DEPARTMENT OF HEALTH AND HUMAN SERVICES
FOOD AND DRUG ADMINISTRATION
CENTER FOR DRUG EVALUATION AND RESEARCH

DERMATOLOGIC AND OPHTHALMIC DRUGS

ADVISORY COMMITTEE

Friday, August 27, 2004
8:00 a.m.
5630 Fishers Lane
Room 1056
Rockville, Maryland 20852

Jennifer A. Dunbar, M.D., Acting Chair
Kimberly Littleton Topper, M.S.

MEMBERS:

Paula L. Knudson
William Gates, M.D.

CONSULTANTS (VOTING):

Scott M. Steidl, M.D.
Jeffrey Lehmer, M.D.
Vernon Chinchilli, Ph.D.

CENTER FOR DEVICES AND RADIOLOGICAL HEALTH OPHTHALMIC DEVICES PANEL MEMBER (VOTING):

Jose S. Pulido, M.D., M.S.

PATIENT REPRESENTATIVE (VOTING):

Elaine King Miller, Ph.D.

INDUSTRY REPRESENTATIVE (NON-VOTING):

Peter A. Kresel, M.B.A.

FDA STAFF:

Jonca Bull, M.D.
Wiley A. Chambers, M.D.
Jennifer D. Harris, M.D.

4

DR. DUNBAR: I would like to call the Dermatologic and Ophthalmic Drugs Advisory Committee meeting to order to review NDA 21-756, for Macugen, and I would like the committee members to introduce themselves. I am Jennifer Dunbar, from Loma Linda, California, and I would like the committee members, starting on my left, to introduce themselves.

DR. GATES: I am William Gates, from Nashville, Tennessee.

DR. LEHMER: I am Jeffrey Lehmer, from Bakersfield, California.

DR. PULIDO: Jose Pulido, Rochester, Minnesota.

DR. STEIDL: Scott Steidl. I am a retina specialist from the University of Maryland, in Baltimore.

MS. KNUDSON: Paula Knudson, with the Texas Health Science Center, in Houston.

DR. CHINCHILLI: Vern Chinchilli, Penn

5

State Hershey Medical Center.

DR. BULL: Good morning, Jonca Bill, Director of the Office of Drug Evaluation V, in the Office of New Drugs here, at FDA.

DR. CHAMBERS: Wiley Chambers, Deputy Director for the Division of Anti-Inflammatory, Analgesic and Ophthalmic Drug Products.

DR. HARRIS: Jennifer Harris, medical Officer, same division.

MR. KRESEL: Peter Kresel. I am the industry representative, Irvine, California.

MS. TOPPER: Kimberly Topper, FDA, the Executive Secretary for the committee.

DR. MILLER: Elaine King Miller, Amarillo, Texas.

DR. DUNBAR: Now we will ask Ms. Topper to read the conflict of interest statement.

MS. TOPPER: 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 even the appearance of such at

6

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

Dr. Jennifer Dunbar has been grated a waiver under 18 U.S.C. 208(b)(3) and 21 U.S.C. 505(n) for her spouse's ownership of stock of the sponsor. The stock is valued from between $25,001 and $50,000.

Dr. Jose Pulido has been grated a waiver under 21 U.S.C. 505(n) for his children's ownership of stock in the sponsor. The stock is valued from $5,001 to $25,000.

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

In the event that the discussions involve any other products or firms not already on the

7

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.

We would also like to note that Dr. Peter Kresel has been invited to participate as a non-voting industry representative. Dr. Kresel is employed by Allergan.

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. Thank you.

DR. DUNBAR: Now we will ask Dr. Chambers to give an introduction of the issues that we will review today.

DR. CHAMBERS: Thank you, Dr. Dunbar. Let me start with welcoming everybody. Good morning. I want to particularly welcome the advisory committee members, and the time that they have taken both to review the material and to both

8

travel and attend today.

[Slide]

We are here today to discuss Macugen, and this is the Dermatology and Ophthalmology Advisory Committee meeting. Those of you who think you should be some place else, we would welcome the open seats if you want to give them up.

My name is Wiley Chambers. I am the Deputy Director for the Division of Anti-Inflammatory, Analgesic and Ophthalmologic Drug Products, and it is our Division within the Office of Drug Evaluation V that will be reviewing this application today.

[Slide]

This application, unlike many others--or at least the section that we will be reviewing today, unlike many others, is part of the continuous marketing application Pilot 1 NDA submission which was part of PDUFA 3, which is the Prescription Drug User Fee Act that was enacted into law in 2002. This allowed for the presubmission of individual modules in different

9

sections that would then be reviewed, and comments given back. This would not be a final action but it would be comments on a particular section, with the goal of speeding ultimate approval of particular applications by being able to give interactive comments early on. The action on the actual NDA will only be taken after all the modules are submitted and reviewed.

[Slide]

Today's discussion is clinical only. We are only dealing with the clinical section. We are not dealing with the pharm. tox. section. We are not dealing with the chemistry manufacturing section. So, no one should expect that we will take an action on this NDA today, tomorrow or the next day because there are other modules which are being reviewed in their own time course.

The expectation is that we will give comments back to the sponsor of the application within approximately six months of the time when the module was submitted, and so we have scheduled this meeting to deal with the clinical issues and

10

our clinical feedback. As you will hear later on, we have particular questions that are geared toward this application, but we are looking primarily to see have we missed anything; are there other areas, while we are still within the review period, that we should be looking at further, or are there issues that you think need to be further explored before an application would be acted on one way or the other?

[Slide]

I am going to spend some time today going through basic clinical trial design issues for products for macular degeneration in general.

[Slide]

The Division gives guidance as trials are performed on a way to do a particular trial. We don't believe there is a single method to do all clinical trials. We have tried to give what we think is a good way to do trials that will give answers that we can then interpret. We clearly recognize that there may be additional ways and there may be reasons to have variance from what we

11

recommend. But just so that everybody is in the same page, I am going to go through what we generally recommend to sponsors of trials so you know where there are potential differences, which you may either agree with or disagree with, but more for informational purposes.

We ask that trials be parallel on design trials; randomized by person as opposed to randomized by eye; double-masked, meaning at least the investigator and the patient are masked to what treatment they are receiving; and to try to incorporate dose ranging within the study development plan. That does not mean every trial but it means that there be an exploration to dose ranging.

[Slide]

The inclusion criteria for at least wet macular degeneration, using that term as broad as that is, is that we expect patients to have choroidal neovascularization documented by fundus photography and/or angiography. We expect there to be specific observable features, including

12

membranes greater than a particular defined size and with particular diagnostic features such as leaking on fluorescein, such as leaking on indocyanine green or ICG, but define a particular population for which we could then label the product.

[Slide]

We try to get the trials in total to be as general as possible while still identifying a population that the product works for. Patients with concurrent ocular diseases that may be associated with choroidal neovascularization we think should be excluded to avoid any kind of confounding issues. In this particular case that generally means excluding people with presumed ocular histoplasmosis and excluding high myopia, primarily because these things can also cause choroidal neovascularization and we want to try and figure out which disease the product is working on.

[Slide]

We ask for replication. So, we want safety and efficacy, supported by at least two

13

independent trials of at least two years duration. We are looking for robustness in the findings. We want independent trials, and to that extent we mean geographically separate so that we know the product does not just work in Washington, D.C. or does not just work in Boston or one particular city where the water supply is unique. These trials conducted to date were each multicenter trials and so, obviously, clearly meet that criteria.

Actually, before I go on let me say one thing about the two-year trial. We have asked for trials to go on for two years and we have had discussions at this advisory committee before about how long trials should go on for. We have recognized that endpoints may be acceptable at a one-year time point but we have asked that trials continue on for two years. So, while you may not hear two-year data, you can rest assured that the trial will continue to go on for two years and we will ultimately have that information which we will factor into our decision. But we believe that, because of the age of the population, one year is a

14

significant portion in the rest of their lives. Consequently, if the product is showing benefit at one year we believe we could potentially approve a product and label it as working for however long it works for, but we think that duration needs to be at least one year, but have not been wedded to anything more than that. If you end up disagreeing with that, as with anything that I say today, please feel free to make those comments to us.

[Slide]

The clinical trial program we think should be able to identify adverse events that occur at least at a one percent adverse reaction rate. People may argue that one percent is too low, too high. It is, for lack of a better figure, what we have picked. That means you need at least 300 patients studied fully through that to be able to determine that. We generally recommend at least 500 patients so that we are not dealing with, "well, I've got 299" or "I've got 298" or "I've got 301." We know in this population, because of the natural age and normal life span, people are not

15

going to necessarily survive through the trial--just not related to the drug but related to other reasons. So, we start out asking for people to do trials of 500 patients or more.

We like the concentration to be studied that is going to be marketed, we like concentrations that are above what is going to be marketed to be studied to try and exaggerate potential adverse events so that we can get a handle of potential adverse events that may occur, even if they are not going to occur on the final product that is approved, so that we have some idea of what to look for. And, we would like the frequency of dosing to be at least as frequent as proposed for marketing. You will see in the trials we discuss today dose-ranging studies that look at different concentrations.

[Slide]

The duration, as I mentioned, should be at least 24 months but, as I also said, the endpoint could be as short as 12 months.

[Slide]

16

We do not require multicenter trials. It is certainly easier to enroll larger number of patients with multicenter trials. Our preference, if a company is going to do a multicenter trial, is that there be at least 10 patients per arm per center. We have set that number so that we can look at investigator interaction. Now, that is frequently a difficult thing, to enroll that many patients per arm per center, particularly if you have a multi-arm study and you are doing dose ranging. That dramatically increases the number of patients you would have at a particular center.

You need to recognize that if we do not have that many we are probably not going to be able to look for investigator interaction at any one particular center. We will do some other things to look at that question but to get a true, you know, is there one investigator that is disproportionate to other investigators really requires more patients than you will see in these particular trials. This is not an uncommon problem that we have. We don't have a solution. Generally, if you

17

are able to enroll a large number of patients at any one center you probably wouldn't do a multicenter trial. So, again, we welcome suggestions on how to get around this.

[Slide]

Stratification is not necessary. If there is a chance of imbalance in factors that someone believes may influence the results, and in this case there have been discussions about whether occult versus classic potentially would influence the results or whether baseline visual acuity would potentially influence the results. We have suggested that people stratify so that they have a higher chance of having an equal distribution between the individual groups--again, not required. The hope is that randomization will take care of it but stratification frequently helps.

[Slide]

Control agent--we have asked that at least one of the clinical trials that is performed demonstrates superiority to a control. We have not defined what that control has to be. It could be

18

the vehicle; it could be a sham; it could be a lower dose; it could be a different product. By saying at least one trial has to demonstrate superiority, that means we also potentially would accept an equivalence trial. In today's discussion we are going to deal primarily with superiority trials but, recognize, we potentially would accept either a superiority trial or an equivalence trial.

