the Type 1 diabetes trials.

In search for a possible explanation, the patient's narratives were reviewed in detail. As a reminder, patient narratives are a brief description of the events and circumstances which led to a subject's withdrawal from the study, the subject's death, or a left-threatening event.

A total of 20 motor vehicle accidents and other driving-related events occurred in conjunction with hypoglycemia were thus identified along with several other injuries.

Therefore, the agency has requested an analysis of all MVAs and trauma occurring in association with hypoglycemia during the pramlintide clinical program.

These events have been presented to us in the following categories. The first category is MVA related events. This is a self-explanatory category and involves the patient behind the wheel loosing control of the car and sustaining a collision.

The second category is automobile related adverse events with no motor vehicle accident reported. As this title suggest, this is a less specific category and it may involve a patient behind the wheel who is able to stop in time before a collision occurs, or an event taking place in the car in the parking lot or an event occurring about the time the patient is entering or leaving the car.

The third category is other accidents and injury-related events. This category includes a whole range of non-MVA related trauma such as falls, fractures, and lacerations.

Shown here are the MVA and automobile-related adverse events presented for the entire pramlintide clinical program. The MVAs are presented as total numbers and MVAs association with hypoglycemia. The automobile-related events were all presented as associated with hypoglycemia.

Before going any further, it should be noted that these events have been collected from the entire clinical program including, for instance, patients in uncontrolled studies and pre-Phase 1 studies.

Out of the 2,573 patients in the pramlintide group, 1.8 percent experienced MVAs. Out of the 904 patients in the placebo group, .66 percent experienced an MVA. There was a slight predominance of 1.4 approximately pramlintide to placebo for total MVAs.

.66 percent of patients in the pramlintide group had an MVA associated with hypoglycemia. Only .22 percent of the placebo groups had an MVA associated with hypoglycemia, roughly a three-fold difference. All automobile-related hypoglycemic events occurred in the pramlintide group, eight versus zero.

We feel that both categories represent in essence different facets of the same range of events of hypoglycemia in the context of driving activity and they should be analyzed together. This would give a completely different percentage.

I would also like to point out that a balanced analysis of these events should involve the pramlintide to placebo comparison limited to the Phase 3 controlled studies.

Such an analysis is presented in this slide as driving-related events associated with hypoglycemia in the Phase 3, Type 1 diabetes trials. Out of 1,179 patients in the pramlintide group, 1.53 experienced such an event in association with hypoglycemia. Out of the 538 patients in the placebo group, .37 percent experienced such an event.

It should be kept in mind that these driving-related events have not been actively ascertained during any of the clinical trials and they may be grossly underestimated.

In summary, pramlintide use in addition to insulin is associated with a four-fold increase in driving related events in Type 1 diabetes patients.

Shown here is a summary of the non-MVA or other accidents and injuries which occurred during the Type 1 diabetes pramlintide clinical program. Again, it should be noted that these events have been collected from the entire clinical program including uncontrolled studies and pre-Phase 3 studies.

Out of the 2,570 patients in the pramlintide group, 7.65 percent had such events. Out of 904 patients in the placebo group, 5.86 percent have such events. There was a predominance of trauma in the pramlintide group.

The patients who had trauma associated with hypoglycemia showed a percentage of .39 percent versus .2 percent, also slightly higher. However, I would like to point out that we have identified at least two more patients in the database that fit these category. In study 137-121 patient 5030 sustained a fall associated with hypoglycemia and a facial laceration.

In study 137-117 patient 6303 fell from a tree during a hypoglycemia episode, sustained a broken elbow, and required surgical intervention. Therefore, we believe that this number should actually be 12.

In summary, pramlintide use in addition to insulin is associated with a four-fold increase in non-MVA injuries in Type 1 diabetes patients.

I would like to finish this presentation with two labeling safety questions. First of all, do we understand how to initiate safely pramlintide treatment and avoid the risk of first month hypoglycemia in both patients with Type 1 and Type 2 diabetes.

In the briefing document, it has been suggested to initiate treatment with 30 micrograms or 60 micrograms of pramlintide per dose in Type 1 diabetic patients and 120 micrograms per dose in Type 2 diabetic patients. Today we heard the suggestion to go even lower than that.

All of the above doses have been shown to be associated with approximately two-fold increased risk of hypoglycemia during the Phase 3 trials.

It has also been suggested, and I'm quoting from the briefing document, that in clinical practice it will be prudent to reduce the patient's insulin dose, particularly the short-acting insulin administered postprandially by 10 to 20 percent at the time of initiation of pramlintide therapy.

While such an approach seems prudent, it has not been tested in a clinical trial and its potential usefulness remains unknown.

Finally, we have to consider this question. How can one prevent the four-fold risk of driving-related events and the four-fold risk of non-MVA injuries associated with hypoglycemia observed in the Type 2 diabetes trials. This is a very difficult question. This is illustrated by this slide.

This is time of driving related events versus HbA1c levels and the X axis represents time within the trial. This is a distribution of all driving-related events associated with hypoglycemia during the Phase 3 Type 1 diabetes trials. In yellow are the pramlintide patients who have been involved in MVAs and in red the placebo patient.

The blue horizontal line represents the hemoglobin 7 or the target hemoglobin for treatment suggested by the American Academy of Pediatrics. The vertical orange line represents the first month of the treatment.

As you can see, there is a lot of cloistering of MVAs occurring during the first month of treatment, approximately one-third of them.

On the other hand, two thirds of all MVAs and MVA related events which occur in the pramlintide group take place during the whole clinical trial. The prediction of which subject will experience such an event, to me, appears impossible at this time based on the information we have.

As a final summary, pramlintide therapy results in the small but statistically significant reduction HbA1c. This reduction is associated with a two-fold increase of severe hypoglycemia during the first month of treatment.

In addition, a four-fold increase in hypoglycemia associate MVAs and non-MVA trauma was observed in patients in Type 1 diabetes.

DR. KREISBERG: Dr. Hobberman will present one overhead transparency also on behalf of the FDA.

MR. HOBBERMAN: I have a couple of transparencies, but I also have three comments about the sponsor's presentation. Following Dr. Roman's talk, I wanted to comment on the sponsor's use of the p-value of .13 to describe the result of an analysis comparing the groups with respect to motor vehicle accidents.

We don't think that the use of a p-value is necessarily a good thing in the light of a very serious safety problem. This was not a planned analysis. The purpose of the analysis seems to be to have minimized the concern with respect to motor vehicle accidents to say that it was not a statistically significant difference.

.05 is not a magic number. It's used in efficacy analyses in protocol specified analyses and clinical trials but we don't think that it's a proper interpretation of the data. I think that Dr. Roman has presented clear evidence that this is a signal that needs to be taken very seriously.

My second comment has to do with the use of person time in multiple events when the sponsor reported rates of severe hypoglycemia. I don't think that was the best way to present the data for two reasons. One is that it is subject to the problem of multiple events on a patient and, therefore, can confound the incidence with the total event rate by counting multiple events per person.

That may be useful in trials or situations in epidemiology where there is gross inequality of follow-up time. In clinical trials that are well-controlled with a very defined follow-up period, it really shouldn't be necessary.

The other reason it's not optimum is that the statistical analysis of multiple event data is very complicated so it's hard to even generate a p-value for that. I think the best way to look at this data is the sponsor did report it in the application which was true incidence timed to first experience of severe hypoglycemia which had plenty of statistically significant results, if you want to work in that realm, in a couple of trials.

Also indicated that on the Kaplan-Meier curves they did supply the incidence in their either to three-year -- I'm sorry, half to one-year trials was in the order of 10 to 15 percent on placebo insulin and up to 20 percent or over on pramlintide. That comes from just looking at their Kaplan-Meier curves which they submitted with the MVA.

The third point has to do with the overheads that I have. This relates to the issue of the target of 7 percent points of HbA1c. What I did was I took the sponsor's data for those patients who essentially completed the trial.

One of the things that I found a little confusing is when the sponsor talked about the number of responders, it wasn't clear to me that this was an ITT data set, i.e., with last observation carried forward, or whether these were numbers that were derived from patients who had actually completed the prescribed trial time.

I think it's really meaningful only to take those patients who completed the trial when you're talking about what happens after 26 months or 52 months -- 52 weeks.

This overhead is the sponsor's completer data for trial 117 in Type 1 diabetes. I'm trying to illustrate three different values. On the vertical axis we have the change from baseline in HbA1c. On the horizontal axis we have the baseline HbA1c.

For the third dimension what I've drawn is these lines. These are not regression lines. These are just plain lines and what they correspond to is constant contours of where a patient ended up, their final HbA1c value.

For instance, if you started out at 7 and you did not change at all, that's that point there. So anybody on this line for any patient you can look at their baseline value. You can go up to the symbol, go across here for the change from baseline, and then simply see where they ended up between these lines. This is 7, 8, 9, and then I didn't bother with lines there.

What this indicates essentially is that as we all know, the baseline values were somewhere between an average of 8 and 9. Here we have the bulk of changes from baseline. These squares, by the way, are the drug. The crosses, or the pluses, are the placebo. You can see there are very few observations of people who fell below a target of 7 in this trial.

This isn't labeled but this is the trial 112. Here we have more drug people on pramlintide than in the previous trial who did fall below 7 at the end of the trial. As you can see, the mass of the data, the density of the data, is anywhere between 7 and 10 if you draw that line.

The import of this partly is getting back to my question about whether the sponsor used an intent-to-treat analysis which carried last observation forward or used completers.

Based on what I found with their completion data, I recall that their slide said 14 percent of the patients in Type 1 diabetes reached a goal of .7. I'm not sure whether the data that I've just shown is consistent with that, or whether they were presenting an over estimate. I'm sure that can be clarified later. Thank you.

DR. KREISBERG: Dr. Cara.

DR. CARA: There was an allusion earlier to the fact that there were two studies in which the insulin dose was allowed to be varied according to standard medical practice. Was there any difference in the number of hypoglycemic events within those studies compared to those studies in which the insulin dose was not changed?

DR. KREISBERG: Does anybody from the FDA want to answer that question or anybody from the sponsor?

DR. MISBIN: This is our time but I think I would defer to the sponsor. It's our understanding in the Type 2 -- these were the early studies. In the patients with Type 2 diabetes, the data was really not captured in the systematic way so it was not actually even reported to us.

In the Type 1 study the incidence that was reported to us was lower than in the other two studies. The protocol for all the later studies state specifically that patients had a touch one meter and that they were told to keep records. There was a great description of the efforts made to capture that data.

In the earliest Phase I study, however, that description is really absent from the protocol so we have data but it seems to me the criteria used and the intensity of capturing the data is different. There was also a change made in the definition of serious hypoglycemia which was not exactly clear when we reviewed the data.

I'm not sure why there is that difference but, in fact, in that trial there is still more in the pramlintide treated patients. The enormous discrepancy we saw in the two later trials was not evident in the first trial.

DR. CARA: And in looking at the incidence of hypoglycemic reactions and other adverse events, did you specifically look at the intent to treat? Did you do an intent to treat analysis or did you look at completers?

DR. MISBIN: I don't understand how you to an intent-to-treat analysis for a safety evaluation. I mean, we scored all of the events that were reported to us and it was not differentiated.

The analysis that I showed you of the responder analysis, those, of course, were completers so that might answer your question but I don't think it was broken down otherwise by the two populations. I would point out that the completer analysis versus the ITT analysis are really very much the same across all the trials. The differences, in my judgement, are not really worth discussing from an efficacy point of view.

DR. CARA: The reason for my question was trying to capture those patients that had dropped out of the trial because of an adverse event such as hypoglycemia.

DR. KREISBERG: Dr. Kolterman, do you want to add anything to this discussion?

DR. KOLTERMAN: With the Chair's permission, I can show a slide that allows a comparison across the three Type 1 trials during the critical period of zero to four weeks if you think that would be helpful. Slide up, please. These are the data that was just presented by the agency from study 137-117 which included two doses that are not being recommended. These are the data from study 137-121. Then here is the data from the lower dose trial where patients were allowed -- where there was no recommendation given in terms of insulin treatment.

The hatched area here, I remind you, represents the presence of our century man who had in excess of 100 events recorded during the study.

DR. CARA: Do you have a similar slide showing the incidence of hypoglycemia rather than the rate?

DR. KOLTERMAN: We do have slides that show incidence. We need just a minute. I'm not sure that the incidence numbers will relate directly to the slide that I just showed you.

DR. ROMAN: If I can just make a point. It is my recollection that in the same placebo group, placebo arm which had the outlier with 120 over all events, there was another outlier who had about 42 or so events. There are really two outliers in that placebo arm. I do not know exactly how many of those events actually happened in the first month.

To go back to the previous question, if I understand it correctly, the question was if there was incidence of hypoglycemia different between the trials in which there was a 10 percent decrease versus no adjustment in insulin. If I remember correctly, it was 117.

Anyway, between study 117 and 121 actually the incidence of hypoglycemia during the first month was pretty much the same, 14 percent, as shown in one of my slides. There was a lower incidence of about 8.5 percent in study 137-112 which employed the lower pramlintide. There was a 30 micrograms of QID. I hope that helps clarify the question.

DR. CARA: Well, yes and no in the sense that the real critical issue is whether or not hypoglycemia is more frequent or less frequent depending on whether you can adjust the dose of insulin based on accepted standard practice.

That also relates to efficacy in the sense that if when you adjust the insulin, do you also see less efficacy of the drug. I would appreciate some comments from the sponsor on that.

DR. KREISBERG: Orville, do you have anymore data you can share with us?

DR. KOLTERMAN: Okay. I do not have the appropriate incidence data on a slide. We can provide that after lunch if the committee would like to see that. We obviously have the data. It's just pulling it together so that we can show it for the same groups.

DR. KREISBERG: There will be plenty of opportunity after lunch I'm sure.

Bill.

DR. TAMBORLANE: This is for Dr. Misbin. Several comments alluded to waning drug effect. My impression was that you have a placebo controlled arm to look at changes in study effect and that you do see changes in outcomes depending on the beginning and end of trial and the extra attention placed to the patient. I would suggest that what you've shown is a waning study effect and that the placebo subtracted difference did not wan.

DR. MISBIN: Your point is well taken actually. If you look at the difference from placebo at four weeks and endpoint, it was not very different. It was around -- in various trials it was around .3, .4, or whatever.

What seemed to change, however, was the statistical difference. At four weeks every arm, every trial across the board was highly significant. When one got to endpoint, however, sometimes it was and sometimes it wasn't significant. Your point, I think, is well taken.

What was being seen, I think, was that there is clearly a drug effect at the beginning but by the end of the trial, I would say just a random variation that you see in patients with diabetes, is fairly great and that, I think, was kind of overwhelming, the statistical difference that one saw between drug and placebo.

DR. GELATO: My question really goes back earlier and I think you were going to try to answer it.

I just wondered when we were shown the data about who reached target and who didn't in terms of the goals of the ADA in terms of 7 percent for glycated hemoglobin whether you could separate out where the insulin levels were able to be adjusted in those patients, in those trials, and where it was kept steady and whether there was a difference in terms of who reached target when you were allowed "to do what good clinical practice is," and that is continue to adjust their insulin.

DR. KOLTERMAN: We're checking to see if we have the targets by study.

DR. TAMBORLANE: I wonder if I could comment while we're waiting because, you know, ADA targets are targets and taking care of patients are a very different thing. I think there was a comment that I'm not sure if Dr. Kolterman made or not that looked at the EDIC results.

I happened to be the PI on the DCCT EDIC at Yale and if you look at the outcomes currently of the formerly intensively treated group in the DCCT who had years of intensive management and training, and now that these patients were returned to the community with, more often, community control when they return to their own clinics, the HbA1c has gone up to 8.2 percent. Setting a target is something that we shoot for. Reality is another thing.

DR. KOLTERMAN: Again, we have the data and we'll have it on an appropriate slide for you after lunch.

DR. GRADY: I would like to ask the FDA a related question. I think the hypothesis has been laid out that hypoglycemia is an early effect of the drug and that it may wan over time.

I wonder if the other possibility isn't that initially patients were on a somewhat higher dose of insulin. They developed hypoglycemia and their insulin was adjusted. Did you look at hypoglycemic episodes among patients who had increases in their insulin dose during the studies?

DR. MISBIN: The change in insulin was very minimal. I think I show that in the responder group. In the placebo patients who responded that were classified as responded, there was an increase of 1.9 percent which is one unit of mean change.

We didn't even discuss changes in insulin very much because they were very, very small. Nothing was statistically significant. They were just so small that I don't think that would add much to the analysis.

DR. GRADY: (Off microphone.)

DR. MISBIN: Oh, yes. There are some patients that do. We did not do a specific analysis on those patients that had a large change in insulin versus hypoglycemia.

I mean, that is certainly possible. The mean changes in insulin across all the placebo groups were very, very small, very small increases which was by design. In pramlintide, of course, because of safety issues there was a small mean reduction.

