November 16, 2010: Vaccines and Related Biological Products Advisory Committee Meeting Transcript
FOOD AND DRUG ADMINISTRATION
CENTER FOR BIOLOGICS EVALUATION AND RESEARCH
Vaccines and Related Biological Products
Advisory Committee Meeting
November 16, 2010
Hilton Silver Spring Hotel
Silver Spring, MD
“This transcript has not been edited or corrected, but appears as received from the commercial transcribing service. Accordingly the Food and Drug Administration makes no representation as to its accuracy”
CASET Associates, Ltd.
Fairfax, Virginia 22030
Table of Contents
|Administrative Remarks - Jack Stapleton, Don Jehn||1|
Topic 1: Pathways to Licensure for Protective Antigen-Based Anthrax Vaccines for a Post-Exposure Indication Using the Animal Rule
CBER Introduction/Presentation of Discussion - Drusilla Burns
|NIAID Studies - Edwin Nuzum||13|
|CDC Studies - Conrad Quinn||34|
|CBER Summary - Drusilla Burns||55|
|Open Public Hearing||74|
|Committee Discussion and Recommendations||74|
Agenda Item: Administrative Remarks
MR. JEHN: Good morning. I’m Don Jehn, the designated federal official for today’s meeting of the Vaccines and Related Biological Products Advisory Committee. I would like to welcome all of you to these presentations. Dr. Stapleton is our chair today, as he noted.
Today’s session is open to the public except between 11:45 and 2:00, during which we will have a closed session. Tomorrow’s sessions will consist of presentations that are open to the public. These sessions are described in the Federal Register notice of September 29, 2010.
I would like to request that any media inquiries be directed to Ms. Shelly Burgess from the FDA Office of Public Affairs.
Now I need to read into the public record the conflict-of-interest statement for today.
The Food and Drug Administration (FDA) is convening the November 16-17, 2010 meeting of the Vaccines and Related Biological Products Advisory Committee under the authority of the Federal Advisory Committee Act (FACA) of 1972. With the exception of the industry representative, all participants of the committee are special government employees (SGEs) or regular federal employees from other agencies and are subject to the federal conflict-of-interest laws and regulations. The following information on the status of this advisory committee’s compliance with federal ethics and conflict-of-interest laws, including but not limited to 18 U.S. Code 208 and 712 of the Federal Food, Drug, and Cosmetic Act, are being provided to participants at this meeting and to the public.
FDA has determined that all members of the advisory committee are in compliance with federal ethics and conflict-of-interest laws. Under 18 U.S. Code 208, Congress has authorized the FDA to grant waivers to special government employees and regular government employees who have financial conflicts when it’s determined that the agency’s need for a particular individual’s service outweighs his or her potential financial conflict of interest. Under 712 of the Food, Drug, and Cosmetic Act, Congress has authorized FDA to grant waivers to special government employees and regular government employees with potential financial conflicts when necessary to afford the committee their essential expertise.
Related to the discussion of this meeting, members and consultants of this committee have been screened for potential financial conflicts of interest of their own, as well as those imputed to them, including those of their spouses or minor children and, for the purposes of 18 U.S. Code 208, their employers. These interests may include investments, consulting, expert witness testimony, contracts and grants, CRADAs, teaching, speaking, writing, patents, royalties, and also primary employment.
For Topic 1 on November 16, 2010, the committee will review and discuss the pathway to licensure for protective antigen-based anthrax vaccines for post-exposure prophylaxis indication using the animal rule. This is a particular matter involving specific parties.
On November 17, 2010, for Topic 2, the committee will review and discuss the effectiveness of vaccinating males and females with Gardasil, manufactured by Merck & Company, for the prevention of anal dysplasia and anal cancer. This is a particular matter involving specific parties.
Based on the agenda and all financial interests reported by members and consultants, no waivers were issued under 18 U.S. Code 208(b)(3) and 712 of the Food, Drug, and Cosmetic Act.
Dr. Margaret Rennels is serving as the industry representative for Topic 1, acting on behalf of all related industry. She is employed by GlaxoSmithKline in Washington, D.C.
Dr. Theodore Tsai is serving as the industry representative for Topic 2, acting on behalf of all related industry. He is employed by Novartis Vaccines and Diagnostics in Cambridge, Massachusetts.
Industry representatives are not special government employees and do not vote.
In addition, there may be regulated industry and other outside organization speakers making presentations. These speakers may have financial interests associated with their employer and with other regulated firms. The FDA asks, in the interest of fairness, that they address any current or previous financial involvement with any firm whose product they may wish to comment upon. These individuals were not screened by the FDA for conflicts of interest.
This conflict-of-interest statement will be available for review at the registration table. We would like to remind members, consultants, and participants that if the discussions involve any other products or firms not already on the agenda for which an FDA participant has a personal or imputed financial interest, the participants need to exclude themselves from such involvement. Their exclusion will be noted for the record.
Dr. Stapleton, I turn the meeting over to you.
DR. STAPLETON: Thank you, Mr. Jehn.
Welcome to today’s meeting, where we will discuss pathways to licensure of protective antigen-based anthrax vaccines.
I would like to have the committee introduce themselves and state where they are from. Dr. Gellin, would you like to start?
DR. STAPLETON: Thank you.
Our first speaker today is Dr. Drusilla Burns from the FDA. I would like to go ahead and start the session.
Agenda Item: Topic 1: Pathway to Licensure for Protective Antigen-Based Anthrax Vaccines for a Post-Exposure Prophylaxis Indicating Using the Animal Rule
CBER Introduction/Presentation of Discussion
DR. BURNS: Good morning.
I would like to provide a short introduction to the session today. We will be discussing a pathway to licensure for protective antigen-based vaccines for a post-exposure prophylaxis indication using the Animal Rule. I know that some of you probably aren’t all that familiar with the Animal Rule. It is relatively new and hasn’t been used yet for a vaccine. But before I go into a little bit of background about the Animal Rule itself, let me tell you a little bit about anthrax and protective antigen-based vaccines.
Of course, anthrax is one of the most feared of all bioweapons. There are many reasons for this. Among them are the fact that the spores are very stable, they are very easy to disperse, and the disease, especially the inhalation form of the disease, has a high mortality rate if not treated.
In the inhalation form of anthrax, the spores are taken up into the lungs. Once they get to the lungs, they are engulfed by macrophages, which transport them to lymph nodes. While the spores are in the macrophages, they germinate, and eventually the vegetative cells will escape the macrophages and find their way to the bloodstream, where they produce very large amounts of a toxin known as anthrax toxin. It is believed that this toxin plays a critical role in disease progression.
This slide shows how the toxin works. It’s a tripartite toxin composed of protective antigen, or PA, and two catalytic moieties, lethal factor, or LF, and edema factor, or EF. PA first binds to the receptors on the cell surface, it’s processed, and at this point it heptamerizes and can bind LF and/or EF. Then the holotoxin form of the molecule is internalized. When it hits the acidic environment of the endosome, PA actually forms a pore in the membrane through which LF and EF can escape to the cytoplasm. LF cleave MAP kinase kinases and can lead to death of certain cell types. EF is a very potent adenylate cyclase and increase cyclic AMP to very high levels in the cell.