We prefer a vehicle control given our druthers of different choices, and we prefer that because it minimizes the bias. There is some animal evidence--we are not aware of any human evidence to date but there is some animal evidence that mechanical manipulation may initiate inflammatory mediators that may help the condition. Consequently, by not having something that simulates that same pathway, there may be some influence going on by the way you deliver the product, in this case the intravitreal injection, that may be a positive effect. But there are ethical issues, and I am sure we will probably get into some of that, with giving vehicle controls.

19

One of the most common reasons cited for not giving a vehicle control is the risk of endophthalmitis. We recognize that there is a theoretical risk of getting endophthalmitis in the vehicle group. The clinical trials that were performed here had cases of endophthalmitis that were in the active control group.

I just want to be on record for stating that, to the agency's knowledge, we have not had a case of endophthalmitis in the vehicle control group in any trial that has run that, and there have been trials that have run it. So, we continue to think it is not unethical to run a vehicle control. Should we get an endophthalmitis case, which I am not hoping for anyone, we may change that opinion but at the present time we continue to recommend vehicle controlled trials.

We do reluctantly accept sham controls, but we have put a condition any time we have accepted sham controls, and that has been that we have wanted additional doses, in other words, more than one dose tested to try to aid in the masking

20

of the trial. You will see that in the case of these trials today there are multiple doses, in addition to the sham, that is conducted. Again, we recognize that having a sham increases the chance of bias influence in the results, although just having a sham does not necessarily create bias.

[Slide]

Dose ranging--we prefer to try and bracket the dose that will ultimately be marketed, in other words, study doses that are higher and study doses that are lower than that which will be ultimately marketed so we get a better understanding of the drug product.

[Slide]

Efficacy has been discussed a lot. We have a number of parameters that we readily accept as being acceptable. We have other parameters which we think may in the future be acceptable or we will be willing to entertain if there is validation, and validation does not necessarily need to occur in this particular trial. The thing that we readily accept as being important is a

21

change in visual function. So, our guidance to people when we are having discussions about clinical trials is that there be statistical significance in clinical relevance in visual function at more than one time point. By visual function we mean visual acuity, visual fields or color vision.

[Slide]

The evaluations we expect to be carried out include, obviously, best corrected distance visual acuity. By that, we generally mean using a chart that has equal number of letters per line and equal spacing between lines. The ETDRS is one type of chart that meets that, and based on the validation information that was conducted at a four meter distance so that is our preferred both distance and test but we are willing to recognize other equivalent tests of best corrected distance visual acuity.

We expect best corrected visual acuity to be measured at every visit, and we expect those visits to occur no less frequently than every three

22

months.

We expect to have dilated seven field fundus photography sometime during the trial. We expect to have fluorescein or indocyanine green depending on what exactly is being studied during the trial, and we have not specified exactly when that has to be. We expect dilated ophthalmoscopy to be performed both for evaluation and for safety at every visit. We expect a dilated slit lamp exam for the same reason. We expect to have endothelial cell counts, not necessarily in every trial but somewhere within the development plan, and have at least one study that includes it at the beginning and end of the trial, and the same thing standard systemic clinical and laboratory evaluations.

[Slide]

Two meters versus four meters has been a source of a lot of controversy. It is my understanding it stems primarily from the practicality of being able to have exam rooms that are four meters. In my father's day and age, it would have required 20 foot length and his exam

23

rooms were set up to do that. That is not the current trend now. People use exam lanes that are much shorter. But the subject has been studied. It was the source of a lot of discussion in the past, and there is a paper that set out four meters as a standard that was published in Ophthalmology in 1996 for exactly the purpose of discussing what the best distance is.

It does not mean that you can't theoretically correct. You know, two meters, four meters--you can use the same distance and make the charts smaller so you are looking at the same angle that gets subtended. The issue is the variability that occurs when measuring at two meters versus four meters and the potential for any bias if the patient is allowed to lean. Now, if we would strap down or lock every patient into an exam seat and never let them move at all, it probably wouldn't be an issue but we don't do that. Just so people get a feel, at a two meter distance 17 inches is equal to one of one line. Those of you sitting in the various seats, if you are leaning backward or

24

leaning forward, just sitting in your same seat can easily do 17 inches. We don't have any reason to believe that people are attempting to bias the results or attempting to lean, and visual acuity is a very common measure in ophthalmology so everybody is aware to try to keep people from leaning or keep that from influencing what goes on. But studies have been done that show poor reliability at one meter versus four meters. So, the assumption is that there is also more variability at two meters than there would be at four meters. The overall impact on a particular trial is not known, and the only way to know that for sure would be to do both two meters and four meters, which we do not have data to discuss today.

We think it is more significant for those trials that have a feature that allows there to be a potential in masking, such as sham. We think it is more of an issue in an equivalence trial than it is in a superiority trial. These trials that we are talking of today are superiority trials; they are not equivalence trials. But there are issues.

25

[Slide]

Our recommended endpoints to date have all been, as I mentioned earlier, visual function. We think at some point in the future we will end up accepting anatomical changes but we have not yet found anatomical changes that correlate directly with visual function. So, currently we readily accept doubling of the visual angle, which on the ETDRS chart at four meters would be 15 letters or more; a halving of the visual angle, in other words, showing improvement in vision; a quadrupling of the visual angle, which would be 30 letters or more. These are all looking at percentage of patients that have this particular finding because we think a doubling of the visual angle is a clinically significant difference that would not occur within the variation of day-to-day visits.

[Slide]

We have also been willing to accept a difference in the group mean. We do not know exactly how much of a difference in group mean would be clinically significant so for

26

consistency's sake we have said we will readily accept a mean change of 15 letters. That does not mean that something less than that may not be statistically significant. We are just not ready to accept without question anything less than 15 letters.

[Slide]

Let me just briefly talk about equivalence trials just so you know the full scope of what we have talked about with individual sponsors. We believe it is possible to do comparison with an active agent which has already demonstrated repeated success. Visudyne is currently approved for predominantly classic choroidal neovascularization in atrial macular degeneration and a couple of other things. So, for that particular indication we would accept an equivalence trials if one wanted to conduct it. The way we have set up equivalence trials is that we have asked that at least 50 percent of the established treatment effect be preserved so that 95 percent confidence intervals be drawn around

27

those parameters to protect at least 50 percent of the treatment effect. Again, it is not a particular issue for this product but it may be an issue for other products.

The analyses that we ask to be conducted always include intent-to-treat with last observation carried forward and per-protocol with observed values only. We recognize these as differences in the data available for analysis. The intent-to-treat last observation carried forward is the fullest data set we can obtain. It is everybody that was randomized in the trial and it is creating a value for everyone whether real of extrapolated. A per-protocol analysis is the minimal data set. It is only those patients that fully met the protocol and only the values that we have there.

We don't believe that either one of these two analyses is the best analysis or is the most proper or is the most representative. We think they are extremes and we ask that both be conducted and we look for differences between these two

28

analyses. If there are no differences between these analyses we assume that, regardless of how much inclusion/exclusion, your results are pretty much the same and you can accept either one. If there are differences we ask for additional analyses to try and explore which one is likely to be telling a better picture or why it is telling a different picture.

Other analyses which you would have seen in the briefing package include things like worst-case analyses where we treat all dropouts in the control as being successes and all dropouts in the test product as being failures. This is not a correct test. This is not an accurate test. We are making assumptions in the worst direction to look and see how robust the findings are. We don't expect the product to win on a worst-case analysis, but it does give us an idea of what the limits of potential analysis results could be.

[Slide]

As a general rule, we ask for alphas to be 0.05. This is the common 5 percent for two-tailed.

29

In other words, p is less than 0.5. We ask for power to detect a difference to be 80 percent or greater, and we ask that any time anybody looks at the data, any kind of look any time during the evaluation that there be an adjustment in the statistical plan, in other words, correction for that alpha for any look that occurs. All of our analyses that you see in any of our data sets will include these features.

[Slide]

The last one I want to talk about is pediatrics. There is an agency initiative to try and include, when possible, pediatric patients in the drug development of particular products. So, I am covering it for completeness. In this particular case, choroidal neovascularization is rarely seen in pediatric populations and we have not asked the sponsor of this application or any of the applications that just deal with choroidal neovascularization to include pediatric patients because the population we don't think is relevant in this particular case. But as a general rule we

30

do ask for pediatric patients to be included during the development.

I am happy to take any questions and, again, I thank everybody for your time, and look forward to a fruitful discussion.

DR. DUNBAR: Are there any questions at this point regarding Dr. Chambers' presentation? If not, at this point then I would like to open the forum for the sponsor, Eyetech Pharmaceuticals and I will ask that the sponsor introduce each of their speakers within their presentation.

DR. DYER: Good morning.

[Slide]

Today we will discuss the first anti-VEGF therapy for the eye and the first treatment to target the underlying biology of neovascular age-related macular degeneration. Pegaptanib sodium achieved statistical significance for clinically meaningful, prespecified primary endpoint in replicate trials with strong supportive

31

data in secondary endpoints.

The efficacy was against usual care controls, and this pharmacological agent also shows a favorable safety profile and provides a treatment benefit to many patients for whom no effective therapy presently exists.

[Slide]

My name is David Guyer. I am from Eyetech Pharmaceuticals. I previously was professor and chairman of ophthalmology at the N.Y. School of Medicine and a practicing ophthalmologist specializing in macular degeneration.

Also speaking today will be Dr. Tony Adamis, who was an ophthalmologist on the full-time faculty at Harvard, and is now with Eyetech. He ran the ocular angiogenesis laboratory as well. Our risk/benefit section will be presented by Prof. Don D'Amico, from Mass. Eye and Ear Infirmary at Harvard.

[Slide]

Neovascular age-related macular degeneration represents 90 percent of the severe

32

vision loss from this disease. Many patients note a loss of independence and inability to read, to ambulate and to recognize faces of their loved ones. This occurs because when the disease forms abnormal blood vessels that leak blood and fluid waviness or blurred vision can be seen in the central area that sometimes can lead to a scotoma or blind area centrally that prevents them from seeing straight ahead, and in up to a third of patients clinical depression can be noted.

[Slide]

The devastating effects of this disease were well summarized in a book by Henry Grunwald, who was the former editor-in-chief of Time Magazine and U.S. ambassador. In the book, "Twilight: Losing Sight, Gaining Insight" Mr. Grunwald said, "after a lifetime during which reading and writing have been as natural and necessary as breathing, I now feel the visual equivalent of struggling for breath."

[Slide]

Macular degeneration represents a major

33

public health problem and urgent unmet medical need. It is the most common cause of irreversible, severe blindness in developed countries. Ninety-five percent of retinal specialists believe that macular degeneration represents an epidemic, and there are 200,000 new cases a year in the United States alone, and a prevalence of up to 1.6 million patients with active bleeding. Limited treatments are available and 85 percent of retinal specialists are dissatisfied with current treatment options.