DR. SAMPSON: Dr. Roman, I would like to follow up just one of your slides in terms of the reduction of hypoglycemia after the first four weeks. You indicated this took place in the context of prior nausea related patient withdrawals. Was that statement -- one would almost draw an inference from that. Was there some data behind that that you were trying to project with that statement?

DR. ROMAN: What I was trying to refer to was the fact that the previous slide which showed that during the first month of pramlintide treatment if you look at the present of patients who due to the nausea in pramlintide groups compared to placebo groups, there was almost a 17-fold difference in all studies combined.

You lose about 7 percent of patients in study 121 to nausea in the first month and about 6.5 in study 117, and roughly about 5.2 in study 112. The sense I'm getting from that is that the initial structure and internal organization of the group is changing a little bit because you lose patients and you don't lose them to a nonspecific reason. You lose to a symptom that is directly related to the drug. Nausea, as you know and has been presented before, is the major adverse event in terms of frequency of pramlintide which results in quite wide discrepancies between the pramlintide and placebo group. I made the comment in the context that you lose some patients which are sensitive to the drug.

DR. SAMPSON: Was there any way of looking at hypoglycemia in these patients that withdrew early for nausea and comparing that to those that remained?

DR. ROMAN: We don't have that analysis. I do not believe I came across about occurrence of hypoglycemia in those subjects. I don't know if Dr. Kolterman has more information on that.

DR. KREISBERG: Can you answer that?

DR. KOLTERMAN: I can comment on two things. The point that is on the table now is if you look at patients that did not experience nausea, you do not see the increase in hypoglycemia. If you look amongst patients experiencing severe hypoglycemia, it's roughly a 50/50 split. Half experience nausea and half do not experience nausea.

With regards to the changes in insulin use patterns and the occurrence of hypoglycemia, I have a slide that focuses on patients who increased their total daily insulin dose by 10 percent or more during the conduct of the trial that has hypoglycemia on it. This issue about stable insulin, as I outlined in my presentation, the goal of that was to come up with a group of patients that limited the variability of insulin as a confounding factor in the HbA1c analysis.

It turns out that only at best 40 percent of patients were able to do that during the duration of one of these long-term controlled trials. That means that 60 percent of patients varied their insulin.

For patients in the type one program treated with the recommended dose, there are 122 patients that increase their insulin by more than -- no, this is increase. I'm sorry. We have the wrong slide. There is a similar slide. It's roughly the same number of patients who will have increased their insulin by more than 10 percent.

Sorry, I was misled. We do not have that slide. Again, I think we can have that for you after lunch.

DR. KREISBERG: What I would like to do now, unless there are any other compelling --

Dr. Levitsky.

DR. LEVITSKY: I've been trying for a while. My question for you, Dr. Misbin, would be in the FDA briefing document, and it was alluded to briefly here, there were some concerns over the correlation between the data summarized and supplied by the company and the data when you went in and reviewed them. Is this still a concern for you?

DR. MISBIN: Are you talking about the inspections?

DR. LEVITSKY: Yes.

DR. MISBIN: Yes, it is a concern. I don't know if we are supposed to discuss this.

DR. LEVITSKY: Can we talk about it? It was in the briefing document I got.

DR. MISBIN: We need to hear a judgement from the Chairman whether this can be discussed.

DR. KREISBERG: The Chairman? Let me consult with the FDA.

DR. TAMBORLANE: I have another informational question in the meantime.

DR. ORLOFF: Dr. Kreisberg, might I suggest that if this is going to come up, it could come up in the discussion after lunch. The sponsor, I believe, would be happy to address the integrity of their database at that time.

DR. KREISBERG: Okay.

DR. TAMBORLANE: I have another informational question about the hypoglycemia and accidents. I apologize because there were a lot of data in a short period of time. My impression was that you were throwing in the results from all of the exposures to pramlintide. Have you looked at the data with respect to the recommended doses as well? Did I miss that or did you present that?

DR. ROMAN: I looked at the doses and practically almost all patients were -- let me make two points. One point is that if you look at all the MVAs, as I said, they were not captured perspectively. That being said, some studies had more driving-related events than others. Of course, the data is going to be skewed toward those studies and those doses using those studies.

Now, to answer your question, study 121 all patients had 60 micrograms, 90 micrograms, 60, 60, 60. It seemed to be occurring maybe in the 60 range and one in the 90 range. In study 117 they had the 90 microgram range of pramlintide.

In study 112 they occurred as low as 30 but there were only two of them which were captured in that study. In study 121 it's really not a clear trend that higher dose is associated with events to me. It's very limited data, though.

DR. TAMBORLANE: Could the company come up with data related to that? You showed some issues with that at the lower doses. At least the hypoglycemia rate seemed less at what you're recommending. Do you have other safety issues?

DR. KOLTERMAN: Yes. I have a slide that will address this issue so we have the slide up. This is a listing of the motor vehicle accidents associated with hypoglycemia.

Now, associated with hypoglycemia means that hypoglycemia was recorded by the patient on the same day within the same 24 hour period of the motor vehicle accident. You can see that the accidents that occurred during the initial four weeks by our analysis of the data occurred only at the higher doses.

The same is true of weeks four to 12 in that the accidents that did occur on the lower dose of 30 micrograms four times a day occurred after 12 weeks of therapy. These accidents here by 112(e) accidents are accidents that occurred into the second and sometimes the third year of exposure to the compound.

If you limit the 30 microgram analysis to just those that occurred during the double blind placebo control period, you have two events right here.

DR. MISBIN: It should be pointed out, I think, that 30 micrograms was only used in one trial. Isn't that right?

DR. KOLTERMAN: That is correct. It was used in one placebo controlled trial.

DR. MISBIN: Of the Phase 3 trials that we are considering, there was one trial in which it was given 30 micrograms QID. The other two trials we have no data on 30 microgram dosing.

DR. KOLTERMAN: That's true but as was on the slide, the data related to 30 micrograms is supplemented by data from two open-label safety studies that provide extra additional exposure at that dose.

DR. KREISBERG: I'm going to cut off this type of questioning at the present time. We should have enough time this afternoon to cover this and all other questions for discussion. We are approximately 45 minutes behind schedule and I would like to move to the open public hearing.

I would like to ask each of the speakers to come to the microphone at the front of the center isle and disclose any financial conflicts that you might have and limit your comments to three minutes.

DR. LEVETAN: Good morning. I am Claresa Levetan. I am a endocrinologist who practices in Washington, D.C., at MedStar Clinical Research Center, which is affiliated with Georgetown Medical School. I have been a clinical investigator and a consultant for Amylin Pharmaceuticals since 1995.

I am here this morning for one reason and one reason only. My patients have told me that pramlintide has given them back their life. In Type 1 patients, in Type 2 patients, in patients on pumps, in patients with A1c of 7 percent, and in patients with A1c of 11 percent.

I have had 60 patients receive pramlintide through the clinical trials. My patients travel from as far as London to continue their participation in the open-label pramlintide trial.

The comment I hear most consistently from my patients is that pramlintide reduces the glucose fluctuations and swings. Yes, both Type 1 patients and Type 2 patients did experience hypoglycemia during the initial stages of some of the randomized trials, but this only occurred in my patients during the time period when the protocol mandated that their insulin dosages go unchanged.

I found that many of my patients actually had sizable reductions in their insulin dosages by study end. During the open-label trial, unlike the randomized trials, we reduced insulin dosages at the time pramlintide was initiated, and avoided all of the serious hypoglycemia previously seen during the first weeks of therapy during the randomized trials.

In my practice, I use the continuous glucose monitoring system which measures interstitial glucose every five minutes and records 288 glucose readings per day.

To further evaluate the potential benefits of pramlintide that have been described by my patients, I utilized the continuous glucose monitoring on a patient who entered the open-label trial shortly after this monitoring system became available.

I have attached the sensor data of this patient who is a Type 1 patient who was one of the patients who fortunately did achieve an A1c of 6.6 percent which was the goal, below 7 percent as was mentioned this morning. This patient does also use an insulin pump.

Despite the A1c of 6.6 percent, he experienced wide swings in glucose fluctuations and hypoglycemia. By day 27 on pramlintide his insulin requirement before meals was reduced by 18 percent and he had a 54 percent reduction in both the high and low glucose excursions from the mean compared to baseline levels before pramlintide.

Currently the most serious and potentially life-threatening limitation to patients with diabetes is hypoglycemia. I believe that amylin plays its greatest role in ameliorating both hypoglycemia and hyperglycemia via different mechanisms of action that insulin as evidenced by the different structure and locations of the amylin receptors compared to that of the insulin receptors.

Far beyond pramlintide's ability to lower HbA1c, I believe that this hormone plays its greatest role as the fine tuner of glucose regulation. Pramlintide's benefits include the inhibition of glucagon which reduces both postprandial hyperglycemia and enhances the liver's stores of glycogen which is reduced in patients with diabetes.

Thus, pramlintide's benefit may result from diminishing both the high and the low swings seen among patients with tightly -- that are seen even in tightly controlled diabetes patients.

These benefits would not be evident by looking at average blood glucose values, nor reflected in HbA1c. The DCCT demonstrated the benefit of reducing high and low glucose excursions independently of a change in HbA1c on reductions in diabetes related complications and I have attached that data.

In summary, as a clinician whose practice focuses on patients with brittle diabetes, I strongly urge this panel to approve the usage of this hormone in patients with Type 1 and Type 2 diabetes because pramlintide is the first and only adjunctive therapy which addresses many of the shortcomings in insulin therapy today.

In my patients, pramlintide resulted in sustained reductions in the hypoglycemic event rate after insulin dosages were reduced beginning at four weeks.

Secondly, it resulted in a sustained reduction in hyperglycemia. It also resulted in a sustained reduction in insulin requirements. And also in enhanced glucose control without weight gain and a sustained weight loss that occurred after the reduction of insulin.

I request that all of the data that was presented this morning and hypoglycemic event rates be analyzed after four weeks of therapy. My strong feeling is that you will see the clinically important endpoints which my patients have seen. I thank you very much for your time and consideration.

DR. KREISBERG: Thank you.

Can we have the next speaker, please.

DR. WUERTENBERG: Good afternoon.

DR. KREISBERG: Good afternoon. You have three minutes.

DR. WUERTENBERG: Okay. My name is Anna Wuertenberg and I am a patient of Dr. Levetan's and a participant in the current open-label trial on pramlintide acetate, trial number 137-140, at the MedStar Research Center.

I not only didn't withdraw from the study, I went out about a week after I joined it and bought 100 shares of stock. I very much appreciate having this opportunity to speak with the committee about my experience with pramlintide.

I have been a Type 1 diabetic for 26 years and for 26 years I have been unable to get control of my diabetes.

I won't give you all the gory details but I will tell you that except for the new "basal" insulins, I've tried every insulin and every combination of insulins on the market. I've tried them up to five shots a day. I've used a pump since 1993.

I test regularly. I actually test religiously. I exercise almost as religiously. I follow my meal plan. Five years ago I had to start working at home four days a week because I could no longer go into work.

Despite all my efforts, despite all the efforts of a number of caring and competent doctors and certified diabetes educators, things have gotten worse, not better and I have been unable to control my diabetes.

I'm very emotional because hearing what I've heard in the last hour as made me very angry because it just hasn't been my experience with pramlintide. Even with the pump before pramlintide I had to take a comparatively large amount of insulin with every meal. I had to use very low basal rates for the rest of the time. This was totally ineffective and I had lots of severe hypoglycemia, especially at night. I was never able to find a more effective balance. I just couldn't get there.

I can't overstate how debilitating this was. By the time I began the trial, my blood sugars still ranged within a range of about 300 points every day with the most severe lows at night. I was unable to exercise. I was working only an hour to two hours per day and sitting in sort of a semi-static sense for the rest of the day. This was my last ditch effort before disability retirement.

Since starting pramlintide at the end of January 2001, my blood sugars have grown progressively more stable. I currently take an injection of pramlintide three times a day. Doing this on top of using a pump is kind of a bore and it's not always convenient to take a shot and then eat immediately afterwards. Let me tell you, it's worth the trouble.

My last A1c is 6.4 but I want to reemphasize what Dr. Levetan said. It's not just the score, it's the fact that it's happening in a range that is much smaller. Before it was truly an average of the very high and the very low.

My insulin dosages at meal have been reduced a little more than Dr. Levetan indicated. My dinner insulin dose has gone down 70 percent. My bolis at breakfast is 30 percent and it's somewhere in between for lunch. I've had very good luck that way.

I've also been able to reduce my overall dose of insulin despite being able to raise my basals a little bit without having hypoglycemia. I am now able to sleep through the night.

I've increased my workload and my turnaround times at work so dramatically that I'm now working a 50-hour week. If you want to include a warning on this drug, tell them it may make you too productive for your own good. I get lectured about it now when I go in.

My life on pramlintide isn't perfect but I'm getting closer to achieving the holy grail of control all the time. I now have several good days in a row and the bad days are fewer and fewer and they are much less bad than they ever were before.

I can't believe how much better I feel. I not only encourage you to approve pramlintide, I beg you to do it. This could change a lot of lives and we diabetics have been waiting since Banting and Best for something that will.

DR. KREISBERG: Thank you very much.

Can we have the next speaker, please.

MS. BENESH: Thank you for allowing me to have the opportunity today to speak in behalf of Symlin genericly known as pramlintide produced by Amylin.

My name is Susan Benesh. I am 53 years old, and I have managed my Type 1 juvenile diabetes mellitus for 38 years. This chronic illness was very traumatic for me and affected my entire family. It meant a complete change in lifestyle for all of us.

Although insulin was being used to treat diabetes at the onset of my diabetes, it was still tricky to keep glucose levels under control. Many factors such as diet, physical activity, stress, and hormonal changes make it difficult to control blood sugar levels.

Because of continued research and developments since my diagnosis, new products have come on the market to help control glucose levels. Research and development is essential to find a cure for this dreadful disease.

More and more people of all age groups are being diagnosed everyday with diabetes. In order to aid in glucose control and avoid daily multiple injections, I began an insulin pump user. When the pramlintide study was initiated and my physician approached me because she felt I was a good candidate for the study, I agreed to participate because I was excited about the possibility that this new drug would enhance and perhaps prolong my life.

The insulin pump has helped me considerably and I was more than willing to take the pramlintide injections because of the success I was experiencing in the pramlintide study.

Pramlintide not only helped with my glucose control, it lowered my cholesterol and enabled me to stabilize my weight. I participated in three pramlintide clinical trials.

The second trial conducted was terminated by Amylin Pharmecuticals because of funding issues. It took less than 30 days for me to notice a tremendous negative difference in my health without pramlintide. Although I was able to maintain my HbA1c below 7, I easily gained 20 pounds, my total cholesterol level increased, hypoglycemia became frequent again, and my HbA1c rose at each subsequent testing.

Prior to participating in the study, my HbA1c was in the double digit range despite my hard work to keep my glucose levels under the best of control. During the study my HbA1c was a low as 5.9 and not higher than 6.9. It was very disheartening and depressing to be denied use of pramlintide.

For a year following the termination of the study and much pleading by me and other study patients, my physician was able to obtain pramlintide once again for only a few study patients on an open-label trial basis. I was very, very fortunate to be one of those patients. I am presently using open-label pramlintide.

Pramlintide as been responsible for controlling my glucose levels, smoothing glucose levels to a point of experiencing much less frequent insulin reactions or high blood sugars, and aiding in the reduction of insulin intake for optimum control.

In closing, I am one diabetic patient among millions of diagnosed, and yet to be diagnosed, diabetics who has greatly benefitted from using pramlintide.

I hope that the information I have provided in addition to the overall study results of pramlintide will convince this advisory committee, as it has me, that even though I have to deal daily with diabetes, there is an opportunity that I will have a better quality of life with pramlintide. Thank you.

MS. ASHCRAFT: My name is Rose Ashcraft. I have been an attorney for about 20 years. I was diagnosed with diabetes at age 34 in September of 1987. I am a Type 1 diabetic. I participated in clinical trials for pramlintide at MedStar Research Institute. I am now taking pramlintide on a open-label trial.

I have heard several comments today about why pramlintide is not good enough. While I am here to tell you that many medical regimens for diabetes are not good enough, insulin is not good enough but it's the best that we have and we're glad that we have it.

I have two major problems with managing my blood glucose levels and they are reflected in my HbA1cs. The first problem is the dawn phenomenon. None of the insulins that we currently have on the market are adequate with the dawn phenomenon.

I began insulin pump therapy in January of this year in order to attempt to get a handle on that problem. The second problem I have with my blood sugar management is after-meal spikes and hypoglycemia. Actually, there are three problems.

The pramlintide has made a very significant impact on my after-meal spikes and upon my hypoglycemic incidents.