Because the toxin is believed to play such an important role in disease progression, the thought is that if you neutralize the action of the toxin, you will prevent disease. For this reason, most anthrax vaccines are based on PA, the non-toxic component of the toxin, with the idea that if you use PA as an immunogen, you will elicit toxin-neutralizing antibodies.
Currently we have one vaccine that is approved for a pre-exposure prophylaxis indication. That is AVA, or anthrax vaccine adsorbed. It’s given five times over the course of 18 months.
AVA is made from culture supernatants of an avirulent form of Bacillus anthracis, and the primary component is protective antigen. Also being developed are vaccines composed of purified recombinant forms of PA. As of yet, no vaccine is licensed for a post-exposure prophylaxis indication.
What would happen in a post-exposure scenario? In a bioterrorism event, exposure would likely come without any warning, and a full course of antibiotics -- that is, 60 days -- would be initiated as quickly as possible. Then the vaccine would be administered on an accelerated schedule.
There are a couple of things to know here. First, antibiotics kill vegetative bacteria very well, but they do not kill spores. Spores can lie dormant for long periods of time. So the vaccine would be given to protect against the residual spores that germinate after the full course of antibiotics.
To license any vaccine for a given indication, we need information that the vaccine is efficacious. This is a little bit difficult to get for anthrax vaccines because, obviously, human efficacy studies are not feasible because of the low incidence of naturally occurring anthrax and human challenge studies would not be ethical because, of course, the disease has a rapid progression and it’s lethal in nature. In cases such as this, a new rule was promulgated in 2002 that allows animal efficacy data to be used as a basis for approval when human efficacy studies are not ethical or feasible to conduct. This is known as the Animal Rule.
The Animal Rule states that FDA may grant marketing approval of a product for which safety has been established in humans -- that’s a very important component; you still have to demonstrate safety in humans -- but you can approve based on animal studies if they establish that the product is reasonably likely to provide clinical benefit in humans. So that’s the standard of the Animal Rule.
In order to use the Animal Rule, however, four criteria must be met. I have listed them verbatim here from the regulations, but let me just go through and paraphrase a few of them:
- First of all, you have to understand how the organism causes disease and how, in this case, the vaccine would prevent the disease -- or have a reasonable understanding of that.
- Secondly, in most cases, you need more than one species and show protection in more than one species.
- Third, the endpoint must be related to the desired benefit in humans. In the case of anthrax vaccine, this would likely be enhancement of survival.
- Finally, the fourth criterion of the Animal Rule, which we will spend most of the day discussing and which is the most difficult criterion to fulfill -- let me read it verbatim to you: The data or information on the kinetics and pharmacodynamics of the product or other relevant data or information in animals and humans allow selection of an effective dose in humans.
What does this mean? That means, if you take the animal data together with, in the case of vaccines, human immunogenicity data, you will be able to determine that the vaccine dose that you are giving in humans is reasonably likely to provide clinical benefit. The Office of Vaccines at CBER has taken the position that in order to determine whether a vaccine dose will be efficacious in humans, the vaccine dose given to humans should elicit an immune response in humans comparable to the immune response achieved in animals that were protected by the vaccine.
CBER has cosponsored several public workshops where scientists in the field have gathered to discuss how best to implement the Animal Rule in regards to anthrax vaccines. The first of these was held in April 2002 and the second was in November of 2007. At these workshops, the discussion led to consensus on certain very important points regarding the four criteria of the Animal Rule.
First, in regards to the first criterion, the pathogenic mechanisms of the Bacillus anthracis were discussed at length, as was information on the mechanism of protection by PA-based vaccines. They were thought to be reasonably well understood.
Secondly, in regards to using more than one animal species to show protection, all of the available animal models were reviewed, and it was felt that nonhuman primates and rabbits are appropriate animal models to use in the case of anthrax vaccines, since the pathology and the immune response are similar to those seen in humans.
Third, regarding study design, in those workshops the animal study designs were discussed and agreed upon, including the route of exposure, which would be aerosol, and the endpoint, which would be survival.
Importantly, at those workshops consensus was reached that the first three criteria of the Animal Rule had been fulfilled.
Which brings us to the fourth criterion, the one that I think is the most difficult to fulfill. That is, how do you put the animal data together with the human immunogenicity data to give you confidence that the vaccine will be reasonably likely to provide clinical benefit?
A number of questions in regards to this issue were discussed at these workshops. For example, what immune markers should be used to link animal protection data to humans? I think many of the participants felt that antibodies would be an appropriate link. Other questions were discussed dealing with animal study design and how animal protection data should be bridged to humans. There was a lot of good discussion, a lot of good ideas. But I should note that there was really no attempt made to come to consensus on these points. That’s one of the reasons why we are here today, to hear your thoughts regarding how best to approach this issue of how animal protection data should be bridged to humans.
Why is it important to come up with a strategy for doing that? It would provide manufacturers with a possible pathway to follow during the development of their PA-based anthrax vaccines for a post-exposure prophylaxis indication. That is not to say that this would be the only pathway they would be allowed to take. Obviously, any scientifically sound strategy would be accepted by CBER. However, by giving them a pathway that has been appropriately vetted by this committee, I think it gives them much more confidence as they move forward that they are going in the appropriate direction. Guidance regarding these bridging strategies would really help them in their design and execution of pivotal studies, thereby facilitating and expediting vaccine development.
Today, for the rest of the morning, we are going to hear from both NIH and CDC regarding the animal and human studies that have been conducted that give you an idea of the data that is out there and information that is available.
We are first going to hear from Ed Nuzum, who will tell us about the NIH animal studies that have been done. Then we will hear from Conrad Quinn, who will tell us about CDC nonhuman primate and some human studies. Then I will come back and give you the CBER perspective on this.
DR. STAPLETON: Thank you, Dr. Burns.
We’ll go straight to Dr. Nuzum’s presentation, from NIAID.
DR. NUZUM: Thank you, Drusilla, and I certainly want to thank CBER for inviting us to present this morning. We have done a lot of work, and we are happy to present our efforts related to today’s VRBPAC meeting.
I also want to note that this DMID effort was led by a small core group within DMID consisting of myself, Judy Hewitt, and Freyja Lynn. Freyja essentially singlehandedly handled all the assay aspects. We really aren’t talking about assays today, but you can imagine that that is a critical component. Those assays were originally developed at USAMRIID, further developed by CDC. We took them from CDC and developed them further. Many of those are validated now. So Freyja deserves a lot of credit for that effort.
I will be talking a little bit about our meta-analysis effort that’s ongoing. Much of that data comes from DOD, USAMRIID, as well as CDC. For some of you that have heard me talk before, some of my slides will be familiar. I think that’s okay. I think the concepts are important. I think they bear repeating. I think we are delivering a consistent message. The more we talk about it, the more we get it out there, the better it is for the community as a whole.
In the presentation today I will be talking about general background, general DMID/NIAID efforts. The data and concepts I’m going to present are meant to be independent of sponsor-associated issues. This is a concept that we have tried to apply from the very beginning. We have tried to involve sponsors all along, but treat them independently and equally. Even though DMID has led the effort, we have invited their input and involvement all along. In many of the calls in the early years, the sponsors were always involved, either both sponsors or one at a time, depending on the agenda.
Because of that approach, my presentation will not include vaccine comparisons, specific vaccine doses, or immune response metrics, or a discussion of how good is good enough.