[Slide]

Macular degeneration represents a progressive disease. Early on in the disease these whitish-yellow spots, called drusen, occur and patients can progress to the neovascular form of the disease which is where pegaptanib is effective. This is an angiogenic disorder and what happens is abnormal blood vessels grow behind the retina where they leak blood and fluid, as depicted here, and, untreated, they lead to disciform scarring where fibrovascular tissue destroys and replaces the

34

normal rods and cones in the retina. At this point, usually moderate to severe visual loss is noted.

[Slide]

Let's discuss the therapeutic options available for patients with neovascular macular degeneration. In the 1980s, the Macular Photocoagulation Study Group showed the beneficial roles of thermal laser photocoagulation. However, very few patients are suitable for this treatment. The treatment is most suitable when the abnormal blood vessel, as seen here on a fluorescein angiogram, is away from the center of the macula, in what we call extrafoveal or juxtafoveal location, because for patients where the blood vessel is dead center or subfoveal the laser scar itself can destroy the very tissue we are trying to save. Unfortunately, most patients with neovascular macular degeneration have subfoveal disease where the blood vessel is dead center.

[Slide]

In the year 2000, photodynamic therapy, or

35

PDT, was FDA approved for patients with subfoveal predominantly classic angiographic subtype. Thus, for approximately three-quarters of patients with neovascular macular degeneration there is no FDA approved therapy, although there is off-label use, with some limited CMS reimbursement, presently.

Today we will discuss the first anti-VEGF therapy for the eye, a pharmacological treatment that targets the protein VEGF that is responsible for the hallmarks of all choroidal neovascularization. Increased levels of VEGF lead to neovascularization and increased permeability, which lead to the clinical features of all choroidal neovascularization, and pegaptanib blocks VEGF.

[Slide]

VEGF is the common denominator for neovascular macular degeneration. Numerous peer reviewed papers have shown that for all angiographic subtypes, by immunohistochemistry staining, VEGF is present in both autopsy and surgical specimens.

36

[Slide]

Pegaptanib sodium is a pegylated modified oligonucleotide. It has a selective vascular endothelial growth factor antagonist to isoform 165. Tony in just a few minutes. It is a sterile aqueous solution in a single-use, pre-filled syringe, which is important for safety reasons. The recommended dose is 0.3 mg of intravitreous injection administered once every 6 weeks.

[Slide]

We will show you today that pegaptanib met a clinically meaningful primary efficacy endpoint with statistical significance in replicate, well-controlled clinical trials, with a favorable safety profile.

[Slide]

I will now ask Tony Adamis to discuss a VEGF overview and macular degeneration pathophysiology.

DR. ADAMIS: Thank you, David and good

37

morning.

[Slide]

In 1971 Judah Folkman first proposed the targeting of a specific angiogenic factor as a way to treat disease, and specifically a way to treat cancer and ophthalmic disease.

[Slide]

It was in 2004, with the completion of pivotal Phase III trials using Avastin which blocks VEGF that this theory was in a definitive fashion proven correct. This drug now was approved this year as a first-line therapy for colon cancer. So, we entered this era of biological anti-angiogenesis therapy.

[Slide]

The target in that trial and in our trial is VEGF, which is an acronym for vascular endothelial growth factor. Prior to that it was called vascular permeability factor. Unlike many other growth factor names, these two are very appropriate in the sense that they describe the central biological functions of this protein. VEGF

38

makes vessels very leaky and VEGF makes vessels grow. The leaky aspect of it was discovered in 1983 by Harold Dvorak and then the neovascularization aspect or biology of VEGF was discovered by Napoleon Ferrara, who has been a leader in this area, and Dan Connolly, in 1989.

Since then, if one conducts a MEDLINE search, there have been over 11,000 published peer reviewed articles on VEGF. There is a large body of knowledge concerning this growth factor. I show you just one example of that here. This is the protein structure of VEGF. We now can determine very precise structure-functional relationships.

[Slide]

The disease we are here to discuss, as David said, is age-related macular degeneration, a very prevalent disease in our society and a very complex one scientifically when one begins to study it. We are beginning to unravel the earlier stages of the disease, the stages where Bruch's membrane is altered and gives you those yellow spots, the drusen that David showed you in a clinical

39

photograph. We are also starting to understand the complex interaction of the different cell layers with the vasculature. But the area or the phase of the disease, the late phase of the disease that we are studying is the neovascular phase where vessels begin to grow up towards the retina. These vessels are abnormal and leaky, and they leak fluid and lipid and they damage the photoreceptors which sense light, and people lose vision and go blind. This process, the angiogenic process, has been very well studied.

[Slide]

As David said, the data indicate that it is biologically plausible that blocking VEGF would have a beneficial effect in this disease in a broad population. When one looks at surgical specimens or autopsy specimens of patients with the disease, what is seen is that the common denominator is VEGF. It is present in all angiographic subtypes and it is present in all active stages of the disease. So, therefore, the hypothesis that blocking VEGF in neovascular MD would have a

40

broad-base effect has some broad biological plausibility.

[Slide]

But those are not the only data that we have. There is a large body of preclinical evidence, roughly 15 years worth, which is summarized on one slide here. Let me just briefly walk you through it. In preclinical models of vessel growth in the cornea, in the iris, in the retina and in the choroid, if one gives a VEGF inhibitor you can prevent the growth of vessels and you can prevent the leak that is associated with those vessels. So, VEGF seems to be required for those processes.

Similarly, if one looks at those normal tissues and now introduces VEGF into the system, either by injecting the protein or genetically over-expressing it, VEGF in and of itself is sufficient to produce the neovascularization or leak that can occur in these tissues.

Then, so that we have some context in which to interpret those preclinical data, surgical

41

specimens and autopsy specimens from humans with actual corneal neovascularization, iris neovascularization, retinal and choroidal neovascularization show that VEGF is expressed at high levels in those tissues at the time when the vessels are growing and leaking. So, the totality of the data supports this approach of blocking VEGF in specifically the disease under study today, age-related macular degeneration.

[Slide]

It gets a little more complicated in the sense that VEGF really refers to a family of related molecules, and I want to talk about one specifically, VEGF 165 which is the target of pegaptanib.

[Slide]

We were faced with the paradox a few years ago, as we looked at the accumulated data concerning the role of VEGF in disease and in the normal state. What we found was that VEGF is required for the normal formation development of vessels during development throughout the body. I

42

am just showing you here two examples. These are the vessels of the normal colon and these, obviously, are the normal vessels of the retina.

[Slide]

In the same molecule, VEGF was shown in a number of definitive studies and laboratories around the world that VEGF is required for the abnormal vessels that can grow in the colon, and this is colon carcinoma, and here is a case of age-related macular degeneration. So, how is it that the same protein can cause these vastly different phenotypes, these different types of vessels? One set of vessels are normal and they don't leak and they behave appropriately; another set looked very different and they behave very differently.

[Slide]

Perhaps, we thought, some of that complexity is encoded in these different isoforms. Let me just explain what those are. There is one VEGF gene but that gene encodes multiple transcripts or mRNAs for VEGF that have different

43

sizes that translate into different proteins. So, one of those major proteins or isoforms is VEGF 165, which just simply means that it is composed of 165 amino acids. Another major isoform, especially in the eye, is VEGF 121. We asked the question could it be that differential expression or synthesis of these isoforms underlies the complexity that we see in the vessels in the normal and the diseased state?

[Slide]

So, in an experiment we conducted and published last year, we studied the retinal vessels. We studied the normal retinal vessels that are developing as the retina forms and we studied abnormal retinal vessels in a model of retinopathy prematurity. This is a model where vessels grow towards the vitreous and leak and are distinctly abnormal.

What we saw was that when normal vessels are developing the isoform expression of the two major isoforms, 120 and 164 which are the rodent counterparts to human 121 and 165, is roughly equal

44

during development. But rather strikingly, during disease when disease vessels are growing there is a shift to almost exclusive expression of the 164 isoform. So, it was an interesting association that we saw of 164 with diseased vessels.

[Slide]

But to really get at the causality of 164 in the production of diseased vessels we conducted the following experiment. In a model of abnormal vessel growth we gave pegaptanib which blocks just 164 and compared it to a non-selective VEGF inhibitor which blocks all the isoforms. We saw that bpth were equally effective in preventing abnormal vessel growth. Here is the control with the abnormal vessels, and both are pretty good at inhibiting that.

We also looked in a model of normal retinal vessel development and, again, gave pegaptanib and what we saw was essentially zero inhibition of normal vessels. We did not affect normal vessels. Whereas, the non-selective VEGF inhibitor had a deleterious effect on these normal

45

vessels in the retina. So, the conclusion we made was that VEGF 164 may be preferentially associated with disease and targeting it gives you a much more selective inhibition in that you are much less likely to affect normal vessels in the developing animal. But I will tell you that there has subsequently been independent support of this, specifically from UCSF, where this is also perhaps true in the adult animal.

[Slide]

To be certain of our conclusion because we used a reagent here, pegaptanib in particular, we wanted to make sure this conclusion was robust. So, we created animals genetically that where we deleted specifically the 164 isoform and these animals were able to make all the other types of VEGFs. What we see here is that these animals have completely normal retinas and normal retinal vessels and they are no different than animals that make all VEGF isoforms. In fact, these animals grow up to a normal age. They can reproduce. There are no abnormalities we can detect, even

46

though they cannot make any VEGF 164.

[Slide]

So, how was a drug made that specifically blocked VEGF 164? Well, pegaptanib is an oligonucleotide aptamer. It specifically is 28 nucleotide in life. Aptamers are molecules that will fold in a very specific fashion. They have a three-dimensional conformation such that they will bind to the target protein of interest--in this case it is VEGF--in a highly specific manner, and in the case of pegaptanib with a very high affinity. This binding occurs extracellularly. The drug does not enter the cell. It is all happening outside the cell, which is where VEGF is residing. These features make it act very much like an antibody but there are some important distinctions, aside from it not being an antibody; it is an oligonucleotide.

This class of molecules, in the published literature and it has been our experience as well, are quite non-immunogenic. In our preclinical and in our clinical examination of pegaptanib we have

47

not seen a single instance when an antibody is raised to it. And, as I alluded to, they have this remarkable target specificity and this simply attests to that.

[Slide]

This shows that pegaptanib is very efficiently binding to human VEGF 165 and murine or mouse VEGF 164, but there is no significant, or essentially no binding to VEGF 121 or related family member of placental growth factor.

[Slide]

So, what we would expect when pegaptanib is administered to the eye is that you would have selective VEGF inhibition of 165 which was associated with pathology and in our animal model spares the normal vasculature, and we would have two very important biological responses as a function of that blockade: vessel growth would be inhibited, as would permeability, and the thinking was this would translate to a better visual outcome.

[Slide]

48

The last thing I would like to talk about is how we chose our dose. This drug is administered to the eye nine times a year, and there are three doses that we chose.