Pramlintide has effectively eliminated after-meal spikes in my blood glucose levels. Without the drug, during the first hour and a half subsequent to meals even with Lispro and the insulin pump, my blood sugars tend to rise rapidly and do not return to normal levels until the Lispro is fully absorbed approximately two and a half to three hours after the meal.

Use of pramlintide with my insulin regimen eliminates the after-meal spike in blood sugars that was consistent for me before I began using the pramlintide.

Secondly, I no longer have severe hypoglycemia, and I even have fewer mild hypoglycemic incidents. Despite my best attempts to use insulin to the best of my ability prior to the combined use of pramlintide with insulin, I routinely experienced hypoglycemia.

Prevention of hypoglycemia enables me to regularly consume fewer calories which, in turn, helps present unwanted weight gain. Maintaining an ideal weight is important to me for two reasons. First, gaining weight because I am diabetic is a real downer and it does not help me in my overall control. Secondly, being overweight puts me at risk for long-term complications.

When I went off pramlintide after the open-label extension ended in 1998, I gained 10 pounds in less than a year but lost it when I began consistently taking the medication again. I attributed this weight loss in large part to the fact that I was not eating extra carbohydrate.

I have continued with pramlintide even with my insulin pump therapy because of the benefits it provides me. The combination of pump therapy with Lispro which ensures a low-fasting blood sugar in the morning and increases flexibility in my routine with pramlintide which evens out mealtime readings and prevents hypoglycemia are giving me freedom and security that I have not had since before I was diagnosed.

It is my hope that pramlintide will be approved by the FDA because I think it's benefits are important to all diabetics who want to maintain good glucose control and live a life free of long-term complications. Thank you.

MR. BROWN: Good afternoon. My name is Chris Brown. I'm a Type 1 diabetic, a disease that runs in my family. Since being diagnosed in 1996 I have given myself at least four daily injections of insulin. Since June of 1998 when I enrolled in a Phase 3 clinical trial and afterwards on a compassionate use basis, I also inject 60 micrograms of pramlintide at mealtime.

The result is something I never thought possible. My blood sugar control is very nearly perfect. In fact, I now think of myself as a fundamentally healthy person who just happens to be a diabetic.

Unless I specifically mentioned it, nothing about my health would give a doctor reason to believe I am a diabetic. I owe this entirely to pramlintide. Assuming pramlintide is approved, I do not expect ever to develop the complications associated with diabetics.

Beyond balancing my insulin with my carbohydrate consumption, I pay my diet no special attention. But what's most remarkable is that with pramlintide, my diabetes requires almost no effort to manage.

Of course, tight blood sugar control is possible without pramlintide but non-diabetics can scarcely comprehend how much effort is required. Taking insulin together with pramlintide means I simply stop worrying about my blood sugar readings. Although I continue to test myself four time a day, I know expect my readings to fall well within a normal range. With daily injections of pramlintide, I am as surprised by abnormally high or low blood sugar readings as a non-diabetic would be.

I go weeks at a time with pre-breakfast blood sugar readings in the low 90s. Two hours after a meal my blood sugars are anywhere from 110 to 135. I have achieved these results while reducing my insulin requirements to about one-half to two-thirds of what they were before using the pramlintide.

I have a much easier time keep my weight down. IN fact, within two months of starting the pramlintide in the summer of 1998, I lost thirteen pounds that I gained trying to maintain tight blood sugar control without it. My HbA1c readings now hover around 5.8.

Pramlintide is as close to a miracle cure for diabetes as I can ever hope to see in my lifetime. I know that, strictly speaking, it is not a cure but I consider myself if not perfectly normal, then at least perfectly healthy again. With pramlintide I no longer think of myself as suffering from a chronic disease.

Of course, I do. Pramlintide has not altered the basic fact that my pancreas does not work. Nevertheless, pramlintide has changed how I think about myself. I am once again a healthy person who just happens to have a metabolic disorder.

Pramlintide has improved my life to a degree non-diabetics cannot begin to comprehend. To me, it represents the biggest improvement in the lives of diabetics since the development of human insulin some twenty years ago.

Pramlintide needs to become a routine element in the treatment of diabetes. Insulin to keep us alive, pramlintide to stave off the diabetic-related complications. The two together should go a long way to making this disease less of a killer.

I cannot imagine ever going back to worrying about my blood sugars. Pramlintide allows me to lead a normal life. I hope that eventually all diabetics will be similarly fortunate. For that reason, I ask that you recommend its approval. Thank you.

DR. KREISBERG: Thank you.

MS. WANT: Good afternoon. My name is Laura Want. I'm a certified diabetes educator and certified clinical research coordinator at MedStar Research Institute in Washington. I have over twenty years experience in diabetes education, management, and research. I have served as a coordinator for Amylin Pharmaceutical trials at MedStar since 1995. I'm a member of their CDE Advisory Committee and I own some Amylin stock.

The DCCT results proved the benefits of tight control to the point that many of our patients will endure severe hypoglycemia in the desperate quest to avoid the long-term complications.

The limitations of diabetes management continue to frustrate physicians and educators as well as patients. No matter how hard we try, it seems impossible to maintain near-normal HbA1c without the complications of weight gain and the frequent hypoglycemia.

Like the four previous speakers whose A1cs have remained below 7 since entering the open-label trial, pramlintide has allowed our patients to maintain type control with these miserable side effects.

Over the past six years I've coordinated Phase 3 pramlintide trials with 60 patients with diabetes. Before the studies most of these patients paid meticulous attention to balancing diet, exercise, and medication to keep the HbA1c as low as possible but struggling with weight gain and suffering frequent hypoglycemia.

Because the earlier trials called for fixed pramlintide and insulin doses, patients reported more nausea and hypoglycemia, especially in the first months of pramlintide therapy. The current open-label trial has allowed us more flexibility. We have found that lowering insulin doses and titrating up to the recommended pramlintide doses has virtually eliminated the problems with hypoglycemia and nausea.

Patients on pramlintide have had improvement in HbA1c levels but that does not fully reflect the benefits of pramlintide. Patients have had less frequent and less severe hypoglycemia.

Self-monitoring records showed reduced postprandial glucose rises as well as lower standard deviation from mean glucose indicating more consistent blood glucoses.

Patients reported increased well-being and energy. A woman whose pre-pramlintide severe hypoglycemia interfered with her daily activities reported she had fewer and milder hypoglycemic episodes on pramlintide.

Patients found it was easier to lose weight without compromising tight diabetes control. Even though the trial requires additional injections, patients felt that the pramlintide therapy dramatically improved their diabetes control and quality of life.

According to our patients, insulin made their diabetes survivable but adding pramlintide makes diabetes much more livable. Thank you.

DR. WOLFE: I'm Sidney Wolfe. I'm an internist and director of Public Citizen's Health Research Group. About 40 years ago it was possible to approve a drug based on testimonials by physicians, by patients such as the ones you've heard. I have no reason to doubt anything anyone has said but a little less than four years ago the law was changed to require the results of randomized control trials. You hear the good news stories. Again, I have no reason to doubt them but you don't hear people who have had some of the serious problems that have been described in the clinical trials coming forth.

As endocrinologists or primary care physicians based on results of the randomized control trials, which is what we are really here to consider, what we recommend for our insulin requiring diabetic patients a drug that had the following benefits and risks, benefits as pointed out compared with placebo lowering of HbA1c of only .3 percent in the four fixed-dose studies.

Increased severe hypoglycemia with automobile driving related adverse offense which if this drug were ever approved be in the thousands if not more as opposed to the dozens that were reported in these trials including crashes and confusion while driving. Most people required paramedic intervention, E.R. visits, and IV glucose administration and one death, as pointed out.

Something not discussed, although in the data provided by the FDA, 11 out of 1,179 patients given pramlintide but none out of 538 given the placebo had nervous system problems. Of these 11 four of the patients had convulsions, three had coma, and one each with ataxia headache, vertigo, and migraines. You've heard discussion of the gastro intestinal problems in the Type 1 diabetics. It was 51 percent nausea in the pramlintide group versus 17 percent in the placebo. Anorexia, again not discussed very much, a serious problem which can contribute to the hypoglycemia, 18 percent in the pramlintide group, and 2 percent in the placebo group.

Although the company seemed to want to trivialize it, there was an increase in diabetic retinopathy in one of the studies, an increase of 19 percent in the group taking the drug, and 8 percent in the group taking the placebo.

The need for more injections goes without saying. It is an additional burden on people. If the answer for our own patients about recommending the drug is no because the risk so clearly outweigh the benefits, the answer to the question about whether the FDA should approve the drug must also be no.

Beyond the question of FDA approval, however, is the issue of any further clinical trials involving new patients such as those proposed by the FDA based on existing knowledge about serious risk caused by pramlintide.

It would be unethical to do a study to pin down more firmly the causal relationship of this drug to hypoglycemia on awareness to further study efficacy or to expose new patients for any other purpose.

One can only imagine what the informed consent sheets for such studies would now have to look like. This drug deserves to be put out of its misery before any more patients are injured or killed in any further clinical trials.

DR. PULLMAN: My name is John Pullman. I've come from Butte, Montana, where I practice as a general internist and I have practiced there for 19 years. I have several hundred patients that are insulin controlled diabetics and I've been an investigator in the Amylin 102, 111, 111(e), 121, 137, 140 extension open-label trials.

DR. KREISBERG: We're having a little trouble hearing you. Could you just speak a little bit closer to the mic, please.

DR. PULLMAN: Surely. My reason for coming here today is to transmit to you the enthusiasm, the 30 or more patients that I have enrolled in these trials have transmitted to me, and the enthusiasm I've felt in my ability to improve their glycemic control.

The open-label trial has probably been the best form since we've been blinded on the others. 17 patients have elected to continue in the open-label trial. Their experience has been remarkable to me after 15 years of frustration up until four years ago.

Weight loss, which can be trivialized to a percent of five to 10 percent, can mean a lot when you weigh 180 pounds and you go down to 162 pounds. People would kill for those things.

The enthusiasm I've seen in Type 1 diabetics who are totally controlled on pumps includes the improvement they have and those terrible instructions we have to give people which is plan your carbohydrates, count a unit of Lispro for every 15 grams you're going to eat.

If you take four units of Lispro, you can't eat more. That is the problem they face. What has happened with pramlintide is the satiety effect has actually allowed them to limit the intake to what they had planned which, for many of us who are fortunate enough not to have diabetes, is a near impossible task.

I think the behavioral aspects in terms of diet are remarkable. I think the improvement in glycemic control are remarkable and I strongly urge the committee to consider approving this drug.

I would like to finish on an ironic note. I know hypoglycemia is a very serious side effect but I find it ironic with the release of Lantis, my clinical experience again just as a practicing internist has been somewhat startling.

I know it came out with the approval that the dose of Lantis as a once-a-day agent be 70 percent of the basal insulin over 24 hours. I don't know a clinician in the area where I practice that dares to use more than 50 percent because of the severe hypoglycemia we see in post-marketing and I find it ironic that it got out with the 70 percent recommendation but I don't think it's anybody's fault. I think the disease continues to humble us in these post-marketing studies. I think it will continue to humble us for a long time unless we start looking for more unique solutions than we have to date. Thank you very much.

DR. KREISBERG: Thank you. Thank everybody else who took the time to come.

Oh, do we have one more? I'm sorry. Two more.

MS. KRUGER: My name is Davida Kruger. I am a certified nurse practitioner at Henry Ford Health Systems in Detroit, Michigan. I have done clinical research for Amylin for the past five years. I've been a speak for them and I do own some stock.

I've also been an investigator for the diabetes control and complications trial, EDIC trial, and presently for the accord trials, all of which look at intensifying tight control for people with Type 1 and Type 2 diabetes.

Over the past five years I have enrolled more than 60 patients in the four pramlintide trials that I have been involved in. Despite, nausea, increased number of injections, none of our patients have withdrawn from any of our clinical trials. Why? Primarily because pramlintide has provided an improved quality of life, something that I haven't heard this panel talk much about today.

With decreased swings in their blood sugars, decreased postprandial rises, less insulin needed, less hypoglycemia, and weight loss. They generally feel better, something that is very important to people's lives with diabetes.

When we had an opportunity to offer open-label in the Type 1 study, we had enrolled 21 volunteers and 18 of those 21 volunteers chose to continue despite the fact that they were on seven or eight injections a day or they were on an insulin pump taking three to four injections a day of pramlintide.

On the three that didn't continue they wanted to go from the therapy they were on and to insulin pumps and that was not permitted.

When the open-label study was stopped, our patients were incredibly disappointed and I still get telephone calls asking when will this drug be approved. They still don't have the control they had when they participated in the study.

I'm also concerned about how ADA standards of care for the HbA1c has been looked at today. We struggle with controlling diabetes and even with everything we know, working with the DCCT, working in the core trial, we need every tool we have to achieve the goals we have and still commit to these standards HbA1c over 7 percent. We all want our patients to be less than 7 percent but we need every tool available to achieve that.

I would never minimize the risk of hypoglycemia. I worked diligently with it in the DCCT and I'm working with it now in the core, and I worked with it in our pramlintide studies.

But, in fact, within the DCCT a small cadre of patients accounted for the majority of the hypoglycemic events. When we worked with those patients and when we adjusted our protocols throughout the nine years of the study, we were able to decrease hypoglycemia.

It's a side effect and a risk to every therapy we use in diabetes and we can do that with pramlintide. We have done it and it has been demonstrated today but we need that tool for our patients.

I've also heard pramlintide called a burden because it needs to be injected three or four times a day. The reality is diabetes is the burden, not pramlintide, not the injections. The reasons patients have concerns about injections, it's not the patients, it's the providers.

I've not met in the 19 years of my practice any patient that will not take an injection if it will improve the quality of their life and prevent the complications of the diabetes. It's how we as providers perceive that injection and how we present our case to the patient.

As a provider diabetes care is an art, not a science necessarily. It all falls together and we need every tool available to help patients improve their outcome and improve their lives. I believe that pramlintide offers that one more tool that we need in the cadre of what we have to provide. I urge you to approve that medication for us today so that we can continue to improve the lives of people with diabetes. Thank you.

DR. KREISBERG: Are there any additional people that wish to make a statement? If not, I would like to thank everybody that took the time to come.

It's now 15 minutes to 1:00. We are going to adjourn for an hour and we'll reconvene at 15 to 2:00. Thank you.

(Whereupon, at 12:47 p.m. off the record for lunch to reconvene at 1:45 p.m.)

 

 

 

 

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

1:57 p.m.

DR. KREISBERG: We probably have a long agenda for this afternoon and we're about an hour behind schedule. We don't want to constrain any of the discussion that goes on so we need to be as expeditious as possible.

We have a break scheduled but whether we take it or not, I think, depends upon how we're progressing.

This afternoon will be devoted to an opening statement by Dr. Orloff. Then we are going to proceed with an in depth discussion which will involve the panelists and the company with regard to the data that was presented this morning as well as other pathophysiologic or physiologic issues.

With that, David, I wonder if you would start.

DR. ORLOFF: Thank you. This item on the agenda is called Charge to the Committee. I think traditionally it involves reading the questions to the committee. I don't think I need to do that.

I thought I would take a few minutes to sort of sum up FDA's concerns and try to set the tone for the discussion and deliberations surrounding the questions. Then any clarifications you need as we go on what we might or might not be referring to in the questions, we'll be happy to offer from the table here.

I wanted to make sure that everyone understands that when we approve a drug, it is on the basis of a determination that we know enough about the safety and effectiveness of the drug so that we can label it for safe and effective use by patients across the populations in which it is indicated.

To go into that in a little more specifics, we need to be able to tell people if they use it according to the directions what they can expect with regard to risk and well as benefit.

But the start of that phrase was "if used according to the directions." We also need to know how to use it. We need to be able to develop directions for use.

I think that a central theme that has come out of the FDA's review of this application, and I think what has been conveyed in the presentations by FDA, is that, as I said, we have concerns about whether the data are adequate, whether the trials were adequate by their designs to guide physicians and patients in the safe and effective use of the product.

We all agree that on average across both types of patients there was a small statistically significant mean reduction in HbA1c in association with pramlintide use when added to insulin.

There was definitive demonstration if efficacy of this drug, albeit on average relatively small. You saw as presented by the sponsor that this was attributable to really across the board for all categories of HbA1c response from baseline an incremental, I guess, population response this from the cumulative response data that they showed.

The question that is -- one of the first questions that we've come up with is whether the trials were by their designs adequate to address the efficacy so that we know about expected benefits of this drug when patients are treated towards optimum glycemic control goals.

Again, the sponsor did show a pulled subgroup analysis of response among patients who started out in the lower ranges, I guess, of HbA1cs within their population of Type 1 diabetic patients, and it looked as though in that subgroup analysis that the response was similar to that for the overall cohort.

However, I would caution you that there are lots and lots of subgroups in a database and this may well be something that needs to be prospectively investigated.