There are a couple of underlying themes I’m going to be talking about. I highlight them here because I think they are very important. Those are correlates of protection and determination of the humanized dose. It’s important to point out that neither of these is specifically mentioned in the Animal Rule, but I think they are certainly implied by the Animal Rule, and if you understand the concepts, if you are able to apply these concepts properly, I think animal model development to support the Animal Rule will be much more efficient in the long run.
The other thing I want to mention here is that there is not a roadmap. There is not a step-by-step process to develop the data and the studies to support these concepts. They are integrated and they are overlapping. For example, in the early years of model development, we did dose-ranging studies, dose-response studies to develop the immune response-protection relationship. Then, as clinical data comes along, you can incorporate that data with the non-clinical data and come to the humanized dose. I’m going to come back to this later, but I just want to point out the importance that I think these two areas represent.
As I said, this has been a DMID-led group. It has been rather unofficial. We have referred to it as the Animal Studies Group. Key elements have included participation by funding and regulatory agencies, product sponsors, and SMEs. Weekly and monthly teleconferences provided for review of draft protocols, standardization of protocols, real-time data review, which I think was important to facilitate discussion regarding next step and subsequent study start. Essentially, as I said, this has been U.S. government-driven, but we have included sponsor awareness and input at every step.
The purpose of our program, of course, has been to support approval of new vaccines using the Animal Rule and developing those tools necessary to do so, and to develop models, generate data, analyze and extrapolate that data in a manner that will provide for prediction of vaccine efficacy in humans. More recently, we have incorporated the meta-analysis which I mentioned, and in that we are trying to bring in as much data as we possibly can, including, as I said, CDC and DOD data.
For example, this slide gives a summary of all the studies that are being considered in that analysis. We have data from all these studies. You might think this is a lot of work and a lot of studies. In fact, it is. But keep in mind that this has occurred over many years, multiple species, multiple vaccines, multiple indications. Our efforts and CDC and DOD efforts are all combined here.
From the meta-analysis, this slide represents the high-level, main take-home message or conclusion we have come to so far. It shows the immune response, in arbitrary units, on the x-axis, with survival on the y-axis. It clearly shows a positive correlation between the dose response and survival, protection for both AVA and RPA in multiple species.
I’m not going to talk a lot more about statistics today, but I do want to mention that the statistical input from NIAID statisticians, CDC statisticians, Battelle has been very valuable for all of the presentations and analyses, interoperations that we are presenting. We have also had valuable input from statistical consultants Bob Kohberger and David Madigan up at Columbia. Actually, our meta-analysis is based on the CDC AVRP data analysis that David Madigan has done of the AVRP program.
Before going much further, I think it’s important to talk about terminology. We realized early on that this was a confusing topic. We have talked about this in our previous workshops. This is a table published by Tom Fleming in 2005, but it’s modified by Bob Kohberger, who is here today, I believe. He modified this to incorporate terminology relevant to vaccine terms. There are several points I would like to make.
One is, based on this terminology, a correlate is different than a surrogate. If you define a surrogate as an endpoint that provides a variable that explains all clinical benefit, then that makes a pretty high bar to reach for the Animal Rule. Correlates can be further divided into a predictive correlate and a basic correlate, where you have just a statistically related clinical endpoint, but if you have a predictive correlate, it’s an endpoint that is reasonably likely to predict clinical benefit. It’s mechanism- and science-based. If it works properly, if it’s a good predictive correlate, it could be used to make predictions across data sets -- multiple species, multiple vaccines, and so forth.
This is our goal. This has been our goal, and this is where we think we are for these efforts, working in this level 3 predictive correlate arena.
This slide is just to further demonstrate this point. Looking at this cartoon down here, if you are doing a clinical efficacy trial, where you can look at endpoints that will indicate clinical benefit, and you can evaluate that endpoint in the same population in the clinical trial, then a biomarker or a surrogate will make that direct bridge for you. In our case where we are using the Animal Rule, we need a bridge that connects clinical benefit in humans to clinical efficacy in animals. For anthrax, in the data that we are going to show today, we suggest the immunological response provides that bridge. So between that bridge and the clinical endpoint in animals, you need a correlate of protection. That’s what we will be talking about today.
Once you get this correlate and this data set, then you can apply it to a vaccine efficacy analysis to come up with predicted efficacy in humans. That’s essentially a relative risk analysis that looks at probability of survival or non-survival with and without vaccine.
The Animal Rule and our approach for applying it: Drusilla has already gone through this, so I’m not going to go through it in detail, but I like to show this slide for a couple of reasons. One is to remind us that the main goal is to provide a way to assess and measure prediction of clinical benefit. The other point I make on this slide is that the Animal Rule is really about humans. Essentially every element deals with humans. I think that in designing studies, interpreting results, every step of the way, you need to think about the human response to both disease and the countermeasure.
Finally, I’ll reiterate Drusilla’s point that this step 4 is the most challenging element here. I think that is because of relevance. That’s something else I’ll talk more about.
To make the model relevant, we have come up with this terminology called “humanized animal dose.” Why is that important? The model needs to be relevant to humans. As Drusilla indicated -- and I’ll try to say it another way ‑‑ if you are looking at endpoints in animals that are parameters and measures that exceed those attainable in humans, the animal data is minimally useful, if useful at all. So it needs to be relevant.
You all need the humanized animal dose for advanced studies. There are any number of studies that will come in the Phase III stage of development -- pivotal efficacy, immunity duration, challenge breakthrough, interference/interaction, time to protection, label changes. The list goes on and on. Those studies are critical. They will be critical for approval. A lot of money is spent, and they need to be done properly. That requires that you use the proper dose, the relevant human dose.
What is a humanized animal dose? This is the definition that a group of us came up with a while back. It seems to have held up pretty well over time. We define it as a dose of final-formulation vaccine that in the animal model elicits protective immune responses which are at or below those achieved in humans, when humans are given appropriately safe dose and regimen of the same vaccine.
This is similar to what Drusilla said. This is a less formal view, but I think it essentially has the same meaning.
We also think it’s important to point out that it’s more important that a humanized dose be obtained than how you obtain it. For example, if you have to dilute the vaccine to get titers in animals that are relevant to humans, then that’s what needs to be done.
Finally, data is needed as soon as possible to guide model development. If you don’t have the clinical data, you don’t know how to really assess and interpret the animal model results. It’s needed, again, to develop the immunological bridge.
If we have immune response data -- and in this case, with anthrax vaccines, we do -- how might you use that data? I think that largely comes down to what kind of data you have. If it’s non-continuous data where it’s easy to identify a point above which all animals are protected or below which they are not protected, then it’s easy and would make sense to use a cutoff value approach. On the other hand, if your data is continuous and you can’t identify an endpoint like that, we would suggest that a logistic regression analysis be performed. That works because it develops this relationship between response to vaccine and protection -- i.e., clinical benefit -- and it does consider confidence intervals in all survivors and non-survivors.
What does our data look like? This slide shows the results for AVA used in rhesus. On the x-axis, the dots represent animals that die and the TNA titers -- again, arbitrary units -- on the x-axis. The dots at the top represent animals that survive. You can see that a few animals survive at very low titers, where others die. Similarly, some die at titers where others are protected. There is no clear cutoff here below or above which you are protected or not. We suggest that this is a continuous type of data set.