[Slide]

Let me show you the data that we had in hand when we were planning these trials. We knew from our pharmacokinetic experiments that when pegaptanib is given to the eye via intravitreous injection it slowly exits the eye and it can be measured in the plasma. Actually, the plasma levels mirror the levels that one sees in the vitreous. So, by sampling the blood you can infer what is happening in the eye.

The other important thing that we learned here is that when the drug exits the eye, at least in this rabbit model, you have thousand-fold less concentration in the plasma than you do in the eye. In a more relevant primate model we saw that this held up in the sense that it was 800 times less in the plasma than it was in the eye.

[Slide]

49

We learned from those studies that the half-life in the primate vitreous is approximately four days. We also had data that we had collected in tumor models and in a model of retinopathy prematurity that when you give pegaptanib intravenously the amount of pegaptanib that is needed to inhibit the VEGF is about 1 ng/mL.

We also had another inhibitory concentration that we had determined in vitro in tissue culture in various assays of calcium mobilization and endothelial cell proliferation. The relevant concentration in tissue culture of pegaptanib that was required to inhibit VEGF was significantly lower. It was 0.01 mcg/mL or 10 ng/mL.

When we started out it was not entirely clear which of these inhibitory concentrations would be most relevant when you are injecting the drug into the eye. So, we postulated that if this is the most relevant inhibitory concentration, then a 3 mg dose, given every 6 weeks would sufficient block VEGF for the entire 6-week period. If, on

50

the other hand, this was the relevant concentration, the 3 mg dose, the 1 mg dose and the 0.3 mg dose would actually all three be sufficient to block VEGF for the entire 6-week period, and perhaps that may translate to a plateau of the dose response.

[Slide]

To summarize what I have just discussed, VEGF appears to be an important control point for neovascularization and vascular permeability, the pathologies that lead to vision loss in age-related macular degeneration. Pegaptanib specifically targets the VEGF isoform VEGF 165, which we believe is operative in disease. I have shown you data from ROP but this has also been shown to be true in choroidal neovascularization, diabetic retinopathy and other conditions. And, pegaptanib dosing is based on pharmacokinetic data which were collected prior to the conduct of this study.

[Slide]

At this point, Dr. David Guyer will return and David will talk to you about our clinical

51

efficacy data from the pivotal trials.

Pegaptanib Clinical Efficacy

[Slide]

DR. GUYER: In this section we will show you that pegaptanib met a clinically meaningful primary efficacy endpoint with statistical significance in independent, well-controlled, replicate trials, with a favorable safety profile.

[Slide]

The macular degeneration program consisted of 6 trials, 1,281 patients and over 10,000 treatments at 117 sites in 21 countries. The dose ranges that were studied ranged from 0.25 mg to 3 mg per eye.

[Slide]

These are the six trials. EOP1003 and 1004 are pivotal trials, sham-controlled, double-masked, randomized trials. There were 622 patients in the predominantly ex-U.S. trial and 586 in trial 1004 in North America. The other four, smaller trials were pharmacokinetic trials and open-label single or multiple dosing trials with,

52

or without PDT, for the total exposed of 1,281.

[Slide]

The Phase I/II program showed that pegaptanib appeared safe in all tested doses and regimens with no dose-limiting toxicities. There were no unexpected retinal or choroidal abnormalities noted by angiography as read by an independent reading center. As Tony mentioned, these trials established the dosing regimen based on pharmacokinetics.

[Slide]

The study objective of the pivotal trials was to establish a safe and efficacious dose of intravitreous pegaptanib sodium in patients with subfoveal choroidal neovascularization secondary to age-related disease.

[Slide]

The development of these pivotal studies was done in conjunction with our expert advisory panel, whose names are listed on this slide.

[Slide]

The study design was two randomization,

53

double-masked, sham-controlled, dose-ranging trials of pegaptanib 0.3 mg, 1 mg and 3 mg and sham. The treatment regimen was every 6 weeks and the prespecified time point for the primary endpoint was 54 weeks. PDT, photodynamic therapy, was permitted per the FDA-approved label at the masked investigator's discretion. Since shams could have PDT, this represented a usual care control group.

Independent monitoring was done both by an independent reading center that confirmed the eligibility prior to randomization, and an independent data safety monitoring committee, or IDMC.

[Slide]

These were the members of the IDMC. It was chaired by Prof. Alan Bird, who is here with us today.

[Slide]

Because of the biology of neovascular macular degeneration and the mechanism of action of pegaptanib, we designed a trial with a very wide

54

range of inclusion criteria which included a broad range of visual acuities, 20/40 to 20/320, and broad angiographic criteria including all subfoveal angiographic subtypes; lesion sizes up to and including 12 total disc areas in size; greater than or equal to 50 percent of the total lesion size needed to be active choroidal neovascularization; and for minimally classic and occult disease subretinal hemorrhage and/or lipid and/or recent change in vision was necessary for inclusion.

[Slide]

Ocular exclusion criteria included previous subfoveal thermal laser therapy, and to avoid older chronic cases any subfoveal scarring or atrophy or greater than or equal to 25 percent of the lesion being scarred or atrophic. Causes of choroidal neovascularization other than age-related diseases were excluded, and if a patient had recent intraocular surgery or was thought to perhaps need cataract surgery in the near future, they also were excluded. Finally, no more than one prior PDT treatment was allowed.

55

[Slide]

The general exclusion criteria included a history or evidence of severe cardiac disease such as myocardial infarction within the last 6 months, ventricular tachyrhythmia or unstable angina; evidence of peripheral vascular disease; or clinically significant hepatic or renal dysfunction; or a stroke within the last 12 months. Our population, however, was very characteristic of your typical elderly population in that 50 percent of the patients had systemic hypertension; 25 percent were on statins; and 20 percent had cardiovascular disease.

[Slide]

Stratification at randomization included study center, a history of prior PDT use and angiographic subtype.

[Slide]

Our primary efficacy endpoint, which was prespecified, was the percent of patients losing less than 15 letters from baseline to week 54, the same endpoint that was used for marketing approval

56

of Visudyne.

This is an ETDRS chart where 5 letters equal 1 line, and the 15-letter change or 3-line change represents a doubling of the visual angle which is a clinically meaningful change to an individual patient.

[Slide]

Our primary endpoint used in intent-to-treat, or ITT, population included patients receiving at least one treatment and a baseline visual acuity measurement. The last observation carried forward, or LOCF, was used to impute missing data. We will also discuss supportive visual and angiographic endpoints, as well as exploratory or subgroup analyses.

[Slide]

This table shows the various study visits. Of note, a telephone safety check was done 3 days after treatment. Tonometry or measurement of intraocular pressure was done both before treatment and 30 minutes after, and fundus photography and fluorescein angiography was done at baseline and

57

weeks 30 and 54.

[Slide]

In order to preserve the integrity of the masking there were two physicians involved in the trial. One physician administered the study treatment and the second physician was involved in any patient assessments or decisions. Patients were also masked in that the sham procedure was identical to the active drug procedure except for the actual penetration into the vitreous. This meant that they had application of a lid speculum, instillation of topical medications, subconjunctival anesthetic, and pressure against the globe using a needle-less syringe.

The visual acuity examiners were also masked to both he treatment arm and also to previous vision assessments, and the reading center was not aware of the patient's treatment arm.

[Slide]

This slide represents the patient baseline characteristics for both trials 1004 and 1003. What we can see in each trial is that the active

58

doses and the sham are well balanced with respect to sex, age, initial visual acuity, angiographic subtype, prior use of PDT and lesion size. The only difference between the two trials was that there was slightly more prior PDT use in trial 1004. That was the North American trial, and that was because Visudyne was approved and reimbursed earlier in the United States than in Europe. Out of 9 possible injections, on average all patients, treated and sham, received 8.5 of the 9 injections, and overall there was about a 10 percent rate of discontinuation in the trial.

[Slide]

We prespecified to use a Hochberg procedure to account for the multiple doses in this pivotal trial. As per agreement with the FDA, it was decided to unmask study 1004 first--that was the trial that was recruited first, thus, the results were available earlier--in order to determine which doses to formally analyze in the study trial study, 1003.

[Slide]

59

So, we proceeded to unmask the first trial, study 1004, and we found for the 0.3 mg dose 67 percent of patients lost less than 15 letters compared to 52 percent of sham. This hit our Hochberg adjusted p value at 0.0031. Note that the 1 mg dose had a similar response rate, about 66 percent. The p value was 0.0273. The 3 mg response rate was higher than the shams at 61 percent, however, it did not hit the necessary p value.

[Slide]

For this reason, prior to unmasking the second trial, it was prespecified to the FDA that only the 0.3 mg and 1 mg doses would be formally analyzed in the second trial. Then we proceeded to unmask the second trial, study 1003.

[Slide]

This study showed replication of the findings of the first trial study, 1004, in that 73 percent of the patients in the 0.3 mg dose, compared to 59 percent of sham, lost less than 15 letters, again hitting our Hochberg adjusted p

60

value of 0.0105. Again, the response rate in the 1 mg group was similar at 75 percent and a p value of 0.0035, and the response rate in the 3 mg group was 69 percent. The p value you see here, 0.1252 was a nominal p value because we decided, as we mentioned, not to formally analyze it.

[Slide]

So, we can look at the combined data and see that 70 percent of the 0.3 mg group, 71 percent of the 1 mg group and 65 percent of the 3 mg group lost less than 15 letters compared to 55 percent of the shams, and for all of these active treatment groups we had low nominal p values.

It is important to emphasize that for the 0.3 mg group we were able to show independent replication in two trials of a statistically significant effect in a prespecified clinically significant primary endpoint.

[Slide]

I would like to turn now to some supportive visual angiographic analyses. There are a variety of ways of looking at various visual

61

outcomes that are standard for reassurance that the treatment effect for showing the primary endpoint is real. As we will present, all of these analyses were in favor of pegaptanib which gives us confidence in this treatment effect. Because the independent trials had the same protocol and demographics, and because we prespecified it in our statistical plan, we will present these as pooled data.

[Slide]

This graph shows the percent responders over time. What we can see is that we were able to show that the active treatment group had a treatment effect over sham not only at our primary endpoint at 54 weeks, but at every studied time point the active treatment group did better than the sham.

[Slide]

This is a graph of mean change in visual acuity. Again, the active treatment group is here, the sham or usual care group showing a progressive decrease in vision, and the difference at 54 weeks

62

was approximately 50 percent in favor of the active treatment group.

[Slide]

This treatment effect was early and sustained, by as early as 6 weeks, which was the first visit after the first injection the pegaptanib groups had already distinguished themselves from the controls and, as we can see here, the 0.3 mg and the 1 mg group had done that with the low nominal p value. This sustained itself throughout the 54-week course of treatment.

[Slide]

Sham eyes were twice as likely to suffer severe vision loss than actively treated patients, as shown in this graph of percent of patients with severe vision loss. We can see the sham controls with severe vision loss compared to the active-treated groups.