In addition, you heard about our concerns regarding a safety signal that arose particularly in the Type 1 patients with regard to an increased risk of hypoglycemia, particularly, I would say, in the early phases of treatment but, as expressed in the safety review, apparently associated with significant sequelae in a number of instances.

I emphasize now that this is a signal. I think everyone can see that to the extent that no one ever really knows what the safety parameters are going to be of a drug when it goes into investigations, admittedly for a diabetes drug, hypoglycemia is something everybody looks for.

To the extent that we don't know the full spectrum of the safety concerns that are ultimately going to crop up, we never have complete ascertainment of all the events of a given type.

It is a signal and I'm not really sure how much of a concern we need to have regarding, if you will, the integrity of the database to have enumerated every single event of the types with which we are concerned. It is safe to say there is a signal there. Perhaps it bears further investigation.

I would also add it is our impression, again something that may bear further investigation, that the hypoglycemic risk is not necessarily restricted to market responders or to patients who entered into the trials with relatively low HbA1cs.

Finally, I would reemphasize the fact of the issue of our concern about whether the trials were adequate by their designs, again referring back to Dr. Misbin's comments about the maintenance of insulin dosing as part of the protocols.

We question whether the trials are there for adequate by their designs to shed light on the safe use of the product under conditions aiming at strict glycemic control.

The questions that you'll get relate to safety, efficacy, approvability, and any recommendations for further studies, clinical or otherwise, that would shed light, or shed further light on this decision by FDA.

Finally, I should have said this at the beginning, but I do want to thank the sponsor and the FDA for their clear discussions of the data. I want to thank the presenters at the open public hearing for their compelling testimony. We are ready to proceed and listen to the deliberations as they go. Thank you.

DR. KREISBERG: Thank you, David.

Well, as usual, I'm sure that many of the panelists and the representatives from both sides will actually have questions. I would like to do this in an orderly fashion and I wouldn't like any one person to monopolize all of this.

What I would like to do is either recognize you to ask your question and then ask you not to ask a second question unless it is related to the question that you previously asked. We will make sure we'll keep going around the table until everybody has had an opportunity to have all of their questions asked and to get full explanations of them.

Having said that, I will primarily try to keep the peace and let people ask the questions that they want. Why don't I start on my left and we'll just go around. If you have a question, fine. If you don't, I'll understand that.

There was a question brought up earlier by the sponsor in response to a question about the database. I would like to say we do not consider the database to be an issue in our discussions but if the sponsor would like to make a very brief statement regarding the database, I would be glad to receive it.

DR. KOLTERMAN: Thank you, Mr. Chairman. On behalf of the sponsor, I would just like to state that we unequivocally feel that our safety database is reliable for the purpose of making regulatory decisions.

There are three issues that have been called to your attention in the briefing book that I would just like to very quickly give you some expanded details on.

First slide, please. One related from an inspection of a site in study 137-112 related to patient 2216 who had a single hypoglycemic event that is represented in the database as nonserious and severe is possibly related. This event was recorded as an adverse event but was not recorded by the subject in the hypoglycemia diary so the event is in the database and is accounted for.

Next slide, please. This was also in study 137-112 related -- that's the same slide. This was in the extension of study 112 where the subject had a severe hypoglycemic event and an associated motor vehicle accident in 1997 involving paramedics.

This event was recorded in the database as hypoglycemia and was classified as serious life threatening. A narrative was provided on the event. This event did not appear as a motor vehicle accident largely due to an issue with the coding dictionary that was employed. The WHOART coding dictionary, which is an industry standard and is accepted by the agency, does not have a preferred term for motor vehicle accident.

Had the patient had any injury that resulted from the accident, the injury would have been captured and would have been coded.

Next slide, please. In the last instance to provide clarification, there's a comment about some missing records for study participants at a site. This is a site that was affiliated with the clinical practice of a clinical endocrinologist who subsequent to completion of the study closed their practice.

The records were removed to storage when we were notified that the inspection needed to occur. The sites were brought back to a common area from storage. In that transfer process it appears that two records were misplaced and were unaccounted for.

Copies of those records exist at Amylin. They were shared with the site and the data related to that subjects are accounted for in the database. This in conjunction with the standard process of auditing and what have you serves as the basis for my statement to you. Thank you.

DR. MISBIN: Mr. Chairman, may I just make a brief comment about that? I think one has to remember whether something is in your database or not, the question is whether it is accessible to our review and we try to find those patients when we became aware of this on the inspections. We were unable to identify those cases in the electronic submissions that you gave us.

Now, I have to point out so there's no misunderstanding one of these patients was on pramlintide, one was on placebo so there is no -- there should be no idea that there was anything inappropriate about the way the data was collected or reported.

In fact, when the inspectors told us they could not find these cases, we went and looked at the submission and we could not find them either. In my judgement, we are really not sure about how many cases of MVAs there are and how many cases of severe hypoglycemia there are.

DR. DATA: I don't know that we're going to solve this issue here. I believe it's a discussion that needs to take place between the sponsors and ourselves relative to whether actually these patients are or are not in the database.

Our understanding is that they are and your understanding is that they are not but I think we need to move on.

MS. KILLION: All right. Let's move on. I have a question as it relates to hypoglycemia which, of course, is a serious issue that we deal with every day as diabetics.

Something that hasn't been clarified to my liking or understanding is, is there something about pramlintide or the patients on pramlintide where the onset of hypoglycemia is very rapid or is there some kind of impairment of the sensitivity to or the awareness of oncoming hypoglycemia that we should be aware of?

I mean, I know when I'm becoming hypoglycemic I can feel it but I know that sometimes that is not the case for all patients. Is there something about this? I wouldn't get in a car if I was feeling I had low blood sugar. If I'm in the car I eat my lifesavers or whatever. What's going on here that these people are having serious hypoglycemia that requires intervention by another party?

DR. KOLTERMAN: We do not believe that pramlintide alters an individual's ability to sense hypoglycemia. That statement is based upon results from a clinical pharmacology study that was done evaluating pramlintide treated patients with insulin-induced hypoglycemic challenge.

Neither the counter-regulatory response nor the ability to sense symptoms were altered in that 14-day study. We have slides that Dr. David Maggs from our Medical Affairs group can review with you if you would like.

DR. MAGGS: Good afternoon. Thank you for the question. We have no evidence that pramlintide influences the responses to hypoglycemia or the rate of glucose decent, etc., with regard to responses seen in Type 1 patients.

Slide up, please. We have conducted a series of studies which I'll just quickly touch upon briefly. The first two studies employ the use of pramlintide at super pharmacologic doses so I won't dwell on those two studies for the moment.

The third study, 9308, is a 14-day study that was conducted in Type 1 patients in which they had a hypoglycemic challenge conducted days prior to initiating pramlintide therapy. They received pramlintide at 300 micrograms TID for a 14-day period at which time they then had a second hypoglycemia challenge.

The net results from these three clinical trials but in this trial inclusive, or as indicated here, there was no effect of pramlintide on the glucose decent, the glucose nadir, or glucose recovery.

No effect of pramlintide on counter-regulatory hormone responses, metabolic substraights. Psychomotor testing was also conducted and there was no effect there. Lastly, no effect on hypoglycemia systems.

Next slide, please. This is data taken from this last study looking at the glucose decent, nadir, and recovery. What was carried out in this study during the hypoglycemia challenge was from zero to 100 minutes an insulin infusion was conducted after the time which the patients had relatively stable glucose levels.

During this period of insulin infusion, glucose descended. At the time of glucose nadir the insulin infusion was switched off. A low-dose insulin infusion was substituted and then you saw glucose recovery.

The 10 plot on each figure indicates the effective of pramlintide at this dose in each of the three treatment arms whereas the red plot indicates the effect in the hypoglycemic challenge prior to the initiation of pramlintide therapy.

Next slide, please. This quickly shows you the catecholamine responses noted in these studies where the incremental two-hour area under the curve for the catecholamine during the hypoglycemia challenge, the catecholamines, of course, being the key hormones with regard to hypoglycemic counter-regulation. We saw no difference in the catechol responses during these hypoglycemia challenges.

As Dr. Kolterman indicated, super pharmacologic doses of pramlintide in healthy subjects does not induce hypoglycemia. Under these conditions in Type 1 patients pramlintide does not influence counter-regulatory or hormonal responses.

DR. MISBIN: Mr. Chairman, could we ask the sponsor to present the data from the two earlier studies, the five-day studies that they have not presented yet? The committee really should see the entire evidence rather than just the one study that we all agree is negative. If you are unable to find that, we have brought the data that you supplied to us.

DR. KOLTERMAN: We have the data but I want to point out to the committee that the reason that we showed data from the study that Dr. Maggs just showed is that these studies employed a hypoglycemic challenge prior to the initiation of pramlintide therapy and on the last day of pramlintide therapy. The other two protocols while using much higher doses of pramlintide also were five-day exposure. There is data in the literature that shows that one hypoglycemia challenge alters the response to a subsequent hypoglycemic challenge for a period of up to 10 to 12 days. The data that we just showed was from a study of 14 days duration which we think is the most representative data that exist addressing this important issue.

DR. MISBIN: Mr. Chairman, again, we would like the opportunity to show the data if the sponsor does not have the data to show.

DR. MAGGS: Slide up, please. Again, going back to these three studies, let me just highlight the important points in these studies. As Dr. Kolterman indicated, the hypoglycemic challenges were done just prior to the initiation of pramlintide therapy and then days later.

What I would draw your attention to is in the first study pramlintide was administered at 500 to 1,000 micrograms once daily on the days of the study. At the completion of the study, hypoglycemic challenges were conducted at peak or trough levels and/or placebo.

The point being that we are conducting a hypoglycemic challenge where pramlintide levels are either at a high level in circulation or at a nadir the following day.

In the second study, 9303, a similar principal pramlintide administered now QID dosing at 800 micrograms dosing. Again, very super pharmacological. In both instances you can liken this to administering five or 10 grams of metformin to a patient at the time of conducting a hypoglycemic challenge.

If I may have the next slide up. The data that Dr. Misbin refers to is the hypoglycemic symptom scores from these two earlier studies that I refer to and the last study which I drew attention to at the very end.

In the first study we saw subjective symptoms of hypoglycemia scored at the time of the hypoglycemia challenge. Remember, these are patients either on placebo or pramlintide either at peak or trough levels.

In the pramlintide treated patients baseline indicates these hypoglycemia scores in the pretreatment challenges. The second line indicates the symptom scores at the time of the pramlintide challenge.

There is no clinically relevant change in hypoglycemic symptom scores during the course of this particular clinical trial or, in fact, the second clinical trial.

What I think Dr. Misbin is referring to is the number here which falls from seven out of 12 to three out of 12 in the peak study, and in the 9303 the 9303 the score of going from 19 out of 20 in this instance to 13 out of 20 in this instance, the pramlintide peak study, the pramlintide treatment arm for the second study.

This raises the question of hypoglycemic unawareness. Again, I should draw your attention to the fact these were studies done days after an original hypoglycemic challenge questioning the validity of doing a second challenge only days later.

Also, these patients are being studied at a time when they had had super pharmacologic doses of pramlintide administered.

I should also draw your attention to the fact that in 9302 this group of patients had a much lower HbA1c at entry compared to the other treatment arms which, again, confounds the interpretation of this data.

Finally, 9308, which is the most clinically relevant study, where we are actually administering doses of pramlintide somewhat nearer to pharmacologically use. In particular, the 30 microgram arm of the study. You can see that during the course of these two hypoglycemic challenges separated by 14 days, the hypoglycemia scores in 30 microgram treatment arm were, in fact, unaffected or, if anything, slightly increased.

In the placebo arm, again, slightly unaffected or you could argue slightly decreased. The net message from these three studies is that this last study bears the most clinical relevance and there doesn't seem to be any impact on hypoglycemic symptom scores as the studies were designed.

MS. McBRAIR: As the consumer rep, I was particularly taken by the testimony of the patients and the caretakers involved. One of the things I didn't hear anything about was quality of life studies and wondered if any were done and what their results were.

DR. KOLTERMAN: There have been no formal quality of life testing done with pramlintide. We are at the juncture where we feel that we have demonstrated efficacy, provided an assessment of safety, and have planned to undertake formal quality of life studies as part of early post-marketing studies.

DR. CARA: I'd like to clarify a point that I was trying to get at earlier and make sure that I understand it. This question is actually for Dr. Misbin, if I may.

I was asking you about looking at adverse events, specifically hypoglycemia in the "intent-to-treat population" as a way of getting at possible hypoglycemia in people that had been discontinued from the study.

My understanding is that there really is no way to do that so that if somebody had dropped from the study for any reason, that any hypoglycemia that they had had would not go into the final data analysis. Is that correct?

DR. MISBIN: No, I don't think that's true. The sponsor could correct me but I think any hypoglycemic event would have captured in the data that we have regardless of whether the patient dropped out.

DR. CARA: I thought that was only a completer analysis?

DR. MISBIN: Well, no. The ITT analysis, and please, the sponsor will correct me if I misstate this, but that pertains to efficacy. The data I showed on the ITT for A1c reduction, that pertains to efficacy with the last observation carried forward for patients who dropped out.

But any hypoglycemic event that occurred anywhere in the trial in anyone would have been captured assuming that we got it in our electronic submission which, of course, we have already dealt with.

But once anything that was entered into the database got to FDA regardless of when it was, even a patient who dropped out, that would have been analyzed in Dr. Roman's presentation.

DR. CARA: Was all of that data included in the submission?

DR. KOLTERMAN: Yes, it was. Perhaps more importantly is that the analysis of the safety database that I presented this morning was based on the intent to treat cohort that included all patients. If a patient started in the trial, had some sort of an adverse event and dropped out, that adverse event was captured and reported in the data that was presented.

As it relates to severe hypoglycemia, we have done some other analyses. If you do the intent to treat analysis and you compare that with either an intent to treat observed cases which means that you just look at the observed values that end up being entered into the database. You don't try to account for patients dropping out.

Or if you limit the analysis to what we refer to as an invaluable cohort which means patients that started the study and continued all the way through the period of observation, you see similar patterns and you come to the same conclusions. Neither the safety nor the efficacy conclusions that we shared with you this morning are adversely impacted by dropouts.

DR. CARA: I'm sorry. You said that you come to the same conclusions. What conclusions are those?

DR. KOLTERMAN: They are the conclusions that I presented this morning in terms of severe hypoglycemia in the Type 1 population. There is an increase occurrence irregardless of whether you look at incidence or whether you look at analyzed event rates during the first four weeks of treatment beyond the first four weeks of treatment the occurrence of severe hypoglycemia is similar between pramlintide and patients treated with insulin alone.

DR. LEVITSKY: The compelling comments of the patients and care givers who spoke this morning compared with some of the data that we were presented with suggest that there may be categories of patients for whom this drug may be very much more appropriate than others because I could not imagine as a care giver having the degree of intensity and wish for this drug if I had seen a e or 4 percent drop only in HbA1c in the patients I had in a .3 or .34 percent drop. Clearly there are some patients who do very much better. Is this related to the frequency of blood sugar monitoring? We know the frequency of monitoring can reduce the risk of hypoglycemia to some extent.

Have you looked in anyway to separate out what types of patients are likely to be the ones who would stand up and tell us how their life has been changed compared to the others who clearly just added to your statistics?

DR. KOLTERMAN: Yes, I believe I can. We've looked -- the short answer is we've looked in detail for identifying factors that predict who will have a robust HbA1c response. The only parameter that is identified, you know, you look at baseline demographics, insulin types, what have you. The only thing that was identified as being a major predictor is that the higher the baseline HbA1c, the larger the reduction in HbA1c that is usually seen. That is consistent with what is seen with other therapies.

If you look at the entire population of patients treated with pramlintide, as I showed you during the presentation, 70 percent achieve some reduction in HbA1c. Over 90 percent achieve a reduction in HbA1c or improvements in body weight control or a combination of both.

We think as with a number of hormones, given where the patient is in comparison to other factors that can impact, you know, glycemic metabolic control, that the pattern of response that you see is different.

Maybe going forward there will be better ways to identify people that particularly have a nice response. At the present time we looked at all of the data that is available and have not been able to find the magic bullet that would allow us a priori on clinical grounds to select those patients.

DR. TAMBORLANE: I think this sort of segways into a common two related questions that I have.

Dr. Baron talked about the unique properties of Symlin and its ability to control postprandial hyperglycemia. I think one of the messages we were hearing from the patients was the smoothing effect. Where you might have more severe events initially, that there seems to be fewer fluctuations in glucose.

That particularly resonates with me because we have just completed a sensor study involving 56 of our kids primarily who were on insulin pump and using Lispro and well controlled with A1cs of about 7.7.

Despite what we would consider great control, 90 percent of those youngsters had postprandial peak glucoses over 180 which would be our normal target. 50 percent of the kids had glucose values over 300 milligrams per deciliter peak postprandial values.