This is a similar slide that shows rPA in cynos. You have essentially the same results.
I think this might be another good place to mention that this is also why we say we have a good correlate, but not a perfect correlate. There are clearly things going on here that humoral immune response doesn’t explain. We see this in different scenarios, challenge at distant time points, vaccinations with some vaccines that protect that don’t produce titers. There are lots of examples of that. Nonetheless, there is a good correlation, and, as Bob says, this is why you use statistics.
This slide shows rPA in rabbits -- again, a very similar correlation.
I put this slide up again just to remind us of the difference and point out that we would suggest for our data sets that the logistic regression approach is the proper approach.
Many of you have seen this before. It does do a good job of -- it’s a picture that shows the relationship of what we are trying to do. On the left side, this might be the immunological profile in a human population from a clinical trial, for example. This slide shows the protection here, immune response here, the antibody-protection relationship in an animal study. If you develop this relationship from the animal studies, then you can go to the clinical population for an individual’s similar response and predict the probability of that person’s response.
In actuality, what the statisticians do -- and I don’t understand this -- is, they extrapolate these results across the entire clinical data population. That’s how you come up with the average probability of survival or predicted vaccine efficacy approach.
I know Drusilla is going to talk more about this later. It really is the basis of what we are trying to do. It’s a little difficult to understand, I think.
Now I want to move to an overview of our studies and results, to show that we actually have some data to support some of these things. The diluent that has been used in our studies in cynos has been a saline diluent, whereby we have a constant antigen/adjuvant ratio. In rabbit studies we use saline with adjuvant, which gives us a constant adjuvant concentration. The vaccines used have been recombinant protective antigens from various formulations. Of course, AVA BioThrax has been used, although most of that data comes from DOD and CDC.
The general-use prophylaxis scenario is one where vaccine is administered IM prior to challenge, and it’s meant to simulate the classical use of vaccines for prevention of disease.
This slide just gives an overview of the actual design. Again, these have largely been -- especially the early studies -- dose-dependent response studies to look at the vaccine immune response in relation to protection. The species we used were Louisiana white rabbits and cynos. Treatment groups are usually four to six doses, using two- to fivefold dilutions, six to 12 animals per group.
Just as an example of how this evolves, in our initial dose-ranging studies we might have used four- or fivefold dilutions. As we narrow in what we think are humanized doses and we know more about the protection response, we have gone to two- and threefold. So it’s an example of refinement and how the models evolve and mature.
Vaccination is given as two vaccinations, 4 weeks apart, challenged on day 70. This is one of those early decisions -- the target challenge dose -- made in our working groups and the workshops that we had. This lists all the various parameters that are looked at in these studies.
This slide shows the immune kinetics for two rabbit GUP studies that use different vaccines. Clearly you can see a dose-dependent immune response for ELISA.
This slide shows the dose-dependent protection in the same study. It’s pretty evident that there is a nice dose response.
This slide is a regression analysis for the same two studies. This shows the TNA curve. Again, we are not showing titers, but we are trying to show concepts here. Hopefully you will agree when we are done that a lot of this starts to look similar. That’s what we want.
This is a regression curve for the same two studies, but this is for ELISA at week 10.
I put this slide in to show that studies are reproducible. These are the immune response levels associated with 90 percent survival probability in three different studies. Just to reiterate the point, in the pivotal studies, these immune response levels are going to need to be levels that are relevant to humans and would presumably be produced using the humanized dose.
Switching gears to NHPs, this plot shows the TNA immune kinetics for an NHP study. Again, there is a nice dose-dependent response.
This is the same study, but shows the kinetics for ELISA.
This slide shows the protection response, again dose-dependent, for the same study.
This is the regression curve for that same study. In this case two studies are combined. We would like to suggest that the narrowness of the confidence intervals also suggests reproducibility.
This is for week 6. We also have week 10 regression data for TNA, and we have ELISA at 6- and 10-week regression data, all of which -- we aren’t showing those -- are very similar results.
This slide is here to show the similarity between NHPs and rabbits. Drusilla said that the community has agreed that rabbits and NHPs are the two models that are acceptable. The fact that they have this similar humoral immune response, I think, is very important.
Now I’ll talk about the passive immunization study design. This is to assess the protective efficacy of IP administered -- in this case, we used human AVA plasma or IgG purified from human plasma. It was done in rabbits, three study groups, three doses, 7, 14, 28 mg/kg. We dose on day 0, challenge at 24 to 36 hours, and look at the parameters indicated here.
This slide shows the survival curve for this study. Again, it is dose-dependent. These two lines represent the two highest dose groups. In this study the treatment wasn’t completely protective. But it is dose-dependent. It does confirm that the antibody is a good correlate and that there is a relationship between level and protection.
This slide shows three regression curves. These two regression curves are from active immunization GUP studies in rabbits and NHPs at 10 weeks. This blue regression curve is from the passive protection study. We thought we would put these together here because it shows the difference in order of magnitude. Essentially, the passive protection curve is shifted an order of magnitude to the right. So although there is a good positive relationship between antibody and protection, we don’t think that that passive protection model is useful for establishing correlates of protection.
Now I’ll talk about the vaccine post-exposure prophylaxis model. This is a scenario where it’s meant to simulate use of vaccine in a post-event scenario, where antibiotic and vaccine are administered concurrently. In our model the vaccine and the treatment start 6 to 12 hours after challenge. The objective then is to statistically demonstrate benefit of vaccination when combined with antibiotic, where you would get a partially protective regimen that keeps animals alive until the vaccine response develops so that when spores germinate, the protection is available there from the vaccine.
I think perhaps the most important bullet on this slide, and maybe one of the most important ones in the talk, is the fact that the antibiotic and vaccine regimen was not required to mimic human regimens. We have had considerable flexibility to do whatever is needed to get the results we need, with the major limitation that you don’t want antibiotic pharmacokinetics or vaccine immune response above that seen in humans. But beyond that, we have had a lot of latitude to develop this model independent of what is done in humans. Quite frankly, without that latitude, I don’t think this model would be possible.
This describes the model in general. Again, we use rabbits. We challenge on day 0. The treatment starts 6 to 12 hours post-challenge, where levo is given once a day for 7 days and the vaccine is given at days 0 and 7.
We monitor for 28 days. We monitored these parameters.
This slide shows the immune kinetics for three different studies, three different vaccines, using high, low, and medium doses. I put a black bar on here to indicate the days when most deaths occur. Note how that lines up with the rapidly rising immune response, here.
This shows the survival curves for the three rabbit PEP studies. Clearly there is a benefit seen from the vaccines when given in conjunction with antibiotic compared to antibiotic alone. These are the antibiotic treatment days; vaccine given days 0 and 7.
The NHP vaccine PEP model, unfortunately, is not such a great success story. We did try the same approach. The problem is that we see very good survival when any antibiotics are given. We tried various approaches -- the start time, adjusted durations, lower doses. We tried bacteriostatic antibiotics. But we have been unsuccessful in achieving a partially protective antibiotic regimen that would allow us to show added benefit of the vaccine when they are combined together.
The PEP summary is that the rabbit model is standardized and robust. Really, we feel like it’s ready for pivotal studies. But we will need to utilize the humanized vaccine dose in final formulation. The NHP PEP model has not worked out in our hands. We currently are not planning any further work with this model.