[Slide]

At week 54, again, there was a low nominal p value for the 0.3 mg and 1 mg group compared to sham, with progression to severe vision loss which

63

is 30 letters or 6 lines.

[Slide]

This also was seen for legal blindness in one eye, which is 20/200 or worse. We again can see that more sham eyes progressed to 20/200 vision or worse compared to actively-treated groups.

[Slide]

Patients on pegaptanib were also more likely to maintain and/or gain visual acuity. This graph shows the prespecified endpoints of maintaining or gaining vision that is greater than or equal to zero lines gained, as well as greater than or equal to 3 lines gained. These other two endpoints were not prespecified but we can see again in all cases a treatment effect for maintaining or gaining vision compared to sham.

[Slide]

The next few slides will show the distribution of visual acuity change at baseline and compared to week 54. Let's first look for the 0.3 mg group. This was the range of visual acuities at baseline. Yellow is the 0.3 mg group

64

and blue is the sham.

[Slide]

After 54 weeks in the trial we can see that more patients in the 0.3 mg treated group than sham had good visual acuities and more patients with sham than treated patients had poorer visual acuity. So, the shift in distribution was in favor of our 0.3 mg group, and the p value for this was less than 0.0001.

[Slide]

The same is true, as we can see here, for the 1 mg group. This is the baseline visual acuity distribution and at 54 weeks again we can see more 1 mg treated patients than sham having relatively good visual acuities and more shams than treated eyes having poorer vision. Again, this shift in distribution is in favor of the 1 mg group had a p value of less than 0.0001.

[Slide]

Finally, we can see that for the 3 mg group also. Here is the baseline distribution and at 54 weeks again more 3 mg patients had better

65

visual acuities than shams, and more shams had poorer vision at the end of 54 weeks than the 3 mg treated patients.

[Slide]

This is a graph of the cumulative distribution function of vision. What it shows on the bottom is the change in visual acuity up to week 54 and the cumulative proportion on this axis. This shows the robustness of the data as it uses all of the data points for 54 weeks.

What we can see first is this S-shaped curve. This is the blue sham patients. You can see here, for example, at minus 15--that is minus 15 letters which was our primary endpoint, moderate for vision loss, and we see minus 30 which, as we talked about, represents severe vision loss, and we can see the zero or higher time point which represented maintaining vision. What we can see is that, whether we are talking about preventing vision, maintaining vision or gaining vision, there has been a shift in distribution, a shift in the distribution of the sham patients in all active

66

treatment arms to the right, suggesting benefit in all areas. The area between the lines which represents this improvement was highly statistically significant for all three doses, for the 0.3 mg dose less than 0.0001; the 1 mg dose 0.0001 again; and the 3 mg dose 0.0017.

[Slide]

I would like to now turn to the exploratory or subgroup analyses.

[Slide]

It is important to emphasize that this study was powered to test for statistical significance in the overall study population, that is, to test for the primary hypothesis or primary endpoint of all subjects. Nevertheless, it is important to explore various baseline characteristics such as lesion composition, lesion size, baseline vision, age, sex and pigmentation of the iris.

[Slide]

Despite a reduced ability to draw statistical conclusions because of decreased sample

67

size, in some cases as small as 18 patients, multiple subgroup analyses which can both lead to false positives and negatives--despite this no one subgroup drove the overall effect, as we will show you.

[Slide]

We will first look at the 0.3 mg and 1 mg doses as was described in the FDA briefing book. We have also analyzed and prepared the 3 mg dose and if people are interested later we can show you that. We will present this using pooled data because it was prespecified and we will show the individual trials after.

[Slide]

Here we can see for the pooled data at the 0.3 mg dose that in all cases of all patient characteristics the 0.3 mg active treated group did better than sham. This was for sex, age and, consistent with the biology of this disease and the mechanism of action of pegaptanib, for all angiographic subtypes, predominantly classic, minimally classic and occult, as seen here; also,

68

initial baseline visual acuity, size of the lesion, race and pigmentation of the iris.

[Slide]

Here we can see for severe visual loss--the first graph was moderate visual loss or primary endpoint, but we can see that the conclusions we made are supported by severe visual loss, or 6-line loss, 30-letter loss in this graph. The blue are the sham so all had more severe vision loss than actively treated 0.3 mg group for all patient characteristics. So, this supports our primary analysis.

[Slide]

Turning to the 1 mg group, we can see the same thing, that in all patient characteristics the 1 mg group did better than sham. Again, we can see that this information is supported by severe vision loss where, again, sham in all cases did worse than the actively treated 1 mg dose.

[Slide]

Let's now turn to the individual trials. Individual trials which are under-powered

69

inherently have more variability. Nevertheless, we can make the same conclusion, that no one subgroup drove the overall efficacy. Again, for trial 1004 with the 0.3 mg group we can see the very small Ns, sample sizes, for some of these groups and, again, we can see support for using severe visual loss as another important clinical endpoint.

[Slide]

For trial 1003, with the 0.3 mg dose we can see the same thing.

[Slide]

For the 1 mg dose, again we can see, in trial 1004, that in all cases the treated groups did better than the controls and this was supported by the severe vision loss in 1004 again.

[Slide]

And, in trial 1003, again, for moderate vision loss treated patients did better than the blue shams and support with severe vision loss where shams did worse than actively treated patients for progression to severe vision loss.

[Slide]

70

In order to be sure there were no important subgroup relationships, we also performed a multiple logistic regression to identify any potential factors either influencing the outcome or modifying the treatment effect. Subgroups and interactions of subgroups with treatment were investigated.

[Slide]

These are some of the subgroups that we evaluated, age, angiographic subtype, use of PDT, sex, race, lesion size, status of smoker/non-smoker, subretinal hemorrhage, the fellow eye vision loss and lipid.

[Slide]

We found for the 0.3 mg dose that no factors were identified as significant treatment effect modifiers for 0.3 versus sham, and no factors except treatment with pegaptanib were identified as significantly influencing the response, and this had a p value of 0.0003 in favor of treatment.

[Slide]

71

For the 1 mg group we again found that no factors were identified as significant treatment effect modifiers versus sham, and for pegaptanib at 1 mg there was a relationship between treatment with pegaptanib, again at 0.0001, and age which favored patients with less than 75 years of age. This is not to say that older patients did not do better. It just said that there was a favor for younger patients even both appear to respond.

[Slide]

What can we conclude from these exploratory or subgroup analyses? First, we have shown that the treatment benefit appears well-distributed among a broad patient population. Second, the efficacy is not consistently concentrated in or absent from any particular patient subgroup. No one subgroup drove the overall efficacy.

[Slide]

The 0.3 mg dose represents the lowest studied efficacious dose and it met its primary efficacy endpoint with statistical significance in

72

independent replicate trials, as we have shown you. The efficacy was substantiated in every clinically meaningful endpoint tested. We have seen the secondary endpoints. And, the 1 mg and 3 mg doses show no additional benefits over 0.3 mg. Tony will shortly show you that there was no safety difference between 0.3 mg and 1 mg as well. However, theoretically we all know that a lowest dose yields the lowest systemic concentration. So, the sponsor advisory board and independent data monitoring committee endorsed the 0.3 mg dose as a dose that should be selected.

[Slide]

I would like to turn now to angiographic findings. We have mentioned to you that we believe there are two mechanisms of action for pegaptanib, anti-angiogenesis and anti-permeability. As I will now show you, we have anatomical confirmation for both mechanisms of action that support the visual findings we have shown you today.

Let's first look at the anti-angiogenesis. Here is a patient in the trial with predominantly

73

classic neovascularization that showed virtually complete regression. The white large area is the neovascularization. You can see it has almost completely regressed after 54 weeks of treatment. But this is one case. So, let's look at the whole group.

[Slide]

What we can see is that there was a decrease in the lesion size that had a low nominal p value in favor of active treatment for the 0.3 and the 1 mg dose. So, we have anatomical confirmation or support for anti-angiogenesis as a mechanism of action that supports the visual findings.

[Slide]

The second mechanism of action that we described was anti-permeability. Here is another patient in the trial that had significant cystoid macular edema with neovascular disease. We can see the cystoid-like patterns here. This is a sign of a lot of permeability. After 54 weeks of treatment we can see a great decrease in the permeability.

74

[Slide]

Again, we can show that leak size over time was less for treated groups than for shams. The p values here are noted.

[Slide]

In addition, we can look at the change in leakage to week 54 as a sign of anti-permeability action, and we can see that very similar to visual distribution curves I showed you earlier, we can see again that there was less leakage noted more often in actively treated 0.3 mg patients than in sham, and more leakage noted in shams than in actively treated eyes. This change in distribution had a low nominal p value of 0.0004. So, again we have anatomical confirmation for anti-permeability as an important mechanism of action that supports the visual findings.

[Slide]

I would like to now turn to photodynamic therapy, or PDT. I think it is first important to have a historical perspective of the use of PDT in this trial so you can understand some of the

75

challenges we faced when we were designing this trial.

At the time of starting the trial PDT was available primarily in the U.S., and there were certainly ethical considerations that required that PDT be permitted in patients with predominantly classic disease. However, the PDT usage pattern was not yet known.

[Slide]

So, what we decided to do was to create very strict rules for the use of PDT in this trial. What that meant was that patients had to have predominantly classic disease and the masked physician--remember, we had two physicians--the masked physician determined if the patient was eligible for PDT per the FDA label and then whether that PDT was recommended for that individual patient. If so, the treatment was administered per the FDA label.

Now, to ensure that these strict rules were being followed, we had a reading center review the usage pattern and we found that 92 percent of

76

the time the reading center agreed with the appropriate use of PDT in this trial.

[Slide]

PDT use could occur three ways: prior to the study, at baseline, and post-baseline and, actually, any combination of the three. It is important to emphasize that overall the use of PDT was extremely low. Three-quarters of patients were never exposed to PDT in the study eye at any time in the time trial.

[Slide]

Let's examine each one of these three scenarios in detail. First let's talk about prior PDT which was stratified and was balanced at randomization. Also, notice the small numbers again, emphasizing very little PDT use in the trial, 18-29 eyes in the various subgroups, but it was stratified and balanced.

[Slide]

Baseline PDT is the second scenario, and the baseline PDT use was again very similar among the groups. We can see here that for the active

77

treated groups 10-13 percent of patients had PDT at baseline compared to 14 percent for shams and, again, look at the very small numbers, 31 to 40 patients per subgroup.

[Slide]

Finally, let's talk about post-baseline PDT use. Now, it is important to mention that a meaningful analysis of potential post-baseline PDT effects on efficacy is limited to the inherent bias in the trial. What I mean by that is, remember, the patients were never randomized to post-baseline PDT use. In order to really assess the baseline PDT use we would have had to design a trial randomizing patients to PDT and baseline. That wasn't this trial. As an example of this, what is called the channeling bias, a patient with a poor response might be the patient that would be preferentially channeled to get PDT. What this really means is that post-baseline PDT is an outcome variable. So, for this reason, we must treat post-baseline PDT in a different way, as I will show you now.