Here are my two questions. One is, do you have any data during either the Phase 2 or Phase 3 studies looking at glucose fluctuations in the treated patients? The second question is are you considering in your plan any pediatric studies?

DR. KOLTERMAN: The answer to both of your questions is yes in terms of plans for additional studies to explore those issues.

I'm sorry. If I could have the slide that was on the monitor up, please. There is a limited amount of data in the Phase 2 program. If I could have the slide up, please. I think that is consistent with both what we heard from the patients this morning as well as the data that you alerted to from your own group.

Here a group of patients were treated for 28 days with 30 micrograms of pramlintide given four times a day. These patients had 24-hour glucose profiles done. This was in the presensor area so this was a more laborious means of bringing patients into a metabolic ward and doing frequent sampling.

You can see with the patients using similar insulin regimens, controlled meals, that prior to treatment you see this kind of a glucose pattern. Whereas with pramlintide on board, there is less fluctuations of glucoses here in this postprandial period.

In this study we tried looking at the data in a number of different ways to get some statistical measure of glucose variability. As you may understand, that can be a difficult thing to do.

There is a slide that we may find here that plots the variance, just a simple sort of variance, and it is reduced in the pramlintide treated groups.

In terms of pediatric studies, we clearly understand in terms of Type 1 diabetes that is an important population to look at and we cannot just assume that what we have in adults will extrapolate to that population.

In September of 2000 at our pre-NDA meeting with the agency we actually made a commitment to them to undertake a pediatric study following approval of the drug -- a pediatric program.

Slide up, please. We did find this one. You can see here that the variability measures as a standard deviation of the plasma glucose concentrations is reduced in four different dosage regimens of pramlintide treatment versus the same patients treated with insulin alone.

DR. GRADY: Would you put that slide back for just a sec?

DR. KOLTERMAN: Sure. Can we have the slide back?

DR. GRADY: I'm sorry. It really looks like the variability went from, say, 74 down to about 65?

DR. KOLTERMAN: That's correct. Slide off.

MS. McBRAIR: You may have already answered this question, I thin, by some of the discussion but one of the points that Dr. Misbin made, and I guess I was struck by the same thing, when you showed some of your earlier data and showed what a dramatic decrease you got in the postprandial blood sugars, I was struck that overall the change in glycated hemoglobin was relatively small it seemed for the dramatic change in postprandial.

The question I had was do you get a waning effect? Because I know that was one of the things that Dr. Misbin pointed out. Are there some patients where you do see it and in other subsets where you don't? Have you looked at the data that way?

Again, I think it gets to the question of is there a population, especially hearing the patients this morning, who seem to have had such a dramatic increase in the control of their blood sugar from this drug that kind of is somewhat discordant with the overall picture of the data presented in aggregate?

DR. KOLTERMAN: I believe you've asked two questions. Let me address the first one. That is, the apparent discrepancy between the reductions in postprandial glucose observed and the eventual results in reductions in HbA1c.

The reductions in postprandial glucose were done as acute studies, single center studies in very carefully controlled, if you will, metabolic ward conditions. Those are relatively precise measures with a lot of the variability that occurs in the day-to-day control of plasma glucose factored out.

If one looks at those reductions in postprandial realizing that pramlintide does nothing, appears to do nothing to lower fasting glucose and leverage off of data published in the literature, you would anticipate a reduction maintained long-term of approximately .7 to .8 percent of HbA1c.

That, we believe, is the merit of the stable insulin evaluation. If I could have the slide from the Type 1 presentation, stable insulin.

When you look to that 30 to 40 percent of patients that for whatever reason did not vary their total daily dose of insulin by more than plus or minus 10 percent and, therefore, I believe, isolated the effect of pramlintide.

Slide up, please. You, in fact, see a reduction in HbA1c that, in fact, is almost one percent here from baseline initially during the early part of the double blind period where, as Dr. Tamborlane referred to earlier in the morning, there is still a study effect present where everyone has sort of had a come-to-Jesus meeting with the investigator and what have you so there is more glucose monitoring, more attention, yadda, yadda, yadda.

There is a bit of waning in that group by 26 weeks. But if you then look between week 26 and week 52, the line is flat as a table top. Okay? I think this shows durability of response. It also yields a reduction in HbA1c of .7 percent which is exactly what you would predict from the reduction seen in postprandial glucose concentration. That, I think, is the tie.

The other point that I would make relates to the differences of clinical trial design versus routine clinical practice. I think it is clear that there is a reason why the B-cell secretes both insulin and amylin.

The two work together sort of hand in hand. One works on the input side to the system and the other works primarily on the output side of the system. These two need to be used in conjunction with each other and titrated with each other.

Since there has been no other compound like this for the treatment of diabetes evaluated, the initial double-blind placebo controlled trials needed to focus on quantitating drug effect, just what we did right here, to demonstrate efficacy because if there is not efficacy, one is not justified from an ethical standpoint to continue to expose patients to this.

Now that we have demonstrated efficacy, we are in a position to consider single-center trials where things can be explored in considerable detail to define various protocols for really how to optimize the interplay between these two pancreatic hormones. Thank you. Slide off.

DR. TAMBORLANE: I've been looking at the data trying to understand the dose response nature both in Type 1 and Type 2 that allowed you to select the proposed doses that you have put in your materials.

DR. SAMPSON: I'm wondering have you done any -- again, I have not seen it -- some sort of integrated summary of efficacy that would have addressed the issue of dose response. To the naked eye it looks fairly flat in both cases, and yet your choices of doses don't seem to totally reflect that. The specific question is dose response. How did you come by your choices of doses, your recommended doses, both for Type 1 and Type 2?

DR. KOLTERMAN: Right. Okay. The recommended selection of doses for Type 1 and Type 2 for use in clinical practice is based upon observations in the long-term controlled trials. The doses for use in the long-term controlled trials was covered in part in the presentation this morning.

If I could have the slide up, please. It comes from data similar to what we were just talking about in terms of a reduction in postprandial glucose concentration.

This is data from Type 1 patients and plotted as a change from baseline. This is change from fasting, if you will, in plasma glucose values over time following the ingestion of a standard Sustacal meal challenge. There is a dose-dependent decrease in the degree to which these plasma glucose concentrations rise.

Next slide shows that there is, in fact, a dose response relationship and this was -- slide up, please -- tested on the glucose data using an appropriate statistical test that I cannot describe to you as a clinician but I have statistical colleagues who can help us out if need be.

We also see that there is a dose response relationship here for the most commonly occurring side effect, nausea. We think there is evidence in the Type 1 population of a dose response for both efficacy and the side effects.

The next slide shows a similar approach to patients with Type 2 diabetes. Here we used a HbA1c endpoint -- slide up, please -- looking at treatment at the end of 13 weeks where you can see a dose dependent decrease in HbA1c from the range of 30 to 150 micrograms. Again, there is also a dose dependent increase in the nausea profile.

There are similar data at 26 weeks and 52 weeks of exposure that was subjected to, again, one of your statistical tests to document the presence of a dose response that was reported in the study report in the integrated summary of efficacy.

Slide up, please. In another study in patients with Type 2 diabetes, namely study 122, which I showed you data from this morning, here is the plot of the statistical function test, the dose response test, done by the statisticians with a statistically significant p-value here.

DR. SAMPSON: Do you have dose response data for Type 1 patients or does the same question apply to individuals with Type 1 diabetes?

DR. KOLTERMAN: Okay. We --

DR. SAMPSON: At least, I think, the data in the Type 2 patients is different from the data that you see in the Type 1 patients in the sense that in Type 2 patients you need to use higher doses and there is more of a dose response relationship.

I don't get a sense that there is a clear dose response relationship in the studies that you did with individuals with Type 1 diabetes.

DR. KOLTERMAN: Okay. If we could have the slide up, please. Let me try something. This looks across a range of doses. Admittedly, this is not HbA1c data. I've went through the argument with you how the postprandial glucose data correlates with the HbA1c data. This goes from a range of --

DR. SAMPSON: Sorry, but I didn't get a sense that you did for Type 1 diabetes.

DR. KOLTERMAN: I'm sorry?

DR. SAMPSON: I didn't get a sense that you did, in fact, talk about dose response data in relationship to HbA1c data for Type 1 diabetes and that was what my question was about.

DR. KOLTERMAN: Okay. I cannot show you dose response data for HbA1c. If I can have the slide up, I can show you an intermediate marker of glycemic control, namely fractosomine. This is from a 28-day treatment protocol in patients with Type 1 diabetes. Plotted here is the change in fractosomine from baseline for patients treated with insulin alone versus patients treated with 30 micrograms four times a day, 60 micrograms given three times a day, and 60 micrograms given twice a day. I believe that you can appreciate a dose response in terms of fractosomine response here.

I call your attention that the doses that we are recommending are 30 and 60 micrograms in the Type 1 population. We did work at 90 micrograms. It's not being recommended because it's clearly not a well-tolerated dose.

DR. CARA: Can you show your curve for efficacy as related to blood glucose concentrations in Type 1 diabetic again?

DR. KOLTERMAN: The dose response?

DR. CARA: The dose response. I want to see how the 30 and the 60 --

DR. KOLTERMAN: Sure. That's a very fair point. Can we have the slide back, please? Slide up, please. This is data from the 30 microgram dose here, 60 microgram dose here. I'm sorry. This is 10, this is 30 versus 60 here. This is 100 micrograms.

DR. CARA: So it's quite surprising to me that, on the one hand, you are really dealing with sort of the lower minimal doses when it comes to glucose effects and, yet, in terms of lycohemoglobin levels. You are sort of not really testing, at least from this data, what I would consider to be adequate doses.

DR. KOLTERMAN: If we could have the summary bar chart of HbA1c intent-to-treat analysis from the presentation, please.

I apologize, Mr. Chairman. We're having some technical difficulties over here with the computer.

Slide up, please. This is the summary change in HbA1c from baseline to 26 weeks across the three studies in Type 1 diabetes that was in the presentation this morning. This data here is essentially data from a 30-microgram dose compared to doses with 60 micrograms across here. In terms of magnitude of effect, there is little difference here which suggest that in Type 1 patients that 30 micrograms may be close to the top end of the dose response curve.

If we look at weight data from patients with Type 1 diabetes, there is some added benefit in terms of body weight in terms of going to the 60 microgram dose.

DR. CARA: But, again, I'm not convinced that there is truly a dose response relationship. If you look at the 30, 60, and 90 microgram doses, there is not a linear dose response relationship in terms of the correlation of that dose response relationship that you are pointing out here in terms of

glycosylated hemoglobin with the glucose-lowering effect, the postprandial glucose-lowering effect of the drug.

DR. GRADY: Could I ask people, you know, if you look at the third slide from Dr. Misbin, it shows exactly what -- I mean, it really shows the data on this issue. There's just no -- it's pretty much flat. The effect on HbA1c with different doses of the drug is just flat.

DR. KOLTERMAN: You are correct. In the presentation I also pointed out that 30 micrograms yielded plasma concentrations that are similar to circulating amylin concentrations during postprandial period in non-diabetic individuals.

The effects in terms of glucose -- a good part of the effect, a major part of the effect of glucose lowering in patients with Type 1 diabetes may already have been achieved by the time one reaches a 30-microgram dose.

That is part of our consideration in conjunction with patient safety for the recommendation of 30 micrograms as the initial dose for patients with Type 1 diabetes.

The place where there is a difference between 30 and 60 micrograms is in terms of the effects upon body weight. Slide up, please.

This is the corresponding data for body weight to that for HbA1c. Here you see that in study 112 the 30 microgram dose has a somewhat smaller effect on body weight than what the 60 or 90 microgram dose does in either study 117 or study 121. Slide off, please.

DR. CARA: If I were to then compare that to your dose response relationship into what you've just told me, I would say that the mechanism -- I would conclude that the mechanism of action of the drug then is primarily anoretic effect.

DR. KOLTERMAN: Okay. I don't believe that is a correct conclusion because when you look at the inter-relationship between the reduction in HbA1c and the reduction in body weight, reductions in HbA1c are similar. If you divide patients into those that lose weight and those that do not lose weight, the reduction in HbA1c is similar in the two cohorts. Have I stated that clearly?

DR. TAMBORLANE: I think I want to direct this to the FDA and I apologize.

DR. KREISBERG: Wait. I thought it was related to --

DR. TAMBORLANE: No, I have a different question.

DR. KREISBERG: Deborah.

DR. GRADY: When I think about your drug, it seems like an attractive drug for Type 1 diabetics mainly because there just aren't good options. Also because those are the patients at very high risk for complications.

As an internist I primarily take care of Type 2 diabetics. When I think about it with regard to Type 2 diabetics, the main concern I have is that its actions are very similar to those of metformin in that it reduces HbA1c, tends to reduce weight or prevent weight gain. However, it is much less strong in producing those outcomes.

It seems clear to me that the first drug that ought to be chosen in Type 2 diabetics would be metformin. Then the real issue is what additional benefit might your drug have when added in Type 2 diabetics to metformin. I wonder if you have any information on that question?

DR. KOLTERMAN: Yes, we do. I may take us a minute to find the slide. If you remember from the presentation this morning, 20 percent of patients in the Type 2 studies were using either a sulfonylurea or metformin.

If you subset out those patients that were concomitantly treated with either sulfonylurea or metformin, you see a beneficial response in those patients that appears to be above and beyond what metformin or the sulfonylurea is bringing to the patient.

While we are looking for that slide, if we could have the slide up, please.

This is looking across the studies of the patients treated with biguanides and without biguanides compared to the placebo. You can see that the patients treated with biguanides are showing the same response that the patients who are treated without biguanides.

As I teed this up in the presentation this morning, I think that pramlintide has the potential to bring benefit to patients with Type 2 diabetes after they have extracted the benefits, if you will, from the presently available agents.

DR. GRADY: Do you know what the effect is on weight in that same group, the ones taking biguanides?

DR. KOLTERMAN: Okay. I don't believe we have that on a slide but the weight effect is not dissimilar to that seen in the entire cohort. My memory is not good enough to quote it precisely as to whether it's identical or not. I do now that evaluation of this cohort does show that there is a weight loss in the metformin treated patients.

DR. MAGGS: I'm Dr. David Maggs. Just a follow-up comment. There is no reason why pramlintide and metformin can't coexist in the treatment of Type 2 diabetes. As Dr. Kolterman has pointed out, about 12 percent of our Type 2 diabetes cohort were already on metformin when pramlintide was added as he has shown.

We should also bear in mind that not all Type 2 patients can tolerate or take metformin through gastro and internal side effects or lactic acidoses counter indication.

The last piece that I should also point out is although there are similar effects on HbA1c and body weight, weight control or weight loss, they have two very different mechanisms of action. Metformin working glucose production and fasting glucose. Meanwhile pramlintide having an effect on controlling glucose in the mealtime period. There is no reason why these two compounds can't coexist in the treatment of Type 2 diabetes.

DR. KREISBERG: Orville, I wonder if you or one of your colleagues could address the issue of both weight loss and hypoglycemic responsiveness because I think the discussion was going in a direction that suggest that just interfering with the rate of gastric emptying is not likely to cause weight loss.

I mean, if you use glucoside ACE inhibitors as an example, they slow but they don't actually produce malabsorption. I would believe that if the diet were held constant, then there should be no weight loss.

I wonder if hypoglycemic awareness and responsiveness are normal, then it seems to me that the issue here is that they have changed their diet and they are not consuming the same amount of energy as they were consuming. I wonder if you have any trial data that is controlled enough to look at what the impact of this would be on the consumption of calories.

DR. KOLTERMAN: We do not yet have clinical trial data that provide quantitative assessments of that or look at change in the composition of the food selected.

There is, however, a considerable literature in preclinical studies indicating an effect upon amylin itself as a satigenic agent. One of my colleagues can share some of that data with the panel if you would like.

DR. KREISBERG: I would just think for additional studies that you might consider to get at the issue of hypoglycemia as to, (1) the nausea itself may, in fact, be anorexigenic. But if there are other more centrally derived properties of the drug, that a problem that is contributing to the side effect profile in this is that people have voluntarily restricted their intake.

I think that would be important to know that from future studies and it would seem to me it wouldn't take a lot of patients in order to determine what the effect was on energy consumption.

DR. KOLTERMAN: Okay. That is an area that we are interested in pursuing. I think it has actually potential significant clinical benefit to better understand that.

I make the point that patients that never experience nausea -- never experience nausea show weight loss similar to that seen in the patients who do experience nausea.

Again, the nausea, as I showed you this morning, or tried to communicate in my presentation with some clarity, is a short protracted side effect. It occurs primarily during the first four weeks and then dissipates. The weight affect is persisting over 26 to 52 weeks.