This slide summarizes the GUP data. There is quite a bit here, but I think it’s all important. We have conducted a data-driven iterative U.S. government-industry effort that has resulted in an extensive animal efficacy database for anthrax vaccines. This has encompassed multiple studies. It has spanned many years, multiple species, and multiple vaccines. We presented an approach for extrapolation of animal efficacy data to predict efficacy in humans, and we have shown a strong correlative relationship between anti-PA antibody and protection. This forms the foundation for future pivotal studies.
The passive immunization data -- although it does demonstrate a positive correlation between antibody and protection, we do not believe the model is useful for establishing correlates-of-protection levels.
Model studies should incorporate the humanized dose as soon as possible. To attain that, you need clinical data and final formulation as soon as possible.
I actually feel that if these two bullets are able to be implemented, the development of animal models to support the Animal Rule can be fairly efficient -- perhaps more efficient than we have been able to do -- because of the lessons we have learned and so forth.
This is a quote that Freyja likes. It says a lot in one sentence: All models are wrong, but some are useful. The way I interpret this is that we need to develop the best models we can and do the best we can, generate the best data we can. And that’s certainly what we have strived to do.
Finally, I want to mention -- there are far too many names involved in this effort to mention, but there have been many entities, inside and outside of government. We think it has been an interagency and government-industry partnership effort. Everyone involved has contributed, including our consultants.
I’m going to leave this slide up at this point, and I’ll be happy to take any questions.
DR. STAPLETON: Thank you, Dr. Nuzum.
Dr. Nuzum’s presentation is open to questions from the committee. Yes, Dr. Gellin?
DR. GELLIN: You may have mentioned it. Does inoculum size affect any of this? Has inoculum size been standardized as part of this? Does that matter?
DR. NUZUM: For a challenge dose?
DR. GELLIN: Yes.
DR. NUZUM: We have actually looked at that. The short answer is no. There have been various studies done. We have looked at that. Given all the other variables in the model, we don’t feel like that’s a critical one.
DR. GELLIN: This may be a related question, what somebody may be exposed to. The vaccine is to be administered within six to 12 hours. Does it matter if it’s much later than that? How might that affect any of this?
DR. NUZUM: Yes, it does. We have looked at that also. If you wait, then you lose the partial protection that you need. Either too many die or not enough die. The start time for that is critical. If you delay it long enough for the immune response to start from the challenge, but then have a treatment that provides survival, you get complete survival. That window is critical when you start. Similarly, we don’t want to start too soon. There was an effort not to start any sooner than we had to, because then it’s harder to describe it as post-exposure prophylaxis.
DR. STAPLETON: Dr. Moulton?
DR. MOULTON: I also had that concern about, in the PEP study, when the prophylaxis starts. When you look at the horrifying examples laid out in the emergent documents of whole cities being exposed, we are talking about maybe a week before people are going to get this PEP. I was just wondering how the rationale fits into this and if there is any other way to design the studies to deal with that kind of important delay.
DR. NUZUM: Right. In an actual event, obviously the situation isn’t going to be controlled as much as it is in our studies. Our studies, I think, had a pretty narrow focus, to show added benefit of the vaccine. The questions you ask really go more to use policy and the need to get antibiotics and vaccine out quickly. CDC could probably speak to this better than I can. There is lots effort and there have been lots of studies on how you do respond and get product out quickly. In fact, I think studies show that that’s probably a more important variable than anything else, starting as soon as you can. That’s really a logistics issue -- or you vaccinate the whole population. Then that issue goes away. But that’s probably not in the cards either.
DR. STAPLETON: I guess I have a technical question. How reproducible and good are the correlates of the ELISA and the nude assay over time?
DR. NUZUM: They correlate very well. I think Conrad has a slide that shows that.
DR. STAPLETON: If there are no other questions, we will move on to Dr. Conrad Quinn, from CDC.
Thank you, Dr. Nuzum.
DR. QUINN: Good morning, ladies and gentlemen. It’s a real pleasure to be here this morning. Thank you to CBER for the invitation. I’m Conrad Quinn, from the Centers for Disease Control and Prevention in Atlanta.
CDC’s anthrax activities cover a broad spectrum, from research, through clinical, through stockpile distribution and emergency response. The two areas that I’m going to focus on today are the Anthrax Vaccine Research Program, which was initiated in the late 1990s and continued until IND submission in 2009, and our AIGIV, anthrax immune globulin intravenous activities, which are animal models in rabbits.
The Anthrax Vaccine Research Program has been a huge endeavor over the last 10 years or so, involving not just 1,500 participants and human subjects, but multiple sites across the U.S., multiple clinical investigators, and multiple other government agencies, including NIH, FDA, and the Department of Defense in particular.
The AVRP focused on two parts
- A Phase IV human clinical trial, which is double-blinded, placebo-controlled, and focused on changing the route of administration of AVA, or BioThrax, and evaluating our ability to reduce the number of doses. This study was based on a pilot activity executed at USAMRIID in the 1990s by Phil Pittman and Art Friedlander.
- The second component to the Anthrax Vaccine Research Program is this nonhuman primate correlates-of-protection study, particularly looking at inhalation anthrax and the ability of a minimal schedule -- what we propose to be a minimal schedule -- of anthrax vaccine adsorbed to protect rhesus macaques for extended periods. In this case, it turns out to be 52 months.
The second component of AIG was a task that CDC accepted in the early 2000s in response to the 2001 letter anthrax to put anthrax immune globulin into the Strategic National Stockpile for emergency use. The data I’ll show today focus on our animal components to this activity, which is an ambulatory New Zealand white infusion model, using human anti-AVA serum from Department of Defense volunteers. Our target was to achieve proof-of-concept that this material, administered prophylactically at least, had the ability to prevent inhalation anthrax in this model.
Let’s start with the Anthrax Vaccine Research Program. We had three clear tasks, as laid out by our Institute of Medicine committee in 2001. This was primarily to document and assure the efficacy of AVA, to minimize the dose and optimize the schedule, and to find the correlate of protection for this vaccine. This is where the animal models feature prominently.
Specifically, our task was to find immunological markers that:
- Endorse the human clinical trial endpoint, which was the quantitative of anti-PA antibodies, IgGs.
- Confirm that human vaccines are protected -- no small order.
- Identify when that protection is achieved.
- Identify for how long it lasts.
The human and nonhuman primate, rhesus macaque, studies were designed to be integrated to the fullest extent possible with the technologies available. In the humans, we have our Phase IV human clinical trial. I will briefly touch on that, as the focus of the day is on the animal models. In parallel, we have our rhesus macaques, in which we did dose ranging and immunogenicity. We chose dose ranging to try and modulate the immune response of the macaques, to achieve good priming, moderate priming, and under-priming, and to try and build correlation curves, as Ed showed in his previous slides of logistic regression, a relationship between the magnitude or the duration or the quality of the immune response and the ability of the vaccine to protect animals for extended periods.
From both of these cohorts, we measured and built immunological profiles, both for humoral, for antibody responses, ELISA for quantifying IgG, and toxin neutralization activity for quantifying the functional components in that antibody response. We also looked at cellular immunological profiles, specifically T cell memory onset, T cell competence, and memory B cell development and duration through the study.