78

[Slide]

We need to ask was there increased PDT use in pegaptanib patients relative to sham that could suggest that some of the pegaptanib efficacy was derived from PDT?

[Slide]

The answer to this question was no. As we can see, there was no higher use of post-baseline PDT in active treated patients compared to sham.

[Slide]

The second important question about post-baseline PDT use is was there an increase in the average number of PDT treatments in pegaptanib patients relative to sham?

[Slide]

Again the answer is no. As we can see again, there was no higher post-baseline PDT use in active treatment eyes compared to sham.

[Slide]

The third important question, which will be addressed in detail in Tony's safety section, is was there evidence of any adverse events with the

79

co-administration of photodynamic therapy and pegaptanib that could lead to a drug-to-drug interaction? The answer is no--more on that in just a few minutes.

[Slide]

In summary, pegaptanib met a clinically meaningful primary efficacy endpoint with statistical significance in replicate, independent, well-controlled clinical trials.

[Slide]

I will now ask Tony to come up and discuss our clinical safety database.

[Slide]

DR. ADAMIS: This is the entire safety database. This includes the patients from the earlier Phase I/II trials. What you see here is that the total clinical experience to date includes over 1,200 patients in over 10,000 treatments, of which 7,500 are intravitreous injections that we can monitor for the safety. There is a slight imbalance that you will see in that there are more

80

patients receiving 3 mg than 1 mg of 0.3 mg. That is because that was the dose that was used throughout most of the Phase I/II program. In addition, we gave doses of 0.25 mg and 2 mg in those earlier programs as well.

[Slide]

The overall safety is shown here. As regards any adverse events, you can see it is balanced between all treatment arms and sham. There is an imbalance in the serious adverse events. These are largely injection related, and we will talk about those in depth in a moment.

The discontinuations, you will note, due to adverse events are low. They are one percent in both the treated and the sham arms. Similarly, the death rate is balanced to two percent.

[Slide]

Looking at the death rate just a little more closely, we can see that there is no evidence here of a dose response.

[Slide]

Let's look at the most frequent non-ocular

81

serious adverse events. This is a busy slide but the thing to note here is, first, that there is good balance between the treated and the sham arms and, secondly, there is no clustering within a system organ class. This is rather diffuse. These conditions are age appropriate. The mean age of this population is 77 years old that we studied. These people had a number of concomitant illnesses. Fifty percent of them had hypertension; 25 percent were on statins; 20 percent had cardiac disease. So, we believe it is representative of the population.

[Slide]

We looked particularly for VEGF inhibition-related adverse events as these have been reported with other non-selective inhibitors given intravenously at higher doses. We were happy to see that there were no signals here. The most sensitive signal, the one that has been picked up with other non-selective inhibitors in smaller trials than ours, less powered but nevertheless it was evident, was hypertension. You can see here

82

that the rate of adverse events is 10 percent both in the treated and in the sham arms--no signal there for that very sensitive signal of VEGF inhibition. Thromboembolic adverse events are similarly balanced, as are ischemic coronary artery disorders, heart failure and serious hemorrhagic adverse events.

[Slide]

Why is it that we did not see any of these VEGF inhibition-related phenomena? There is a number of reasons. Some of these are theoretical, some are real but in aggregate they provide I think an argument. Pegaptanib is, as I said, selective for VEGF 165 so the other major isoform 121 is never blocked. So, all VEGF is never blocked with pegaptanib, even if you gave it at very large concentrations. It just does not bind to VEGF 121.

Secondly, the concentrations that we see when we put 0.3 mg in the eye are many orders of magnitude less in the plasma and those concentrations are below the inhibitory concentration that our models have told us both for

83

in vitro and in vivo inhibition of VEGF. So, we believe that these are levels that are below the ability of pegaptanib to affect VEGF levels in any sort of substantive way.

Third, as I just said, there was an absence of sensitive VEGF inhibition signals, the most sensitive being hypertension which I showed you but also in our 1006 trial, where we looked carefully at proteinuria, again there is no evidence that this drug is inducing proteinuria in either our clinical population or in our preclinical models.

Then, the report recently of thromboembolic adverse events occurring in cancer patients on chemotherapy and receiving Avastin--we think there are a couple of very different things about our population and that population that was studied. Number one, cancer in and of itself predisposes patients to thromboembolic phenomena. They have indwelling catheters; they are bedridden; and the cancer itself alters the clotting system. Secondly, some chemotherapy has been shown to be

84

vascular toxic, to be prothrombotic. There is a published literature on that.

So, add these two hits to the vasculature and then block all VEGF to prevent the endothelium from healing itself, one can have a theoretical basis for understanding now why thromboembolic phenomena may be more prevalent in a population with cancer and chemotherapy. That is not age related macular degeneration. This is a very different population that is not, by and large, on chemotherapy and do not have cancers.

[Slide]

Let's look at the ocular adverse events. Again, this is a busy slide but we will talk about these events in a little more detail. They are listed here, those that occurred greater than or equal to 10 percent of patients on either pegaptanib or sham. You can see that there is a slight imbalance in eye disorders, and we will talk about these, and you see a number of various adverse events listed here.

[Slide]

85

Let's talk about them in more detail, number one that was listed on the previous slide being eye pain. These patients receive nine intravitreous injections over the course of a year. It is rather remarkable actually that two-thirds of them never reported a single instance of pain. Of those patients, approximately the one third that did report pain, it was mild or moderate in character in 99 percent of them, and only one patient exited this trial describing an adverse event of pain.

The other important thing to note here is that the eye pain in the sham arm, at 28 percent, was significantly higher than what is seen in the fellow eye, 2 percent. So, some of this mild pain that these patients experienced--one conclusion you can draw is that it may be due to the preparative procedure prior to the injection of the drug. As you recall, these patients have a speculum placed in the eye. They have povidone-iodine scrub. They have a subconjunctival anesthetic injection. These things may have contributed to the lion share of

86

the reports of pain which, again, was mild. Then, obviously, there is a difference here. The remainder of it here can well be ascribed to the actual intravitreous injection itself.

Of those patients who reported pain, it was in a minority of their injections, two in both the treated and the sham arms, and the median time to resolution was two to three days which is the time of the follow-up phone call.

[Slide]

With regard to vitreous floaters, there was more than an imbalance here. It was 33 percent in the treated arms versus 8 percent in the sham. Again, there is a slight difference, 8 versus 1, between the sham eye and the fellow eye so some of this may be due to the preparative procedure but a large portion of it, the majority of it, is very likely due to the act of giving an intravitreous injection itself. When giving a 90 mcL volume injection into the eye, in the average human a volume of 4 mL, you are displacing the vitreous and it is perhaps not surprising that as a function o

87

that you are going to induce floater. These floaters never were severe. All of them were characterized as mild to moderate. No patient left the trial because of floaters. It was in a minority of injections, 1 to 2 injections, that these were reported, if they ever were reported, and the median time to resolution was 3 days in the treated arms versus 7 days in the sham arms.

[Slide]

We looked at cataract very carefully. We specifically looked at cataract in only the aphakic eyes. One-third of these patients approximately were pseudophakic. What we saw was that across all treatment arms there was a slight imbalance, with 30 percent of the eyes having an adverse event of cataract versus 26 in the sham arm. This slight imbalance may be partially explained by the fact that the phakic fellow eye also had a slight imbalance, 17 percent in the treated versus 15 percent in the sham arms.

But we looked at this a little more in depth. The type of cataract that one would expect

88

if this was due to a drug toxicity, the type that has been amply described in the literature, is posterior subcapsular cataract. So, when we looked for that specific type of cataract grading, you can see there is zero difference. It is 11 percent in both the treated and the sham arms.

[Slide]

Nuclear cataract was similarly well balanced. In fact, if you remove the eyes that were vitrectomized, which we will talk about in a minute, vitrectomy can cause a nuclear sclerotic cataract to accelerate. This is 18 percent in both arms and there is, indeed, a slight imbalance in cortical of 18 versus 15 percent.

One piece of objective data we have is that the vast majority of these patients came in at baseline with cataract and only 3 patients underwent elective cataract surgery over the 54 weeks of the trial in the treatment arms.

[Slide]

Anterior chamber inflammation was another adverse event. You can see here that there is an

89

imbalance slightly with 14 percent occurring in study eyes versus 6 percent in the sham eyes, and there were zero reports in the fellow eyes. None of these cases of anterior chamber inflammation were characterized as severe. All of them were mild to moderate and we believe they were largely due to the active intravitreous injection and not to the drug itself. The reports of inflammation were all moderate and self-limiting and did not increase during the course of the trial. In fact, there was a slight trend to decrease, arguing that there wasn't a sensitization to the molecule here, in fact, supporting that this was due to the injection itself. The median time to resolution was 8-9 days, and no patient left the trial because of inflammation.

[Slide]

We looked at interaction potentially with PDT and specifically at ocular adverse events. You can see here that the majority of patients did not have the combination of PDT and pegaptanib, but of those who did we looked very carefully at the event

90

rates and the important thing to consider here is the event rate difference in the sham arms plus/minus PDT, and does that difference change in any sort of meaningful fashion when the PDT is given together with pegaptanib. The answer is that from these data there doesn't appear to be a difference in those two measures. The same is true with vitreous floaters. There is a slight difference here and there is really no difference here in the treatment arms.

[Slide]

But let's look at it another way. This assessment is looking to see if there was a report of an adverse event at any time during the 54 weeks. For instance, if the patient had PDT at baseline but had an adverse event at 54 weeks it would be captured and presented in these data. We thought we would try to look at this a little more carefully and see if there was a better temporal relationship. So, now we are looking at data of patients who had PDT plus/minus 2 weeks around an injection of pegaptanib. These events may more

91

likely signify some sort of interaction and, again, there are no alarming signals here.

When one looks at eye pain there is very little difference here and there is very little difference here between the sham and the treatment arms. The same is true for corneal epithelium disorders. For these two specific adverse events one can postulate a mechanism as to why that is. There is, you know, the povidone-iodine prep for the injection which can affect the epithelium and perhaps cause pain. On top of that is a near temporal relationship the placement of a contact lens for doing the PDT, and one could understand why there might be a slight increase here. Again, no patients dropped out because of any adverse events related to a combination of PDT and the use of pegaptanib.

[Slide]

Now let's concentrate a bit on ocular serious adverse events. The three most common we are going to discuss in detail here are endophthalmitis, retinal detachment and traumatic

92

cataract. The ones below occurred at a very low event rate. When the narratives in the cases were looked at in depth there really did not appear to be an association with the use of pegaptanib so we will not discuss them further here unless you wish to discuss it later in the question and answer session.

Endophthalmitis occurred in 12 patients over 54 weeks. That translates to a relative risk of 1.3 percent of patients developing endophthalmitis over the course of one year of therapy. So that we could compare our rate to the published literature this was converted to a per injection rate of 0.16 percent. What we learned is that the rate that we saw is not an outlier; it is within the published norm and reported norm in cases of endophthalmitis in patients receiving intravitreous injected therapeutics.