I think the nausea piece comes out of it, but I think the human data by inference as it now stands indicates that an amylin-like effect exist in man similar to that demonstrated in animal studies that decreases food intake. It appears to be a satiety effect as opposed to a food aversion effect.

DR. CARA: I was just looking at the number of subjects experiencing treatment-emergent adverse events. I am impressed by the number who are actually complaining of anorexia relative to placebo.

I can understand that you haven't quite established what that is due to, whether it's gastric filling or delayed gastric emptying versus a central effect. Again, I just think that needs to be looked at.

DR. KOLTERMAN: Not to cut you off but there is another issue lurking in that term in that that is the preferred term in the coding dictionary to which some rather interesting things collapse. A good number of the things that collapse and are reported in that term, you and I would probably view as beneficial attributes to have in a drug treating patients with diabetes in terms of feeling less hungry, not wanting to eat as much, what have you.

It's anecdotal information but a comment that you heard from some of the patients in the public comment session this morning, you know, believe me, when we had to stop the open-label extension studies, I took a number of fairly irate phone calls and the patients were saying a very common theme is that for the first time since I've had diabetes while using this compound I don't feel like I need to eat all the time.

DR. TAMBORLANE: Again, this is for the FDA. I'm kind of proud of being part of the DCCT that established the A1c as a marker for efficacy in treatment of Type 1 diabetes.

It seems to me that has facilitated studies of efficacy with oral agents in Type 2 diabetes. I apologize if I don't get all the details right. In that case, you can withdraw a patient from their oral therapy.

You can allow their A1c to go up, and then you come in with a placebo controlled study which then may actually return their A1c back to where they were before they started. But compared to placebo or no treatment you have demonstrated efficacy.

It seems to me that the agency is holding a higher standard for a now unique study which the company pointed out that these kind of studies of glucose lowering agents in Type 1 diabetics or insulin-requiring diabetics, particularly Type 1 diabetic patients, is a different kettle of fish.

It seems to me also that you're holding them to a higher standard for efficacy by suggesting that the only way to demonstrate efficacy is in the context of a treat the goal and intensive diabetes regimen. I assume you make guidance to industries. Have you thought about this further and how are you going to deal with this in the future as well?

DR. MISBIN: Well, actually, I think, Dr. Tamborlane, it's really up to you to set the standard. There is no official documents as far as what the standards are. I think --

DR. TAMBORLANE: Me personally? I'm happy to try.

DR. MISBIN: You will have to vote at the end so that is, in fact, something that you will personally have to decide. I think it depends on, yes, that is one way of looking at it, they are being held to a higher standard but one could say why do we have a situation with patients with Type 2 diabetes when in the past we have allowed patients to be taken off of active treatment and, indeed, in some protocols allowed their glucoses to go up to 400.

I think there are clearly many people at industry who believe that informed consent in the general sense covers that issue. It's my personal belief that patients should not be required to accept a substandard treatment simply by virtue of being part of a clinical trial. Furthermore, the relevance of such data, I think, is very much open to question.

We have evaluated three new molecular entities with respect to insulin. Three new insulin analogs. Clearly we didn't do placebo control trials but these were all evaluated in patients that were in reasonably good control.

Patients had HbA1cs of roughly 7.5 to 8 and we just said let's take any one of them, Lispro, Asbart, or Lantis. We basically told doctors to treat how they would treat in ordinary practice and just at the end of the day see if there was a difference between the experimental drug and the standard drug. I've already presented the metformin data so I don't see any reason why we cannot hold companies to a standard that they evaluate their drugs under circumstances which they expect that the drugs will be used.

Now, what kind of label is right for pramlintide? If you use literally the patients that were studied, you would have to say these are patients in whom A1c levels are high and you are saying not to adjust their insulin dose.

That is clearly not the way the patients will be treated. As soon as you allow insulin to be varied, it may or may not be effective. It may or may not be safe. We really don't even have the data to answer that question.

DR. TAMBORLANE: Just to respond to that, I was trying to be very specific with my language, and that is showing efficacy versus safety. The problem I envision is that the sponsors are caught in a box when it's a different entity, not another kind of insulin in which I assume the FDA is willing to accept noninferiority as a criteria for acceptance where you have to show efficacy that they get caught in the box because the harder you work to adjust the insulin dose, it's a confounder.

My impression, only from limited experience with the FDA, is that other endpoints such as decreased variability and some of the quality of life and things like that are not hard endpoints. I assume that when these protocols are devised, that the agency provides guidance to the sponsor as far as what they might be looking for. That is just sort of an information question.

DR. ORLOFF: I think -- I don't think there is any disagreement as to whether or not the principle has been proven that this is not placebo in either Type 1 or Type 2 diabetes. Depending upon the disease target and the nature of the drug and its safety and efficacy profile, a proof of principle, notwithstanding any relationship to actual use, may be sufficient for approval.

What we are asking you actually is whether proof of principle in this instance is sufficient information for you to go ahead and recommend approval because you believe that we ought to be able to label this drug for safe and effective use and that physicians and patients alike ought to be able to use it across the board safely and effectively. We're not holding anybody hostage. It's a question of proof of principle versus how do you use the drug.

DR. MANIGOSKI: I'm Dr. Manigoski. I will second the opinion of Dr. Orloff. I think we don't disagree that this drug works and we have stated this time and time again. The point is whether to prove a principle you need to treat 2,000 patients. Maybe 100 patients will do to show that this works.

Then you have to study a population such that represents the general population and allow you to treat across HbA1c. What we have learned from the studies that have been conducted is that if you start at 9 you can go low to .3. What happens at 8, at 7 we don't have the slightest idea.

Of course, we can cut among the studies or the patients that were incorporated in the study in trying to answer this question but this is not the way we see that we have to look at the data.

We would like to see patients treated in the way in which you treat your patients in addition to these to see what happens. We have not seen that and we are looking into that. We have not forced the company to do the studies in this manner. They have chosen to do the studies in this manner.

DR. TAMBORLANE: Let me ask -- let's make a hypothetical. Let's say that we started with patients who had a HbA1c of 6.8 and they did a study to look at Symlin versus placebo with insulin and it showed that A1c did not change but that the peak postprandial glucoses were substantially reduced on a seven sample glucose profile. Would that be a sufficient endpoint for the agency to say that they have shown a benefit assuming that there was no difference in safety issues?

DR. MANIGOSKI: Assuming that there are not safety issues, we have to look at the issue of what -- okay, how relevant is postprandial glucose. I am not a diabetitologist. Three years ago when I became involved in this project when I went to the conferences and I learned about postprandial glucose, it makes a lot of sense. I say the money is there. If you diminish the postprandial glucose, you have a winner.

I think it makes sense to think in this manner. However, since then we have evaluated rapid active insulins and other compounds that act in the early phase. Although, as you have seen here, you have dramatic changes in the postprandial, at the end the HbA1c changes at a .3.

We still are debating whether what is the role of postprandial. The literature showing that if you correct for this, you may prevent cardiovascular disease. This is open to discussion. I will argue that the quality of these papers are not very good.

I think that the company has stated something also that is extremely important. They stated that this beautiful data and their conditions of clinical research meaning that these beautiful curves -- when I saw the curves I was really impressed -- is what happens when you bring the patient to the clinic, you fast the patient, you give them Sustecal, you give the injection at the time you want, etc., etc. When you leave that condition and you conduct a clinical trial, you see completely different outcomes and this is what we have seen.

There is one more step. You will be the first to recognize that when you have patients and their clinical trials, probably this is the best of care. The question is whether this level, .3, will remain once you have this in clinical practice.

Let me also add that I think clearly the company has shown, and we have shown, that you don't have a dose response relationship. Therefore, it is open to discussion. I argue very hardily that 30 is safer than 60 or 90. I have not seen the data to make that judgement.

I think we have hypoglycemia with 30, with 60, with 90, and under 20. We agree full heartedly with the company that it seems that during the first months you have more hypoglycemia. However, you have motor vehicle accidents that are attributed to hypoglycemia during the 12 months. We don't have a mechanism to discriminate who is going to have it, how he's going to have it, what dose, etc., etc.

DR. TAMBORLANE: I just want to follow up on one point about this clinical research, that it's a different situation. It certainly is. One thing as somebody who is actually trying to do some of these clinical investigations now, Davida Kruger talked about the art and experience that is involved in taking care of particularly Type 1 diabetes.

We look at the one month hypoglycemia data and say that this might be expected because whenever we change a regimen in a patient with Type 1 diabetes, we might see more hypoglycemia, but nobody has mentioned the fact that the clinicians who are caring for these patients, they are naive to this therapy as well.

They don't know perhaps what to expect either. Personally I would take some of the issues with the hypoglycemia reflecting a change in therapy in the patients and also a learning curve for the clinicians.

DR. KOLTERMAN: Mr. Chairman, if I could just interject a comment here. This has been a very nice conversation and the thing that I would like to add is that the experience with Lispro insulin is that it looks much better in the controlled setting of a metabolic unit than what the results with Lispro have been in the general clinical setting.

Also, in terms of the very real question that the panel needs to wrestle with this afternoon is do we know enough to provide reasonable instructions, reasonable informed instructions to clinicians as to how to use pramlintide therapy.

I would like to call upon Dr. Richard Dickey who has been a clinical consultant to us who is an endocrinologist in private practice in North Carolina who has some experience with pramlintide and also treats patients on a regular basis and have him comment upon the introduction of this into clinical use.

DR. KREISBERG: Richard, how much time are you going to take?

DR. DICKEY: How much time do you want me to take, Bob?

DR. KREISBERG: Well, not a whole lot.

MR. DICKEY: Couple of minutes. Thank you for inviting me to speak and thanks to Amylin for inviting me to sit on their team today.

I practice clinical endocrinology in Hickory, North Carolina, and I did a 121 study with 10 Type 1 diabetes patients. The results of the study were not dissimilar from the results that have been summarized today.

The importance of the study is that it was in the clinical practice with patients that I've been treating for some time who had not achieved optimal glycemic control to prevent complications. These patients were offered to participate in the study because of that.

80 percent of the 10 patients in the study had their insulin doses changed in violation of the protocol recommendations but the protocol did not restrict us from doing that. The clinical practice was consistent with what I thought was wise, fair, and appropriate.

The results of the study were similar to those that have been presented. One patient dropped out at two weeks because of nausea. He was on the highest dose. Two other patients dropped out because of the rigorous requirements of the protocol.

Not because of the drug but because of the frequent monitoring, the diary keeping, and because of travel. Some of these patients, as Claresa Levetan pointed out this morning, traveled 100 or 150 miles one way to participate in this study.

The results of the study, and I think this is important because when you do a study to try to obtain approval for a drug, you are looking at means. These are not means. These are 10 individual patients who participated in a study and whose drops in HbA1c were .5 to .8 at the end of the study for half of the patients in my study.

That is a significant improvement. Those patients were mad when I did not offer them to participate in the open-label extension which was offered to me and which I declined. They are still waiting for approval of the drug, as has been mentioned by some earlier today.

I was quite excited about this replacement for a natural hormone and participated actively in the study and have continued to follow this product in its development. I am happy to say that two weeks ago I found out about this hearing and asked to participate in the public comment session and instead was invited by the company to sit here as a practicing clinician. I hope that these findings in my particular study, which mirror the findings and some of the statements that were made by some of the individual patients, will convince you that there are people, in the case of my study,half of the patients that participated, who uniquely benefitted by this drug.

We did have three patients with hypoglycemic events. None of them serious or compromising the patients or resulting in any significant injury but reported as appropriate by the protocol. All of those patients continued in the study to completion. Thank you.

DR. KREISBERG: So we can do this in some order, I'm going to go back to Lynne because we skipped over here. Then I'm going to go to Marie and around and I'll come back to everybody.

DR. LEVITSKY: Okay. I will make one comment but I need to ask Dr. Dickey a question first. You had 10 patients, three of whom dropped out, and half of the seven that were left had a response and the other half did not. Is that correct?

MR. DICKEY: There were 10 patients in the study. There were three who did not complete the study. One dropped out at 44 weeks because he developed sarcoidosis and had to go on prednisone which was in violation of the protocol.

DR. LEVITSKY: Okay.

MR. DICKEY: One who dropped out at two weeks because of nausea. He was on the highest dose. One other who dropped out because of distance, travel, and the rigors of the protocol. What I said was five patients out of the 10 achieved reductions in A1c of .5 to .8.

DR. LEVITSKY: Okay. Thank you.

The comment which I guess I want to address to the FDA, I'm getting a fuzzy memory but I seem to recall that after the approval of human insulin, there was a big human cry about the fact that much more hypoglycemia was seen with human insulin than with beef or pork insulin.

There was real concern and worry about this which you guys had to deal with. It turns out it was probably just the doctors didn't know how to use it yet. When we learned how to use it, this wasn't an issue at all. Could you tell me how you went about sorting that out? It would be interesting because I hear something like that happening here, I think.

DR. MISBIN: None of us were here during the approval of human insulin. This is an extremely contentious area. I will tell you in advance that I am going to sound very evasive.

DR. KREISBERG: You weren't here but you're going to go into some detail about this? Is that what you're going to tell us?

DR. MISBIN: I'll just tell you things that I think would not be controversial. I think it's fair to say there are many people that feel very strongly that there is more hypoglycemia with human insulin than with animal insulins and they are very eager to convince the FDA of their position and we certainly listen to them.

They are very vocal and I don't trivialize this at all. They have diabetes and I don't. It's not up to me to say that they would do better on one product than another.

On the other hand, when one looks at the clinical trials, and this has been done, there is no difference with respect to these events. What is a regulatory agency supposed to do? We really cannot address public policy based on the perceptions of individual cases. We have this responsibility and we take it very seriously.

Now, otherwise the issue of beef and pork insulin and human insulin is something which we really cannot go into. The point that you're asking is very relevant and I wish you would ask it in another way so I would be able to respond.

DR. LEVITSKY: Tell me how I should ask it.

DR. MANIGOSKI: I will try to address that issue. I think it is very important. The point is that we have randomized placebo control studies. Even one arm of the study -- I don't know why because they are blinded -- you have four times more hypoglycemia than in the other. I don't know. Maybe it is placebo and maybe it's the drug. That is the way in which we address the issue.

We would love to know what the mechanism is as the company would like to know whether the mechanism is for hypoglycemia. There are many interesting and world-wide hypotheses but we don't know.

What we do is we look at the data and we see that there is a disbalance, a disproportionate number. The explanation is that maybe people don't know how to use the product. Therefore, the company has to do the studies to show us how to use the product and not allow a product that may induce very life-threatening conditions to be in the market to let physicians to figure out whether these things do happen or not.

The same with the motor vehicle accidents. This was a finding we never expected this to happen but suddenly when you look at the data, you or 11 or 12 or 14 -- I don't know how many -- on one arm and one or two on the other.

The patients were randomized and this happened in one arm and the other. If this would have happened the other way around, we would say placebo is very dangerous and may lead you to have a motor vehicle accident. Unfortunately, I think, the results were different.

DR. KOLTERMAN: Mr. Chairman, I would like to show a couple pieces of data what was asked for earlier that I think helps address this concern about hypoglycemia and in the Type 1 population.

The first slide -- slide up, please -- is one that was included in the presentation that look sat the event rates of hypoglycemia over time during the first four weeks of therapy where you see a dose dependent rise with increasing doses. Again, I call your attention as I did this morning to the data for the 30 microgram dose where there is much less of an increase in hypoglycemia.

If we now look at how patients did with that dose over time -- next slide, please. Slide up -- you see the data here that on the previous slide and you can see that there is really a very -- any increase in severe hypoglycemia needs to be taken seriously. Nothing that I say is intended to downplay this but there is a significantly less increase in severe hypoglycemia here with the 30 microgram dose which leads us to the recommendation that the primary focus for patients with Type 1 diabetes be on the 30 microgram dose and in some situations one might want to initiate therapy even at a lower dose.

DR. MANIGOSKI: Excuse me. These are events, not patients. Am I correct? Events not patients?

DR. KOLTERMAN: You are correct, Dr. Manigoski. These are events and not patients. We continue to feel as was done in the DCCT that this is the most meaningful way to look at the data. We have incidence data. We can pull up incidence data for the study.

DR. SAMPSON: Could we see the incidence data? Could we actually see the incidence data? Would you put that up, please?

DR. KOLTERMAN: Yes.

DR. CARA: Could I make a comment in the meantime? Yes, it is. If I'm not mistaken, it wouldn't surprise me that in this study you did not see hypoglycemia over time because this is one of the studies where patients were actually able to adjust their insulin.

DR. KOLTERMAN: Right. They also were able to adjust their insulin during the first four weeks during the period of initiation. Basically our recipe for addressing the issue is to initiate therapy with a lower dose and to decrease insulin with the initiation of therapy as I had mentioned in the presentation this morning.

Slide up, please. This is the incidence data for severe hypoglycemia during the first four weeks in the three Type 1 studies. I believe this pattern is similar to that that was seen with the annual event rate data.