With rhesus macaques, of course, we were able to challenge and, between challenge profiles and immunological profiles, then build immunocompetence models, from which we developed immune correlates of protection, and connecting the human immunological profiles, both cellular and antibody, with the correlates-of-protection models from the animals, to build a predictive correlate -- to use Ed’s term from the previous presentation -- for humans, which will contribute another piece to the jigsaw in understanding how this vaccine works and the type of protection that we can expect to achieve with it.
This table shows the outline of the human clinical study overlaid with the parallels in the rhesus macaque study. In the human clinical study we had six groups. This is the original schedule from 1972, which we call 8-subQ, or for our interim analysis component, which we published in 2008, 4-subQ, which covers just up to the 7-month period, 6-month vaccination. Our first comparison was between subQ and intramuscular injections, the same schedule of 0, 2, 4 weeks, 6 months, 12, 18, 30, and 42 months. For the interim analysis, we collapsed these groups. We call them 3-IM, and what they have in common is that they do not have the week 2 dose. This is replaced with an injection of saline. We have placebo-controlled groups who received saline either intramuscularly or subcutaneously.
A major component of the clinical trial was the safety evaluation, which I will not discuss today. That is why we have IM and subQ controls and placebo.
Where the human and macaque studies continue to overlap is that the macaques receive this minimal schedule of 0, 1 month, or 4 weeks, and month 6, 26 weeks.
In addition, where we dropped booster doses in the humans, we challenged the rhesus macaques -- month 12, month 30, and at 42 months. In reality, due to technical issues in the early parts of the study, this was carried out to 52 months. So we have four-year protection data for this vaccine in this model.
Building the immunological profiles: As I alluded to in the earlier slides, one of our tasks was to modulate the immune response of the rhesus macaques by giving dilutions of the vaccine, to over-prime, under-prime, and have, hopefully, a gradation between those two extremes to relate to protection.
In the humans, we had variable schedules ranging from the 8-subQ, the original schedule, down to the 3-IM and a booster at three years. We quantified the antibody levels by enzyme-linked immunosorbent assay and their functional activity by toxin neutralization activity. We also looked at the Th1 and Th2 disposition in humans and rhesus macaques to determine if both arms of the immune system were being triggered by this vaccination. We looked at the onset and duration of T cell memory and their competence throughout the duration of the study, 43 months in humans and 52 months in macaques. Onset and duration of B cell memory -- this is an important component. When is long-term memory established, and how long does “long-term” last?
Finally, the holy grail, as Ed alluded to -- the data mining and bridging between the genera, correlates of protection in rhesus macaques by statistical modeling and then defining the surrogate or protective correlate of protection in our human vaccinees.
Looking at the neutralization assay and the ELISA assay, a question raised in the previous presentation was, how well do they correlate? These data show, on the x-axis, the anti-PA IgG as measured by ELISA, correlated with the neutralization activity of the same time points. These are from the human study. These are all time points, all groups, all dose schedules. We see a very strong positive correlation at all time points, with no discrimination between different dilution groups or study time points or schedules. So the relationship between magnitude of response and the ability of those antibodies to neutralize anthrax toxin in vitro is consistent for the duration.
The macaque data are very similar. They are not identical. There is a different relationship between humans and macaques in terms of neutralization power per unit of IgG. The macaques are slightly higher.
Looking at the macaque data, these are the dilution groups. We have a full human dose, a 1-in-5 diluted dose in saline, a 1-in-10 diluted dose, 1-in-20, and 1-in-40. It’s a little crowded here, but here’s the schedule -- 0, 1-month, and 6-month vaccinations. These are the classic textbook-like immune responses to vaccination given intramuscularly.
The important point from this slide is that in the three groups identified here, which are our human dose, 1-in-5, and 1-in-20, continued out for 52 months, we have sustained measurable, although low, levels of anti-PA IgG. These are all ELISA data. In the other two groups, we terminate the study at this point, 30 months, and challenge, and in the lower dilution groups, at 12 months. But again we see that, for as long as we are able to measure, we are able to detect antibody levels.
So myth busting number one: Although the response diminishes, it does not disappear. These are geometric mean concentrations.
Looking at the response to challenge -- and I have reduced some of the groups here for clarity -- here we have our human dose group in black, rhesus macaques 1-in-5 in red, and the 1-in-40, the lowest dilution group, lowest antigen load. What we see is that when we challenge these groups, either at 12 months, 30 months, or 52 months, we have robust anamnestic responses, exceeding the peaks achieved at the maximum response to vaccination -- significantly achieving. In this 1-in-40 group, we go from a circulating level at the time of challenge of less than 3 µg/mL to almost 2.6 mg/mL -- physiologically improbable, but we have confidence that that is a real response.
Likewise, in the human and 1-in-5 doses, we go from significant levels at the time of challenge, by geometric means, to significantly much higher in vaccination response levels post-challenge. Even out at 52 months, four years after the first vaccination, when we have again very low but measurable and significant levels of antibody circulating, phenomenal responses to challenge. These challenges are all targeted to be 200 to 400 LD50s of Bacillus anthracis Ames.
If we look at cellular responses, again focusing on macaques, the only ones I’m showing today -- although these are representative of the B cell responses as well -- are T cell competence, as measured by stimulation indices. Along the bottom we have the controls. Here are our vaccination points of 0, 1 month, and 6 months. What we see is that after the second vaccination, we have measurable, above-background levels of T cell competence-induced vaccination, which are elevated and remain sustained through the 6-month vaccination, out to the end of the study at 52 months. There is a lot of noise in here. There are not statistically significant differences between groups, but they are statistically significantly different from controls at all time points after week 8 and from month 6 onwards.
So T cell competence is maintained from the minimal schedule given IM for all dilution groups, even the lowest antigen loads.
If we look at the bottom line to these, how well did this vaccine perform in terms of challenge? These again are group geometric means. We actually had several groups at each dilution, staggered challenges over the period of the study. Here they are summed into one table for one point. What we see is that the undiluted human dose, not challenged at 12 months, but at 30 months and 52 months, gives us high levels of challenge, ranging from 100 percent at 30 months to 80 percent at 52 months.
In the 1-in-5 group -- again, not challenged at 12 months, but challenged at 30 and 52 months -- we have 100 percent protection. In these groups of eight to 10 animals per challenge, 80 and 100 percent are not statistically significantly different from each other, but they are significantly different from controls.
The 1-in-10 group challenged at 12, 30, and 52 months -- again, significant levels of challenge, 80, 66, and 60 -- a sort of trend here, but not statistically significantly different across time.
The 1-in-20 and 1-in-40 challenged at 12 and 30 months, respectively -- again, high levels of challenge, even at the lower -- high levels of survival, though not complete at the early time points with these low levels of antigen vaccination.
It’s worth pointing out at this point that in terms of the magnitude of response, measured by ELISA, the 1-in-5 dilution in macaques is roughly equivalent to a humanized dose. Measuring by TNA, the 1-in-10 dilution is roughly equivalent to a humanized dose, in terms of magnitude of response, comparing macaques to humans.
So significant levels of protection out to 52 months.
Let’s look briefly at the human data. These are the antibody levels measured by ELISA from our 1,500-person human clinical trial. In blue we have the minimal schedule of 0, 1-month, and 6-month priming, with a booster at 42 months, so three years after the completing of the priming series here at 6 months. This blue line is analogous to the schedule in the macaques.