As important as the rate is what happened to these patients, what was the outcome. One patient lost severe vision in this trial as a function of their endophthalmitis, 1/12, which

93

translated to a rate of 0.1 percent over the course of the year. Seventy-five percent of the patients who developed endophthalmitis elected to stay in the trial.

Traumatic cataract--you can see there were five cases of it and there were five cases of retinal detachment, of which three were rhegmatogenous in nature.

[Slide]

I show you here the specific details of all 12 cases of endophthalmitis. What you can see here are the starting visions, the visions prior to the event, and the change in vision from just prior to the event which probably most accurately captures the visual loss related to the endophthalmitis itself. What you can see is the one patient who lost 11 lines as severe vision loss.

Let me just tell you anecdotally what happened. It was a protocol violation. It turns out this patient had an active lachrymal sac infection prior to the development of the

94

endophthalmitis and the injection of the mediation, and had an active lachrymal sac infection after the event of endophthalmitis. The patient should never have been enrolled because that was an exclusion criterion.

The other patients, as you can see, were treated aggressively and their visual outcomes tend to be perhaps a bit better than what you would expect for a case of endophthalmitis. In fact, there are some patients here who gained one or two lines of vision.

[Slide]

How were these patients diagnosed, and were we able to identify the endophthalmitis relatively early? This slide shows you exactly what happened. Three patients were identified in their follow-up phone call at days three-four post injection. Eleven patients presented to their physician's office with complaints, and this happened between days two and five. Two patients came in and were diagnosed in a routine follow-up. The endophthalmitis cases I am describing here are

95

the 12 in the first year and the ones that have occurred subsequent to that which I am going to talk about.

[Slide]

We have been following the endophthalmitis issue very carefully and I would like to provide you with an update on where we are beyond the 54-week time period. As I just said, in the first year 0.16 percent of injections, or 1.3 percent of patients, developed endophthalmitis. In the second, and now some patients have entered the third year of this trial, there have been five additional cases as of July 31st of this year, and there has been one case in our Phase II diabetic macular edema trial. So, if you look at the total now, it is 18 cases of endophthalmitis with a denominator of over 14,500 injections, and the rate now is reduced somewhat to 0.12 percent per injection.

In the first half of this trial when we saw the case reports of endophthalmitis we convened an expert panel of ophthalmologists and retinal

96

specialists who work in the endophthalmitis area and we decided that we needed to heighten the awareness of the need for strict adherence to an aseptic protocol when one is giving an intravitreous injection. In fact, there was a letter sent to IRBs and a formal protocol modification mandating the use of a sterile drape, of a speculum, of the use of povidone-iodine. When we did these things and we analyzed what the potential effect could be, what we saw was that prior to that protocol modification being adopted at all sites between August of 2001 and May of 2003 the rate was 0.18 percent, and after that protocol modification the rate has now fallen to 0.03 percent.

Can we ascribe the decrease in the rate to the change in the protocol? Not necessarily. There was more than one variable that was changing here. At the same time that we instituted this protocol modification and heightened awareness about the aseptic technique there was a dramatic uptake in the number of intravitreous injections

97

being given for off-label use in diabetic macular edema with steroids, triamcinolone in particular. So, the knowledge base and the experience of retina physicians increased rather dramatically at the same time that we saw a drop in our rates.

[Slide]

The visual outcome for the cataract cases is shown to you here. For the one patient who lost 7 lines of vision, it was ascribed to progression of macular degeneration. All of these patients, in fact, had successful cataract surgery.

[Slide]

The visual outcome of the retinal detachment cases is shown here. All of these were successfully repaired and you can see the cases of rhegmatogenous detachment which most likely were injection related. The visual outcomes were quite good.

[Slide]

Intraocular pressure was examined. As I said earlier, it is not surprising if one injects 90 mcL into a 4 mL closed space that you will see a

98

transient rise in pressure. In fact, in ophthalmology it is common with almost all procedures that pressure spikes tend to occur. Well, they occurred here and the transient rise in mean intraocular pressure at the first prespecified measurement, 30 minutes, was 2-4 mm across the treatment arms.

It is important to note that the mean intraocular pressure returned to pre-injection levels one week following the injection, which was the next visit, and that 90 percent of patients, approximately 90 percent of patients, never had a spike above the prespecified threshold of 35 mm and any patient who did have a spike was not allowed to leave the physician's office till the pressure was below 30 mm.

Very importantly, there was no evidence of a persistent increase in intraocular pressure over one year. The drug did not seem to alter the outflow of the eye in any way. In those patients who did have a spike, the question was if you had a spike was it because somehow the drug was altering

99

the outflow mechanisms, and if that was the case you would expect to see an increased incidence during the course of the trial as it progressed. As the data show you here, that is not the case. It doesn't appear to increase over time and, in fact, may have been dropping slightly.

[Slide]

This slide simply shows the mean intraocular pressure values over time for all three treatment arms and sham, again giving us some confidence that the drug is not inducing a rise in chronic IOP.

[Slide]

We have a safety update for you regarding angiography. Colored photographs and angiograms were looked at in the independent reading center at the University of Wisconsin. We have looked at up to 97 percent now of our month 18 angiograms and 92 percent of our two-year angiograms to get a sense of is there any evidence of cumulative toxicity. The results are that there is no evidence whatsoever of alterations in the normal retinal or

100

choroidal vasculature as a function of the drug being in the eye now for up to two years, nothing that deviated from the natural history of age-related macular degeneration and no alterations in the normal vessels.

[Slide]

The safety update, which was just concluded in the past week by the independent data monitoring committee, has reviewed 100 percent of the patients through month 18 of this trial and 97 percent through month 24, and there have been no deviations from sort of the pattern of adverse events, the ones that we saw in the first year of the trial. There have been no new safety concerns except perhaps for a slight increase in the number of retinal detachments. There were 6 that were reported in the second year of this trial.

[Slide]

To summarize the non-ocular safety, there was a very low discontinuation rate due to adverse events. It was one percent and it was balanced in the treated and the sham arms. Non-ocular serious

101

adverse events appeared to be similar in rate and character between pegaptanib and sham, and the mortality rate, as you saw, for the 77 year-old population was similar between pegaptanib and sham.

[Slide]

As regards ocular safety, I think what we can conclude is that the majority of the ocular adverse events were judged to be procedure related. They were transient and mild in character and largely self-limiting. There was a low discontinuation rate due to ocular adverse events and the serious adverse events were infrequent. They were rarely associated with severe vision loss and were mostly procedure related. Finally, there were mild transient and predictable, manageable increases in intraocular pressure but no evidence of a long-term rise in intraocular pressure.

[Slide]

At this point Prof. Don D'Amico, who is a practicing retinal specialist at the Massachusetts Eye and Ear Infirmary, will come and discuss the risk/benefit profile for pegaptanib.

102

DR. D'AMICO: Thank you, Dr. Adamis. Dr. Dunbar, members of the advisory committee, ladies and gentlemen, with your permission I would like to introduce myself a little more fully and my perspectives so that you can have the clearest context in which to place my remarks.

[Slide]

With regard to this study, while it was in progress I was invited to be a member of the safety committee and later became its chair. At the conclusion of the study I was asked to be a member of the scientific advisory board. I perform a virtually identical role for the Alcon Corporation, chairing their safety committee in the evaluation of their anecortave product. I also advise them on surgical themes and instrumentation as well. Finally, I am a consultant to the Iridex Corporation serving as a member on the safety committee for the transpupillary thermotherapy trials and their PTAMD or laser for drusen trial. I hold no equity in any of these companies nor any

103

of their competitors.

[Slide]

I would like to also share four perspectives that will inevitably influence my remarks and may be helpful to you also in your evaluations. First, of course, I was a member of the pegaptanib safety committee. Secondly, I have had a career-long laboratory, as well as clinical, interest in endophthalmitis and the effects of administration of intravitreal medications. I am, as introduced, an academic in the field of retinal diseases and therapy. But perhaps most importantly and most germane is that I have a very active retinal practice at the Massachusetts Eye and Ear Infirmary and care for many patients with macular degeneration.

[Slide]

As has been said, neovascular AMD is quite a source of human suffering. At the 20/40 level of visual acuity driving privileges frequently become impossible for a patient. At 20/80 or worse difficulty is even present in trying to read large

104

print. And, 20/200 or worse is a commonly accepted level of definition of legal blindness at which it is difficult to recognize faces and independent function is threatened.

[Slide]

How extant is this problem in the world today? In a very careful meta-analysis of the most comprehensive studies recently reported by the Eye Diseases Prevalence Research Group, they looked at studies in the United States, Western Europe and Australia over an 11-year period.

[Slide]

Based on their analysis, it is the leading causes of blindness in U.S. adults in patients aged 40 years or older. You see that slightly over half are due to age-related macular degeneration.

[Slide]

They then applied their model to the U.S. Census data for both 2000 and projected to the future. In a morning filled with numbers, I will spare you all the numbers here, but using a definition of 20/200 or worse as blind and 20/40 or

105

worse as visually impaired, there are 3.3 million Americans with visual impairment today. In the future there will be approximately 5.5 million American with visual impairment at some level, again slightly over half, due to age-related macular degeneration. So, it is clearly a problem.

[Slide]

As such, it merits our highest attention as physicians, researchers, etc. to try to find treatments and even cures. This slide is color coded and it lists the candidate therapies for neovascular subfoveal age-related macular degeneration. Therapies which have demonstrated effectiveness in replicate clinical trials are shown in yellow. We have laser photocoagulation, photodynamic therapy with Visudyne and the data you have just heard on pegaptanib. The great majority of interventions are listed in white, which indicates ongoing study with various degrees of promise, and it includes surgical options, as you see here and a variety of other laser treatments, as well as other pharmaceuticals, many of which are

106

nearing the end of their clinical trials. There are also some abandoned therapies that were ineffective and combination strategies, as you see in the lower right, are becoming of increasing interest.

[Slide]

Looking at the established therapies, there are two. One is photocoagulation with thermal laser which has been effective in extrafoveal, juxtafoveal and subfoveal lesions. However, in subfoveal lesions this therapy has been abandoned due to the immediate destruction of central vision following treatment and is no longer in clinical use. Photodynamic therapy with Visudyne is approved for subfoveal predominantly classic lesions.

[Slide]

In addition, evolving clinical practice, in a hope to provide improved care for patients with macular degeneration, has led to a new accommodation therapy which has become widespread. That is the combination of a PDT treatment with

107

Visudyne in association with an intravitreous induction of triamcinolone in the peri-PDT period. This treatment has had some very promising early pilot results but the literature is quite minimal at present. Nevertheless, it has become a common treatment in clinical practice.