The advantage of the annual event rate data, we believe does it really apply to this issue because the amount of patient exposure is pretty much constant across the different treatment groups.

When you look at longer periods of time, the annual event rate becomes useful to compensate for differences in exposure. Also the fact that single patients may have more than one event. Thank you.

DR. GELATO: I want to go back to a comment that you just sort of alluded to when you said that based on the data that you just showed us that you might want to start with 30 or perhaps lower.

Have you looked at lower doses because, as was pointed out before, there doesn't seem to be a dose response when you look at glycated hemoglobin. I wonder could you start with a lower dose? I mean, has that been looked at? Would that change your incidence events with hypoglycemia? Then slowly titrate patients up.

I think one of the things we are all concerned about is how do you do this. I'm getting the sense that maybe you could go with a lower dose and maybe that would help this event or the hypoglycemic problem. Do you have data for that?

DR. KOLTERMAN: There are two pieces of data that go in that direction. In an early clinical pharmacology study patients were dosed with a dose of 10 micrograms for 14 days. With the 10 microgram dose it's probably too low for the efficacy.

There was evidence of efficacy but it did not in the small number of patients achieve statistical significance. With the 10 microgram dose, there was little nausea and there was not an issue with hypoglycemia. That is one piece of data.

The other piece of data is that for business reasons the long-term open-label extension studies that we had under way in 1998 had to be canceled for financial considerations. There was, as you heard this morning, or heard allusions to this morning, an outcry amongst a good number of the participants and they felt they had something that was valuable to them being removed.

As it became available we offered to investigators, who had sizable numbers of those patients, the opportunity for another open label sort of extension, if you will, an interrupted extension study with a period in between.

That is the means by which the patients who spoke this morning appear to be gaining access to the drug. In that protocol, a number of investigators have chosen to initiate therapy with 15 micrograms for a period of one to two weeks and then escalate to 30 micrograms.

This, in many ways, is not different from the experience that we are familiar with as clinicians in terms of dealing with insulin in this patient population.

DR. GRADY: Well, I'm going to dispense with the question. Do you have something else?

DR. KOLTERMAN: Dr. Baron would like to add a comment to the last point we were making.

DR. BARON: I would like to make a comment as an endocrinologist. One of the observations that I made looking at this program was that the patients with Type 2 diabetes were given a higher dose, yet had a much lower incidence of nausea.

One has to wonder why even at lower doses patients with Type 1 diabetes had much more nausea. We don't really know the answer to that but one of the things that occurred to me was that patients with Type 1 diabetes have absolute amylin deficiency and, in this case, on average 19 years.

It's not uncommon, at least in endocrinology, that when we place a hormone there is hypersensitivity initially. Clearly in patients who are hypothyroid, for example, and elderly, we are very careful to introduce the hormone back slowly. The Type 2 patients are relatively amylin deficient but not completely amylin deficient. While there is absolutely no proof to what I say, I think to me that makes sense, again, as an endocrinologist.

DR. GRADY: Thanks. Well, I just want to compliment everybody on the display of data. I think it has really been very clear and very helpful. I think where I am now is maybe more in the discussion phase. If I could just voice my concerns, that is, it seems to me this drug does have efficacy. We are not arguing about that, but the efficacy is rather small.

I could probably get the same efficacy relatively easily by increasing insulin doses. But the advantage your drug has there in contradistinction to insulin, decreases weight and may decrease insulin levels. Both of those may be good but maybe marginally good. What we are trading off against those potential benefits are, of course, the dangers of hypoglycemia.

The three things that really worry me about what we understand about hypoglycemia with this drug are, (1) I am still really worried that there is lots of nausea associated with this drug, nausea and anorexia.

I'm worried that those two things are associated with the mechanism of hypoglycemia because many of the patients with nausea and anorexia dropped out early in the study and may subsequently have developed hypoglycemic episodes that we would not have captured because they were dropped from the study.

(2) It may be that physicians will get more experience using your drug over time, but it could also be that once patients get outside of the carefully controlled situation of a clinical trial that we will see yet more hypoglycemia.

Thirdly, and maybe most importantly, I haven't seen any evidence, at least, that your approach of decreasing insulin levels early on when you start your drug will really reduce the incidence of hypoglycemia. Maybe that will work. I don't know.

It could also be that once insulin levels are subsequently increased later in the use of your drug, that we will again see hypoglycemia. I think those are sort of the three real concerns about worries that hypoglycemia is going to be even more prevalent and more severe than it was in the clinical trial setting.

DR. KOLTERMAN: The primary consideration in that regard, I think, is that the -- I'll make an observation, cite some data, and then Dr. Dickey would like to add a comment to this.

The observation is that hypoglycemia occurred in the setting of a double-blind clinical trial where patients were being encouraged to maintain their insulin regimens constant. That's not the way any of us would practice in a patient initiating therapy.

We would know that the patient was getting the drug. The patient would know that they were getting active drug. We now know what the side effect profile is and you could take appropriate actions.

I mentioned earlier in my comments that there were two long-term open-label safety studies done with the 30 microgram dose. In the open-label setting it appears that hypoglycemia was significantly less of an issue than what it was in the 30 microgram dose treated in a double-blind manner.

The reason that I cannot pull that data for you is that I would be comparing apples to oranges and that the capture of hypoglycemia data in those open-label studies were not comparable to what was done in the double-blind study.

Now I would like to ask Dr. Dickey to make his comment.

MR. DICKEY: Thank you. I just wanted to share the concern that I have with hypoglycemia in any patient irrespective of pramlintide. You mentioned that you could improve the control of the patient by increasing insulin.

My experience in trying to do that, and I mentioned that in the patients that I entered into the trial, I think that I had done the best I could with the tools available for the treatment of those Type 1 patients, namely, insulin, diet, education, and I still was not able to achieve my goal. That is why they met the criteria for the protocol of over 8 percent.

In trying to increase the insulin in the patient to improve control, I believe that you are likely to increase the risk of hypoglycemia. It is a special concern of mine and I have published and I have currently an article in review about that very risk. Not about new drugs but about the risk of hypoglycemia as we become more aggressive in our treatment, namely with insulin.

A recent article in Diabetes Care pointed out that the -- this is no news to you, I'm sure, that the peaking of insulin, for example, of NPH, and I believe it's going to turn out to be true of Lantis as well, is not as predictable as we would like.

My concern is that in treating these patients, I can smooth the curve, as some of the patients alluded to earlier, smooth the curve, reduce the standard deviation, something that we all want to do. Reduce hypoglycemia, reduce hyperglycemia with the use of this drug.

That is one of the main reasons I'm excited about the drug. It reminds me that when I was thinking about beginning using insulin pump therapy as a solo practitioner, I was very concerned about not having another endocrinologist within 60 miles and doing insulin pumps.

I went to two or three meetings sponsored by an insulin pump maker and heard patients repeatedly say from the podium that their hypoglycemia was less on an insulin pump even though their A1c was lower. That impressed me. I believe the same thing can happen with pramlintide. That's another reason I'm excited about it.

I would beg you to consider that if we've done the best we can with other means of therapy for the Type 1 patient, namely insulin, education, and diet, and exercise, then I'm not sure that you can do enough to do better than this drug can do even though the magnitude of the reduction in A1c is relatively smaller than you would like to see. Thank you.

DR. KOLTERMAN: Mr. Chairman, if I could just remind the panel that we did present a recommendation for the initiation of therapy in the presentation this morning.

Slide up, please. This was to choose an initial dose. The 120 microgram dose appears to be appropriate for patients with Type 2 diabetes. We chose the words well or carefully, 30 micrograms or lower in patients with Type 1 diabetes.

Dose frequency determined by the patient's meal pattern. Some patients only eat two meals a day, if that is the patient's lifestyle, dosing twice a day, three times a day, or the patient eats three meals. Type 1 patient three meals and a snack, four times a day.

The drug is given 15 minutes before a meal. Insulin reduction with initiation of therapy 10 to 20 percent of the postprandial short-acting insulin dose, not dissimilar to the recommendations in the labels of other compounds that are approved for use as add-on therapy to insulin.

Next slide, please. This fits into, we think, allowing the physician the patient to compensate for both the nausea and the satiety effect based upon things that most patients and physicians are already familiar with, namely blood glucose monitoring and using this data to judge appropriate modifications to the patient's insulin regimen.

With chronic therapy this part of the loop continues, the blood glucose monitoring and appropriate adjustments of insulin to facilitate the patient's going to target. Thank you.

DR. CARA: I want to throw out a comment in response to your questions and maybe get your thoughts on it and see what you think and sort of have you tell me whether this is totally off base or not.

I think there is a way that I've struggled in trying to see this medication specifically as a diabetic medication or as a diabetes related drug. I don't think it is. I think the bottom line is this is an anoretic agent that works either through essential mechanism or perhaps through delayed gastric emptying. It's primary efficacy is in decreased food intake with some weight reduction. I think a secondary effect is actually reduction of glycosylated hemoglobin as a result of better compliance with meal plans.

As a result, there is a direct dosage response between the anoretic effect and the drug dose. The side effects are related directly to the biological activity of the medication; hypoglycemia, anorexia, and nausea.

As a result, the dose has to be tapered individually based on side effects and the clinical response of the patient. If you think of it in those terms, and you think of it in terms of that paradigm, I think it tends to work better. Does that make sense?

DR. KREISBERG: Let me tell everybody what I would like to do since I'm the Chair. It's 20 minutes to 4:00 and we have seven questions to address with subsets of each question.

What I would like to do is invite the panelists to -- I'm just going to go around the table -- to make a concise statement about your feelings about the drug, studies that you might think would shed more light on the mechanism of action.

We are going to move around. Hopefully that will be in 25 words or less. We'll then begin to vote and that will be associated with more discussion.

Having said that, can I start with you.

MS. KILLION: 25 words. Okay. That's four to five right there. Diabetics are always struggling for balance and I think that's the theme of this drug, balancing risks and benefits. Interestingly enough, I think the parallels are really high. Hypoglycemia happens. It happens to all of us, some more than others.

My concern is what can you do to minimize the risk of hypoglycemia in its severe instances in the four-week window where that risk seems to be most amplified. I really haven't heard anything. They said, well, reduce your insulin but we don't know what that effect is so we need to see something on that. Maybe that will work and wouldn't that be great but we don't know and that's a scary thing.

It's a very seductive drug for diabetics. To have a new option for Type 1 is very inspiring since we haven't had -- we've had different kinds of insulins but not anything that acts in a different way or in conjunction with.

And to have weight loss instead of weight gain, that's manna from heaven. I guess in addition to even a small reduction in your HbA1c anything that is going down is good. To have your weight loss, to use a forbidden phrase, that is icing on the cake.

But my concerns are still how do you get safely through that high risk period if you accept the notion that it does go down over time? Two, many patients -- this is a high-level concern for me, this second one. Many patients, probably most patients, do not have the level of care, the exquisite level of care that you get in a clinical study.

They are going to people who are not diabetic specialist and they are saying, "Here's a drug." You get all kinds of responses to that. I worry about that risk of hypoglycemia being amplified in a much broader population. Those are my comments.

MS. McBRAIR: I agree with everything that Rebecca has said and I think she's making some interesting comments. I certainly think that the patients themselves have the right to decide if they want to live with the increased number of injections, if they can handle the nausea side effects.

I, too, am concerned about the hypoglycemic events. I think we could study that more. I think we can help education patients and physicians that that could happen and to be more aware to make sure that they have effective monitoring in place, that they are staying on top of it now that we know that's an issue. I still think we should be looking at quality of life and if, at least, the anecdotal reports that the physicians and the patients have reported here on quality of life is true, then it's another reason to approve this drug and look to make sure we know what the indications are for some of the folks.

DR. CARA: I've already told you what I thought to a large extent but I think, too, that there are still a number of questions that need to be resolved. I'm just going to add a couple to what other people have already commented on.

I think issues related to really establishing the characteristics of the patients that are likely to respond to this medication and that might best benefit from this medication still needs to be established. I still think that we need to learn more about its dose response characteristics and its true efficacy.

I think there is a fine line between biological effect and adverse effect that still needs to be teased out. I would also like to know more about what type of insulin regimens might work best with this sort of medication. I think the other issue is trying to get at sort of more concrete evidence of intolerance to the medication other than just nausea or anorexia which are very subjective terms would be very helpful.

DR. LEVITSKY: I think the company has set the background for a drug that could very well be quite important in the management of Type 1 and Type 2 diabetes and their evidence for efficacy is good.

What I would like to see is a study which evaluated quality of life, nutritional aspects including quality and quantity of nutritional intake and change during the course. I would like to see something that looked in a chronic kind of way at maximum amplitude of glucose excursions so that you could see the effects that the patients anecdotally reported.

I would like to examine a dose which perhaps was slightly lower in the patients with Type 1 diabetes. I think it could be done in a double blind way with flexible insulin dosing without hurting the patients. It's possible that it would take a dose slightly lower than the 30 micrograms to pick this up. I think if that were done, that there could be really powerful evidence as to how to use this drug and also its safety characteristics.

DR. TAMBORLANE: I'd like to thank everybody for a very interesting day, actually, and also join in both the FDA and the company. I agree that efficacy has been demonstrated. There are the obvious concerns about safety.

However, it's my feeling that these safety issues related to hypoglycemia, although incredibly important were also predictable and potentially manageable. I think ultimately the drug is approvable. The issue has to do, like everybody else said, this first month of therapy.

It seems to me that would be a very easy and relatively quick study that could be done that would just focus in the first four weeks that you could test placebo versus drug and cut the insulin dose by 20 or 10 percent and have some sort of answer within a few weeks.

DR. GELATO: There are just two things that I like about this drug very much. One is in the Type 2 patients you get weight loss. I think that is important. The other thing that I like about it is in the Type 1 patients if you can truly smooth out their glucoses, I think that is important because I think that is a problem.

I think this drug has a lot of promise. I think other people have already stated some of the feelings that I have. I would like to see a lower dose tried first where you actually do titrate their own insulin and hopefully get around the problem of the hypoglycemia because I think the drug really does have a lot of promise.

As many people have said, for Type 1 there really isn't anything else out there. I certainly would like to see more done with it and feel, too, that it should be a drug that should get approved.

DR. SAMPSON: I would agree with the agency's analysis that efficacy has been established. It seems to me it's worthwhile. Just reiterating what I said earlier, it would be nice to see a true clinical dose response group including lower doses. The safety differences were noted in the first four weeks. I thought I heard the agency also argue that differences, though they are less, still exist beyond four weeks.

It seems to me that one would like to see perhaps a placebo controlled dose response study in maybe a setting that is more compliant with the ADA recommendations.

But also you have provided a strategy, I think, for handling hypoglycemia in the first four weeks. That certainly could be tested in a clinical trial to see if it reduced the incidence or made the comparison to placebo in terms of hypoglycemic events disappear or get very small in the first four weeks.

DR. GRADY: I think it is also a very exciting drug. I find it a little hard to think about it in terms of Type 1 and Type 2 diabetes. I think for Type 1 diabetes it's an exciting drug and we are worried that the benefits may not exceed the risks. I think all of us have this idea that somewhat more study of minimizing the risks would be use. I think particularly a randomized control trial in which insulin doses were decreased to begin with and allow to vary as they normally clinically would with very careful measurement of hypoglycemic episodes, quality of life, and glucose excursions. Those are the main things that we've heard today that are really the big benefit of the drug.

I agree while I know that you can't rip out and do that in a couple of weeks, I think it could be a fairly short-term trial perhaps.

Secondly, when I think about Type 2 diabetics it is a somewhat different situation because there I think the efficacy that you demonstrated was pretty much the same and there was very little risk. The risk benefit there is much better.

However, we have many more options for Type 2 diabetics. The idea that many of our Type 2 diabetics are going to use three or four injections a day I guess I find -- I mean, I doubt that is going to happen. Of course, we should let them make that decision.

I guess I would like to bring the committee back to it seems like everyone was talking about Type 1 diabetics and I would like to -- we are also charged with the question of what to do about approving a drug for Type 2 diabetics. I would just like to see what other folks have to say about that.

DR. KREISBERG: The Chair can --

DR. KOLTERMAN: I just wanted to make one comment, Dr. Kreisberg. That is, that the sponsor sees value in a study evaluating changes in insulin with the initiation of therapy.

Also exploring lower doses and had planned to do that and feel it would be an appropriate study to be done very soon. We would hope that it can be seen prudent to approve the drug with the understanding that that study would be done.

DR. MISBIN: Dr. Kreisberg, may I make a brief statement also? Very briefly. I think several people have alluded to the fact that this is America and people have a choice as to whether they want to take several injections and whether they will tolerate nausea or not. We completely agree with that.

I just want to point out that if you take the motor vehicle accident incidence serious as we do, that there is a four-fold risk in patients on pramlintide, then not all the adverse events would be in patients.