The orange, or mustardy color, is what we call the 5-IM. This group received 0, 1 month, 6 months, and then a booster at 18 months, and no further vaccinations until a booster at 42 months.
The other groups received either full schedule, which is 8-subQ, full schedule subcutaneous, 8-IM, full schedule intramuscularly, or 7-IM, which had the full schedule expect the dose at week 2. 5-IM with the booster dose is in orange; blue, the minimal schedule; and, of course, our placebo controls.
In terms of characterizing the immune responses, the first thing to notice is that in the minimal-schedule group, out at 42 months post-vaccination we still have, as in the macaques, low but measurable and statistically significant levels of circulating antibody by geometric mean concentration. This represents 66 percent of the cohort at that time point. When we vaccinate -- these are pre-vaccination doses at 42 -- we get phenomenal anamnestic responses. This group here, the minimal schedule, goes from 6 µg/mL circulating antibody to 433. This is approximately twice the level of the original-schedule dose, 8-subQ. The other dose responses are graded in between. The 5-IM, which received the minimal schedule with a booster at 18 months, goes from 21 to 216. Likewise, the 7-IM goes to 254; 8-IM, 320; 5-IM, 310 -- so significant anamnestic levels of response at 42 months in face of a sterile challenge or a booster.
Looking briefly at the T cell responses in humans, it’s similar to the macaques. Control levels are along the bottom. After the first vaccination, at 0, 2 weeks later we are able to measure T cell competence, which is sustained following a 6-month vaccination, in all schedules for the duration of the 42-month study -- again, noisy data, but statistically significantly different from controls. So as in macaques, we see that the human T cell response is sustained for the duration of the 42-month study.
Moving next to our passive transfer studies in rabbits, we developed with the University of Illinois in Chicago this ambulatory infusion model for the pharmacokinetics component of the study, where the animals contain the drug substance in a little backpack and we have infusion over the period required to deliver the dose. The attraction of this is that it can be used for multiple infusions or multiple drug combinations.
For the AIGIV study, we took our human anti-AVA IgG fraction -- this was prepared by Cangene in Canada. The recipients of the vaccine had received at least four doses of AVA subcutaneously by the original schedule. The rabbits received one single intravenous infusion of variable length, depending on the dose. The materials were infused 24 hours prior to aerosol exposure, again with 200 LD50s of Bacillus anthracis Ames. We looked at three dose levels of the human AIGIV, and we used the Flebogamma protein control, recognizing that the volume and total protein stresses on the animals may be a significant contributor to the outcome of the study.
The ambulatory model allows minimal invasion and maximum access and multiple long-term fusions, should that be necessary.
The data: Here we are with different groups. The time points of the study are along the x-axis, up to day 14, and the human anti-PA levels measured by ELISA. In red and black we have our control groups, Flebogamma with no challenge and Flebogamma with challenge. In blue we have the highest dose level, which is 20 mg/kg, the medium dose level, 10 mg/kg, in red, and the lowest dose level, 5 mg/kg, in green.
One hour after infusion, the maximum Cmax recovered was 600 µg/mL for the high dose, 300 for the medium dose, and 145 µg/mL for the low dose. At the time of challenge, circulating levels were 290 µg/mL, 161, and 65.
We see that the clearance of the human IgG follows this curve. Interestingly, here, surviving animals in the low-dose group started to mount their own immune response. It’s reflected here in the cross-reaction between the human conjugate material, the reporter enzyme, and the circulating rabbit. We are in the process of confirming this as a rabbit-specific activity.
These are the levels of circulating antibody at the time of challenge. Notice that these are significantly different from the challenge levels in the active immunization challenge studies in the previous slides.
When we look at survival in these groups, 20 mg/kg, 292 µg/mL at the time of challenge, 10 mg/kg, 161 at the time of challenge, 5 mg/kg, 65 µg/mL at the time of challenge. A hundred percent survival in these two top groups, so these levels of circulating antibody protect when given prophylactically against 200 LD50s of Bacillus anthracis Ames. In the 5-mg/kg group, with 65 µg/mL circulating at the time of challenge, we have 56 percent survival.
Flebogamma as a total protein control also delays time to death significantly over controls, for a nonspecific effect, we presume. We did not detect any anti-PA IgGs in this material.
The point here is that these levels of 65 µg/mL circulating at the time of challenge do not give full protection. Yet these are levels that are significantly above those that we saw circulating at the time of challenge in actively vaccinated macaques in the previous study. For this purpose, we would say that although the correlates-of-protection aspect of this is that we know that antibodies are important to protective response, they are not the full picture. They are only one facet of the immune response. When given passively, they are required to overcompensate for the missing components, the cell-mediated components.
Taking these data and moving forward to what we can project to humans, the approach we have taken has been with David Madigan, as Ed alluded to earlier. David has done extensive analyses, not just of the macaque data, seeking the correlates of protection in that model, using least-squares regression analyses and Lasso procedures, but also Bayesian multilevel models to bridge between the genera, to bridge those data from the rhesus macaques to the humans.
This is worth an entire presentation in its own right, and as a non-statistician, I will not even attempt to go there.
What I will show are some additional data from our statisticians at CDC, specifically Chuck Rose and Lydia Foster, who have used linear mixed-effects models to look at the decay curves of these antibody responses, which brings us again to one particular aspect of bridging between humans and rhesus macaques.
What Chuck and Lydia have found is that if we model the decay curves of these responses, irrespective of the peak, in a linear model, plotted here with natural log IgG against natural log week, we find this interesting relationship. In the blue line we have the rhesus macaque responses to a full human dose. Overlying here we have the 1-in-5 diluted -- these IgGs are measured by ELISA -- the 1-in-5 diluted group responses’ decay curves, the 1-in-10, and the 1-in-20.
Underlying here is the human dose. What we see first is that the human dose in humans and the 1-in-5 diluted dose in macaques are superimposable, indicating that measuring from a modeled peak response to the third vaccination, these decay characteristics are essentially identical. They have similar intercepts and they have similar slopes. The human dose in macaques -- higher peaks because more antigen is delivered, but again the same decay characteristics. Similarly for the 1-in-10. Now the slopes start to drift away from the model, if you like, and as the antigen loads get different, so the slope starts to become significantly different from the human and the 1-in-5 diluted group in the rhesus macaques.
So how do we use these data? At a very simplistic level, if we now segregate those responses into what animals survived and what animals didn’t and use receiving-operator characteristics to establish a putative threshold, we find that a threshold of 74-µg/mL response gives us a good distinction between high levels of survivors -- 55 out of 58 animals above this line survive; 18 out of 36, 50 percent of the animals, below this line did not. This is a very simplistic approach to determining what a discriminating factor might be in terms of immune response.
The other interesting component of this chart is that animals with low responses and high decay curves tend to cluster in this lower quadrant, whereas animals with high responses and slow decay curves cluster in this quadrant. Just by eye, to put in a discriminate factor here, we see that we can start to segregate between what the characteristics of a good response are versus the characteristics of the response that might not protect you, even though it has significant levels of antibody.