[Slide]

Intravitreous injections are quite common in my world as a retinal specialist. They were employed and were actually the pathway to great success in the therapy of endophthalmitis, and are still continued widely in use for that indication. We also utilize intravitreal injection as a treatment of retinal detachment, as well as administering agents for CMV retinitis. However, there has been great expanded use recently in office practice of intravitreal injections as regards the use of triamcinolone acetodine for diabetic macular edema, retinal vein occlusions, uveitis, as I have just mentioned, in association with photodynamic therapy.

[Slide]

108

Pegaptanib represents the potential for a new approach, a pharmacotherapy, and what are the advantages of pharmacotherapy? They are both general and specific. In general, pharmacotherapy offers the prospect of treatment at a molecular level with improved targeting of the disease process and, more importantly, limitation of the collateral damage that invariably occurs with larger scale interventions such as surgery or laser.

Pegaptanib quite specifically is based on very extensive basic science into the most widely accepted, central disease processes in AMD, namely neovascularization and leakage, with consistency across multiple experimental models and studies.

[Slide]

As a member of the safety committee, we looked for three specific areas in great detail. One, were there any potential systemic side effects from receiving an anti-VEGF medication? Secondly, were there intraocular drug-related side effects from this VEGF medication? Thirdly, were there any

109

mechanical side effects or complications from the intravitreal injection procedure itself?

[Slide]

We did find serious ocular adverse events related to the injection procedure. As you have heard, there were 12 cases of endophthalmitis. This incidence rate is quite comparable to that in published series for intravitreal injection with the other forms of intravitreal injection therapy that I have mentioned previously. One of these patients had severe visual loss. Nine of the patients continued in the study and elected to continue receiving study medication. Finally, after protocol modifications, the incidence is clearly trending downward.

There were five cases of retinal detachment, which were repaired and some were related to the underlying macular degeneration itself. Traumatic cataract was seen in five cases and all were surgically repaired without sequelae.

[Slide]

So, in these 22 serious ocular events, we

110

considered them in the context of 7,545 intravitreous injections performed in 1,190 patients by 117 centers worldwide, and many of those centers had more than one injector. We felt that this denominator indicated substantial safety for this procedure.

We also found no evidence of systemic side effects, no evidence of ocular drug-related side effects, and the majority of other adverse events were mild and transient within the eye. The serious ocular adverse events were infrequent and manageable. So, we concluded that there was a very favorable safety profile that, in addition, may be further improved by education and additional training.

[Slide]

If we look at severe vision loss, again to understand the context of these adverse events, if a patient presented to the trial and received sham, that is, usual care, there was a 22 percent risk per year of suffering severe visual loss. When they were enrolled in the pegaptanib group that

111

risk was reduced to 9.5 percent per year.

[Slide]

In the endophthalmitis, retinal detachment and cataract serious ocular events that we saw, the risk of severe vision loss, that is 6 or more lines of vision, was 0.1 percent, indicating substantial order of magnitude less risk from endophthalmitis than from the real problem here which is the macular degeneration itself.

[Slide]

Regarding efficacy, you have heard a detailed presentation and I will just summarize. There was significant reduction in moderate and severe vision loss compared with sham. There was promotion of vision stability and gain in a proportion of patients. There was efficacy with broad-based entry criteria including a range of subfoveal neovascular AMD lesions. And, the benefit of intravitreous pegaptanib therapy was early and sustained.

[Slide]

As we have seen, in this slide baseline

112

visual acuity is on the left. Sham is indicated in purple and pegaptanib in grey. At 54 weeks there is a definite shift in the 0.3 pegaptanib group to preservation of better vision on the left of this chart compared to the visual acuities observed with sham.

[Slide]

I am not a biostatistician but I will try, for myself and for all of us, to place these results in some kind of a wider context. What could this mean? No one knows exactly how many new subfoveal neovascular lesions occur a year, but 120,000 per year of new treatment-eligible patients is probably a reasonable estimate. If those patients were to behave similar to the gathered group enrolled in this trial, we could make some statements, and here they are:

Pegaptanib potentially prevents severe vision loss, that is a loss of 6 or more lines of vision, in 15,000 additional patients per year in the United States compared with usual care, based on a 57 percent reduction in the rate of severe

113

vision loss with pegaptanib.

[Slide]

Secondly, reaching a level of 20/200 or worse within the treated eye, we could call that blindness in the treated eye. Pegaptanib potentially prevents treated-eye blindness in an additional 22,800 patients per year in the U.S., again compared with usual care, based on a 38 percent reduction in the rate of treated-eye blindness with pegaptanib.

[Slide]

In conclusion, from the perspectives available to me and now available to you, I have concluded that pegaptanib will have a significant impact on AMD in regard to both individual patients with AMD lesions that would become amenable to treatment and, secondly, in its effects on visual function and its preservation in the aging U.S. population.

The positive results in this trial indicate the beginning, and not the limit, of pharmacotherapy for AMD. I agree with the

114

sponsor's recommendations that the benefits of pegaptanib therapy for AMD strongly outweigh the risks. Thank you.

DR. DUNBAR: Thank you to the sponsor and Drs. Guyer, Adamis and D'Amico. At this point I would like to open the floor for discussion and questions for the sponsor from the committee members, and ask that you will speak your name into the microphone as you ask each question. Are there any questions from the committee members? Dr. Chinchilli?

DR. CHINCHILLI: Yes, I don't know much about the disease and the patients that were recruited for the two trials so, please, bear with me. Could patients have AMD in both eyes? I mean, roughly what proportion of patients that were in the trials had that situation?

DR. GUYER: In general, for neovascular age-related macular degeneration usually one eye becomes active at a time. If the patient lives long enough, they often will get it in the second

115

eye. In this particular trial the investigator would choose--in a very few number of cases where there would be active disease that was eligible for the trial, the doctor would make that decision. If we look at slide D-82--

[Slide]

--here we can see some of the baseline characteristics. In two-thirds of the patients this was the worse eye that was treated. Again, no patient was treated in both eyes at the same time. But in the lifetime of a patient there could be some overlapping times where they have an active lesion and the second one becomes active. Some patients are fortunate enough not to get it in their second eye but, unfortunately, if they live long enough many will.

DR. CHINCHILLI: Thank you.

DR. DUNBAR: Dr. Pulido?

DR. PULIDO: A superb presentation; very interesting results. Just two questions. Number one, glaucoma was not an exclusion criterion in the study. So, some of the patients had glaucoma and

116

AMD. Do you have any data as to the effect of chronic injections on the small subgroup of patients that had glaucoma?

DR. GUYER: I will let Tony answer that.

DR. ADAMIS: We were interested in that question as well. Slide OS-31.

[Slide]

We looked specifically at patients with a history of ocular hypertension and/or glaucoma, and then followed their pressures throughout the entire 54 weeks. What we saw was that there was no change in their intraocular pressure as a function of treatment.

DR. PULIDO: The other question probably is to you as well. There are some recent articles--here is one from Nature, May: VEGF delivery with retrograde transported Lentivector prolongs survival in a mouse ALS model. Here is another one: mural protection of ischemic brain by VEGF is critically dependent on proper dosage. Here is another one. So, we have gone under the assumption that VEGF and VEGF 165 is specifically a

117

cytokine for angiogenesis, but there is more data to show that there is an independent effect directly on neural tissue, separate from its angiogenic effect. ERG was not a part of this trial. You did some ERGs on some dogs, from what I saw here. I don't know how many, how long, etc.

So, considering the neuroprotective effect, from your data--it is wonderful--that the angiogenesis is important, critical to take care of this significant problem in patients. But my concern is the long-term chronic dosaging considering that there is an independent effect of VEGF as a neuroprotective agent.

DR. ADAMIS: As always happens in science, what seems very straightforward becomes more complex, and what you quote is absolutely correct. I think that is Peter Carmeliet's paper in Nature. But what has been learned in about the last five years is that neural cells have VEGF receptors and VEGF may be neuroprotective for certain tissues. Certainly, in the ALS model that is the most convincing story to date. Whether the effect is

118

direct or not is still being debated in the scientific world, but it may well be direct because of the VEGF receptor on the neural cells.

We were interested in this as well. Even before we got into the scientific question as part of our preclinical safety testing, there was a 9-month dog study where the dogs received 3 mg injections every 2 weeks bilaterally. Then they had ERGs done and there were no abnormalities seen there. So, that gives us a little bit of comfort but, more importantly, recently we examined this issue and looked specifically at the isoform story. We presented a paper at ARVO last spring where we showed that in a model of retinal ischemia if one gives a pan-isoform, non-selective VEGF inhibitor, you can in fact induce some neural apoptosis. But when we gave pegaptanib in the exact same setting there was no induced apoptosis. So, again getting at this thesis, the important thing with pegaptanib I think is that you are sparing some VEGF to allow it to have its physiological or perhaps these rescue functions that can occur in the eye. So,

119

that gave us an additional measure of comfort that we are not going to have neural toxicity.

DR. PULIDO: But the question still arises have you done long-term ERG studies on these patients?

DR. ADAMIS: Oh, I am sorry, no, we have not done those in these patients.

DR. DUNBAR: Mr. Kresel?

MR. KRESEL: My disclaimer is that I am not a statistician and so I am not sure if this is even an appropriate way to ask this but I am going to ask it anyway. You did a great job of looking at endophthalmitis which, you know, obviously is one of the things that people have concern about, and referred to a decrease in patients that was only five cases in years two and three. My question is how many patients continued therapy that far? So, did the number of patients decrease and, therefore, the percent not go down? Because what we saw is a cumulative number that, of course, did go down.

DR. ADAMIS: It is a fair question. The

120

number of patients was decreased in the second year. That is why the metric we used was on a per injection basis. That accounts for any loss of patients and those were the rates that I presented to you today. So, on that basis it does go down. Slide 129.

[Slide]

Just to show you the data, you can see that prior to the amendment on a per injection basis it was 0.18 percent, and then post the amendment it was 0.03 percent but with that additional confounding variable of a lot of off-label steroid injections going on.

DR. DUNBAR: Dr. Gates?

DR. GATES: In the context of the cases of endophthalmitis, could you expand on the initial injection technique versus some of the changes that you made secondarily? Because draping oftentimes means different things to different people.

DR. ADAMIS: Correct. The details of the injection procedure are on a slide but let me see if I can recite them from memory for you, the

121

changes. There was a requirement for the installation of an antibiotic drop or dilated povidone-iodine prior to the amendment. Then, after the amendment the drape that was specified is a clear plastic one that adheres around the lids and the lashes, and then the placement of the speculum, and then also asking for a povidone-iodine flush to be done, and then patients received postoperative antibiotics. So, what we tried to instill there was a sense of uniformity in the procedure. There was more latitude prior to that. Those were the changes, to the best of my memory, that were instituted.

DR. DUNBAR: We have more than one-year data, but would you anticipate that the patients will continue every six-week intravitreal injections for the rest of their lives?

DR. ADAMIS: That is an important question. It is one of the questions we ask in the second year of the design. We want to kno