I think I need to remind people that in the DCCT there were three patients who died of motor vehicle accidents, one on conventional treatment, one on intensive treatment, but there was also one person who was not a patient who died as a result of an automobile accident that was attributed to hypoglycemia in the driver who was in the intensive group.

I think, yes, patients have a choice but some of the adverse events will be in people who do not have a choice and are not going to be amenable to black boxes or anything else that patients ordinarily would have the availability of.

DR. KREISBERG: Well, I would like to say that I think the idea that you could improve the effectiveness of insulin without increasing the dose is very attractive. It's a promising concept that I think needs to be pursued. I think the drug that you are investigating has a unique mode of action, although I'm not exactly sure what that mode of action is, but it clearly is different.

I think the efficacy is clear. I think it's small but that may be an understatement of its value. I think that the agency really needs to get into explaining -- I'm sorry, the sponsor needs to get into explaining the discrepancy between the public testimony and the lack of any information like that provided to us by you because I can see value in a drug that perhaps had a modest effect on the HbA1c if it did, in fact, smooth out the curve as Dr. Levitsky suggested that it might.

If it made the quality of life better for the patient, it would be a valuable drug. I think that this drug has that potential if we could understand it somewhat better than we do.

I don't want to trivialize hypoglycemia because I think it's very important but Dr. Tamborlane pointed out that hypoglycemia was a very common complication in the DCCT and I think whenever you strive to get the very best control that you can in diabetic patients, that's always one of the considerations that you have to deal with.

If I recall, close to 30 or 35 percent of the patients in the DCCT actually did have significant hypoglycemic reactions. Many of those were severe occurring in the early morning hours, as I recall.

On the other hand, I took some numbers out of your data and Dr. Grady can correct me because she knows this area much better than I. In your Type 2 diabetic patients you need to threat six or seven to get a one percent decrease in the HbA1c and you need to treat 100 to harm a patient. That's what I came up with in calculating.

I just wonder if you would take another look at your data to see whether you couldn't display it in some way that showed the relative risk benefit of that. For the Type 1 diabetics, it turned out that you had to treat eight to get 1 percent improvement in HbA1c and you had to treat 25 in order to harm a patient.

We have to decide what the risk benefit ratio is, I think, in order to determine whether the benefit is worth the risk. I think there are perhaps some more creative ways you can look at your data to see if you can cast it in more of a risk benefit type of effect.

I think there's some fine tuning that has to be done and I would heartedly endorse a study that was more real life, and that is the use of the drug with intensive management of the insulin alongside of it because that's the way in which we would hope it would be utilized in the community so that we can see whether we have to back pedal on some patients or go up on other patients.

I agree with everybody else that the initial four week period of time appears to be crucial in somehow identifying patients who might be withdrawn from the study subsequently because they show unusually unpredictable reductions in their glucose concentration.

I think it's as likely that we have missed some of the hypoglycemia because of the early dropouts. I also think it's theoretically possible that you've underestimated the benefit of the drug if you could have kept some of them in the study because there may be some relationship between the drop out and the ability to reduce the glucose concentration. I think there's lots of things that need to be done to better understand how we could select a subgroup of patients who might get the most benefit from the use of the medication.

Does anybody else want to make a comment?

DR. CARA: I just want to add one more area that I think would be interesting for you to look at that might be a little bit different from what other individuals have said, and that is that I would really encourage you to look at the efficacy of the drug alone in patients with Type 2 diabetes. That is, non-insulin requiring individuals with Type 2 diabetes.

DR. KREISBERG: If there's no further discussion, I would like to begin to ask the questions. I will read them and then we would like to go around the table. I'm told that each person who votes has to say what their name is and what their vote is.

The first question deals with efficacy and it reads, "Based on the information presented by the sponsor in the NDA, are the data adequate to establish the efficacy of pramlintide in combination with insulin for the treatment of patients with (a) Type 1 diabetes, and (b) Type 2 diabetes." Let's do Type 1 diabetes first and we'll start on my left.

MS. KILLION: Rebecca Killion. Type 1, yes.

MS. McBRAIR: Wendy McBrair. Type 1, yes. Type 2, questionable.

DR. CARA: Jose Cara. Type 1, questionable.

DR. LEVITSKY: Lynne Levitsky. Type 1, yes.

DR. TAMBORLANE: Bill Tamborlane. Type 1, yes. Type 2, yes.

DR. GELATO: Marie Gelato. Type 1, yes. Type 2, yes.

DR. SAMPSON: Allan Sampson. Type 1, yes. Type 2, yes.

DR. GRADY: Deborah Grady. Type 1, yes. Type 2, yes. I only hesitate because there were only 150 people in all of these studies as far as I can tell who were also taking metformin.

DR. KREISBERG: Bob Kreisberg. Yes for Type 1 and yes for Type 2, and I'm going to come back around.

MS. KILLION: Okay. Type 2, yes. Rebecca Killion.

DR. KREISBERG: Jose, we've got to get you to commit yourself.

DR. CARA: Jose Cara. Type 2, yes.

DR. LEVITSKY: Lynne Levitsky. Type 2, yes.

DR. KREISBERG: It says, "If the answer to either of the above is no, what additional data would be required?" It doesn't look like the answer is no to either one but I would ask is there any other data other than what has already been mentioned that anybody would like?

DR. GRADY: I think you need to study more people with Type 2 diabetes who are taking both insulin and metformin.

DR. TAMBORLANE: I would also encourage the use of newer sensor technology to look at fluctuations in future studies.

DR. KREISBERG: Question No. 2, "Based on the information presented by the sponsor in the NDA were the study designs adequate to guide physicians in the effective use of pramlintide in combination with insulin in (a) Type 1 and (b) Type 2 diabetes." Answer both at the same time this time.

MS. KILLION: Rebecca Killion. I think, as we have discussed, we need more information on both Type 1 and Type 2.

DR. KREISBERG: Does that mean no?

MS. KILLION: That would be no.

MS. McBRAIR: Wendy McBrair. For Type 1 and Type 2, no.

DR. CARA: Jose Cara. For Type 1, no. For Type 2, insulin requiring patients, yes.

DR. LEVITSKY: Lynne Levitsky. Type 1, no. Type 2, no.

DR. TAMBORLANE: Bill Tamborlane, a qualified no on Type 1 and yes on Type 2. The qualified no is that I think they have shown that they have identified reasonably effective doses. I think the issue is initial titration starting at lower insulins which we needed to know more about.

DR. GELATO: Marie Gelato. Type 1, no. Type 2, yes.

DR. SAMPSON: Allan Sampson. Type 1, no. Type 2, no.

DR. GRADY: Deborah Grady. Type 1, no. Type 2, no.

DR. KREISBERG: Bob Kreisberg. Type 1, no. Type 2, no.

MR. JENKINS: Mr. Chairman, can I ask are you planning on reading out the tally to make sure everyone kept score correctly?

DR. KREISBERG: Would you like me to do it after each question?

MR. JENKINS: Whichever you would like. I was just asking do you plan to do it because it's always questionable that everyone has the same tally.

MS. REEDY: Question 1, (a) Type 1: eight yes; one no. (b) Type 2: eight yes, one no. Question 2, (a) Type 1: no yes, nine no -- Question 2, Type 1; yes, zero; no, nine. Type 2; yes, three; no, six.

DR. KREISBERG: With regard to question No. 2, do any of the panelist want to suggest any other studies or data?

DR. LEVITSKY: In addition to what we've already -- I think I kind of laid out before what I think would be reasonable which is a broader study in patients who are being managed with multiple insulin injections or on pumps with careful attention to quality of life.

The nutritional status, amplitude of glycemic control, and flexible insulin dosing and a lower dose -- at least two doses, perhaps, in the Type 1s. In the Type 2s there were comments which are reasonable about perhaps selecting out patients who are maximally managed on other agents and then adding this.

MS. McBRAIR: Wendy McBrair. I would just like to include exercise as part of the study.

DR. CARA: Jose Cara. I want to qualify my answer yes in regards to Type 2 diabetes, and that is that it's yes for insulin requiring type patients with Type 2 diabetes and that according to the recommendations, or the proposed recommendations, and that is to start at a lower dose and/or a 20 to 30 percent decrease in the insulin dose. I don't have anything to add in regards to additional data.

DR. GRADY: Could I just say that I think No. 3 here, it doesn't seem that any of us agree that such a trial would have to aim to meet ADA guidelines for treatment of these patients.

DR. TAMBORLANE: I just want to make sure we have all our positions. My position would be that we really -- the most pressing need is the initial four week study looking at the recommendation that the sponsors would like to put in their labeling.

DR. CARA: My understanding was that the labeling applied to Type 2 patients as well. Is that correct?

DR. KOLTERMAN: Making a recommendation for the reduction in insulin dose with the initiation of therapy in Type 2.

DR. CARA: That applies to Type 2?

DR. KOLTERMAN: Correct.

DR. KREISBERG: Orville, could I ask, you have not actually done the titration study using lower doses to acclimatize the patients to see whether that would abrogate in any way the initial nausea. Is that true those have not been done?

DR. KOLTERMAN: They have not been done. A protocol has been devised and the intention was to carry that out while the final resolution of the approval process was being wrapped up.

DR. KREISBERG: I think that would be very important because it would take it out of the realm of your best recommendation to something that is evidence based.

I would also suggest that this idea about a 10 to 20 percent reduction in the insulin dose would also be better if it were evidence based in that in some of your studies you might want to look at the relative benefits in terms of reducing the frequency of hypoglycemia by a 10, 15, 20, 25 percent initial reduction during the titration period.

DR. KREISBERG: Question No. 3 deals with safety. "Based on the information presented by the sponsor and the NDA are the data adequate to define the safety profile of pramlintide when used in combination with insulin in (a) Type 1 diabetes, and (b) Type 2 diabetes."

MS. KILLION: Rebecca Killion. I would say not yet on each one which is sort of a qualified no.

MS. McBRAIR: Wendy McBrair. For most things except for the hypoglycemic events. That's a qualified yes for both.

DR. CARA: Jose Cara. I don't know if this is a qualified yes or no but the bottom line is we don't have enough data on either and I think we need to get more information regarding safety profiles in both situations.

DR. LEVITSKY: No to both. Lynne Levitsky. Same reasons.

DR. TAMBORLANE: Bill Tamborlane. I think we are saying the same thing but I think the safety is okay for other things but hypoglycemia which needs further definition. I would say that is a qualified no.

DR. GELATO: Marie Gelato. No for both.

DR. SAMPSON: Allan Sampson. No for both.

DR. GRADY: Deborah Grady. No for both.

DR. KREISBERG: Bob Kreisberg. No for Type 1. Yes for Type 2. I think everybody has expressed their concern about the issue of hypoglycemia as being the major adverse effect and I don't think anything more needs to be said about that unless you want to.

MS. REEDY: For question 3, part (a), Type 1: yes, one; no, eight. For Type 2: yes, two; no, seven.

DR. KREISBERG: Question No. 4.

DR. ORLOFF: Dr. Kreisberg, I think at this point based on the interpretation in the responses to question 3, we can dispense with question 4. Thank you. Thank me.

DR. KREISBERG: Question No. 5, approvability. "Based on the data presented in the NDA and the adequacy of the trial designs, do you recommend approval of pramlintide for use in combination with insulin in (a) Type 1 diabetes and, (b) Type 2 diabetes."

MS. KILLION: I want to call a friend. Rebecca Killion. I want this drug but I really think based on all the answers that went before, you have to say not yet.

DR. KREISBERG: Is that for both Type 1 and 2?

MS. KILLION: Yes.

MS. McBRAIR: Wendy McBrair. Same answer.

DR. CARA: Jose Cara. For Type 1 diabetes, no. For Type 2 diabetes for patients on insulin, yes.

DR. LEVITSKY: Lynne Levitsky. For Type 1, no, and for Type 2, no, not yet either.

DR. TAMBORLANE: Can I get a clarification of the question for the agency? Can you approve it with the stipulation that they do this dose finding thing or is it just based on what we have at this point? Is there dose reduction in the first four weeks?

DR. ORLOFF: We do approve drugs with Phase 4 commitments for further investigations. The approval, as is, I think, evident, is based upon a judgement of safety and efficacy that we determined from the data that have been produced so far. The question to you is there enough there to approve it and/or is the Phase 4, as we call it, really of ancillary importance to the approval?

DR. TAMBORLANE: Since I think that the first four weeks is actually a manageable problem, I'm going to vote for approval for both Type 1 and Type 2.

DR. GELATO: I'm not going to poll the audience. Marie Gelato. For Type 1, no. For Type 2, at this point I'm going to say no.

DR. SAMPSON: Allan Sampson. For Type 1, no. Type 2 is a harder decision. I think still at this point I would say no.

DR. GRADY: Deborah Grady. For Type 1, no. Again, for Type 2, I'm very borderline but I am going to vote no based on the need for more information, I think, on appropriately treated patients.

DR. KREISBERG: Bob Kreisberg. No for Type 1 not based on efficacy but based on the need for more safety information. And Type 2, yes.

MS. REEDY: On question 5, for Type 1: yes, 8; no, one. Wrong. Sorry. Upside down. On question 5 for Type 1: yes, one; no, eight. For Type 2: yes, three; no, six.

DR. KREISBERG: Question six, I think, becomes unnecessary considering the vote. Question 7 really deals with additional studies that we would like for the sponsor to consider. I think we have discussed that in pretty much detail unless anybody would add further.

DR. GRADY: I'd just like to ask the committee one question. That is, in my mind it's not clear that it would be necessary to decrease insulin doses at the time of starting this drug in Type 2 diabetics. I wonder if the company might just do studies in Type 2 diabetics taking insulin and metformin.

DR. KREISBERG: Jose.

DR. CARA: Can individuals that answered yes to any part of question 5 get a chance to just make a comment on question 6? That's regarding the manner in which hypoglycemia risk might be addressed in the labeling.

In addition to what's already been suggested by the sponsor, I would advise them to make strong warning about the anorexia and the nausea and indicate how that might be implicated in the hypoglycemia that is seen with the medication.

DR. KREISBERG: I would just like to say that, again, this drug has considerable promise and the concept of its uses is very attractive. I disagree with the agency in thinking that the target is how we should evaluate the efficacy of this drug. I think that everything that we've done in diabetes, at least for Type 2 diabetes, is looking for incremental improvement in glycemic control and that the use of drugs in combination or triple therapy gives us an opportunity to augment the efficacy of the previous therapy in patients.

I see this as playing the same role both in Type 2 and hopefully in patients with Type 1 diabetes. I hope that the sponsor is not discouraged by the response of the committee because I think I sense a great deal of enthusiasm from the members of the committee just a sense that there isn't yet enough information to vote with conviction.

Our experience in the past with adverse side effects of drugs makes us skiddish about doing that. I would like Orville to tell us anything that he would like to tell us as a parting shot.

DR. KOLTERMAN: I actually would like to pose a question to the committee. I understand the issue that has caused you concern with regards to Type 1 diabetes. This initiation of therapy and actually the desire to see a demonstration that our recommended approach to the initiation of therapy reduces both the incidence of nausea and, most importantly, the incidence of the severe hypoglycemic reactions.

I did not hear a clear directive in terms of what additional information is desired in the Type 2 population. As I listened to your comments, I didn't hear a similar concern.

In fact, I just heard a recommendation from one panel member that perhaps we don't -- that there is not an indication to reduce the insulin dose with the initiation of therapy in Type 2 patients as we had suggested.

DR. GRADY: You know, I think my concern with the Type 2 is more on the efficacy end because, you know, you did show that nice slide with the similar reduction in HbA1c in Type 2 but, again, it was based on 150 people.

Secondly, it's my understanding that in the Type 2 studies there wasn't a really careful ascertainment or adjudication of hypoglycemia. Is that true?

DR. KOLTERMAN: That is not true. That is true in the initial study with Type 2 diabetes. The second two studies in patients with Type 2 diabetes were done to -- the ascertainment and evaluation of hypoglycemia were done to the same criteria and to the same standard as done in patients with Type 1 diabetes.

That data, as I showed in my presentation, resulted in an increase. There was an increase -- there was no question about it -- during the first four weeks of therapy from .2 to .5. Then after that initial four weeks there was no difference in hypoglycemia.

DR. GRADY: But the absolute effect on hypoglycemia is really small. I am most concerned just that you demonstrate a little bit better that it would be effective in patients the way they are treated today typically with metformin in addition to insulin.

DR. TAMBORLANE: I was just a little confused. There was still a relatively as impressive a problem the first four weeks with the Type 2. I mean, in absolute terms it was a lot less but there was still this problem the first four weeks. That is why -- I mean, my personal recommendation would be to test it out in the first four weeks in the Type 2 also.

DR. KREISBERG: Other comments? We're adjourned.

(Whereupon, at 4:25 p.m. the meeting was adjourned.)