But this is very much a work in progress. There’s a lot more to do. We are always very cautious about assigning magic thresholds above which you are protected and below which you are not. And here we can clearly see that 50 percent of animals below this particular threshold are protected. So again we would caution against magic thresholds of protection.
What did we learn from these studies? In the macaques we learnt that we get significant survival at 30 and 52 months post-vaccination with the humanized doses and the full doses. We know that in macaques and humans the level of antibody and its functional activity are very highly correlated, and they correlate with survival. From the plethora of things that we measured over the years, of the other assays, only the stimulation index was a reasonable correlate, but it was weak compared to antibodies alone. So the antibody measurement is our best correlate of protection at this time.
Nonhuman primate responses measured by ELISA at 1-in-5 in the 3-IM group are equivalent to those measured in the humans and fit the same decay curve. The peak response to vaccination, combined with the decay kinetics, might be a useful step forward in predicting protection in the nonhuman primates, and therefore bridging to the humans -- but a lot more work to be done.
Conclusions: Anti-PA antibodies are adequate, if not perfect -- and this is in agreement with the information that Ed presented previously -- not perfect correlate of protection to bridge between the genera. Bridging from animals to humans may be achieved from the active immunization studies, and the passive transfer studies, although useful, comprise only one facet of the immune response. They are required to overcompensate for the absence of cell-mediated immunity, and consequently may set unnecessarily high requirements for measurable circulating antibody from the active immunization schedule.
We do believe that GUP vaccine efficacy, general-use prophylaxis, does inform post-exposure prophylaxis, but we must also take into consideration that in a real event we will be giving concomitant antibiotics therapy with the vaccine. This is a component which has not been studied, to our knowledge, for this vaccine.
I’ll be happy to take questions.
DR. STAPLETON: Yes, Dr. Durbin?
DR. DURBIN: I have a question on the nonhuman primate study, just to clarify. In the 1-to-10 human dose, there was an excellent boost at challenge, but I believe 40 percent of the animals died. Was there any significant difference between the boosted antibody titers of those animals that survived and those that died?
DR. QUINN: Those data are for the survivors only, because the macaques go down in two to three days, before we can take samples for antibody measurement. Where we have been able to take measurement, there has been no measurable onset of antibody. Basically, the animals die too quickly.
DR. DURBIN: Thank you.
DR. STAPLETON: I want to announce that for this session the questions and answers should not cover material that will be discussed in the closed session.
Yes, Dr. Tacket?
DR. TACKET: In your rabbit passive infusion studies, it appeared that the larger doses suppressed the active responses at challenge. Am I interpreting your graph correctly?
DR. QUINN: In the context of the rabbits mounting their own immune response?
DR. TACKET: yes.
DR. QUINN: That’s correct. At the two highest doses -- we haven’t actually finished looking at the rabbit specific responses, so they may be hidden within that. Those were the circulating human levels. There is a cross-reactivity between the reporter conjugate that we used for humans and rabbits which we weren’t aware of when we started that study. So there may be a rabbit response in there that we have not yet seen. We are currently looking for it.
But on the basis of those data, yes. The answer is yes.
DR. STAPLETON: Yes, Dr. Gellin?
DR. GELLIN: This may be suggested in your last point there. Thinking about the mechanisms of post-exposure prophylaxis, in somebody who has been exposed, spores are waiting to germinate over time. Is there some evidence that the spores germinating and becoming toxins would provide an ongoing boost to the immune system?
DR. QUINN: It certainly did indicate that, yes. But I think in an emergency response, we would hope that people are protected before those spores germinate, although they may get a boost when they do.
DR. STAPLETON: Are there any additional questions from the committee?
I think we’re quite ahead of schedule. Given the timing of this morning’s meeting without a break, I’m going to make the decision to have a 10-minute break. We’ll reconvene at 10:40. Thank you.
DR. STAPLETON: I would like to call the meeting back to order. Our first speaker for the next session is Dr. Burns again, from CBER.
DR. BURNS: I will now give CBER’s perspective and hopefully set the stage for this afternoon’s discussion.
Of course, as I said earlier, we are really here to discuss how to fulfill the fourth criterion of the Animal Rule, and that is how to combine the animal protection data with the human immunogenicity data to assure us that the vaccine dose that is going to be given to humans is reasonably likely to provide clinical benefits. In other words, how should animal protection data be bridged to humans for a post-exposure prophylaxis indication?
I want to define exactly what we mean by a post-exposure prophylaxis indication, for the purposes of discussion today. That is defined as: for prevention of disease caused by B. anthracis spores in exposed individuals who have received a full course of antibiotics. Currently the approved regimen of antibiotics is 60 days.
As I said earlier, OVRR has taken the position that in order to determine whether a vaccine dose is reasonably likely to provide clinical benefit, that vaccine dose should elicit an immune response in humans that is comparable to the immune response achieved in animals that were protected by the vaccine. The first question that comes up is, what immune marker should be used to link animal protection data to humans for PA-based anthrax vaccines?
You have heard a lot of data from NIH and CDC. Just to summarize, I think what they have shown is that antibody levels correlate with protection in animal studies, and from their passive protection studies, antibodies alone, in the absence of other facets of the immune system, can protect. Now, that doesn’t mean it’s the only mechanism of protection that comes into play with active vaccination, but antibodies alone can protect. And it’s really not surprising, given the complexity and the redundancy of the immune system, that other facets might come into play in active immunization. But given the totality of the data, CBER believes that antibodies would be an appropriate immune marker to use to link animal protection data to humans.
If that’s accepted, the question is, how should protective antibody levels be determined? We heard about three different study designs today:
- The general-use prophylaxis study design, in which animals are first immunized and then challenged with B. anthracis spores, and then protection afforded by the vaccine is determined.
- Also there is the PEP study design, in which animals are first challenged, then antibiotics are initiated very quickly, and animals are immunized. After a short period of time -- just enough time for the animals to develop an immune response from the vaccine -- the antibiotics are terminated, which, in the case of the NIH rabbit studies, was after seven days, and then protection afforded by the vaccine is determined.
- Finally, there is the passive immunization study design, in which animals are administered antitoxin antibodies and then challenged, and the ability of those antibodies to protect is determined.
These study designs are somewhat different, but I believe they have some commonalities, especially in regards to the chronology of vaccine response and exposure. In the GUP study design, the exposure to spores is delayed until the immune response to the vaccine has developed. In the PEP study design, antibiotics are given, and those antibiotics kill the organisms that have germinated until the immune response to the vaccine has had a chance to develop. So, in a sense, in the PEP study design, exposure to vegetative bacteria is delayed until the immune response to the vaccine has developed. In the passive immunization design, exposure to spores is delayed until the antibody is in place.
CBER believes that all three study designs provide useful information in regards to vaccine protection and the role of antibodies in protection. However, the question that is key is, for a post-exposure prophylaxis indication, which study should be used to estimate protective antibody levels?
As I’ll go through in just a minute, estimating protective levels using GUP data is relatively straightforward. CBER would ask the question, in reality, is infection in humans caused by spores that remain after a full course of antibiotics fundamentally a GUP scenario? What I mean when I say that is that antibiotics are going to be in place. They will kill all of the vegetative bacteria that have germinated while they are being given. During this time, the vaccine will have 60 days to elicit an immune response, and that response has that time to develop. Only then would the vaccine be needed to prevent germination of residual spores that remain after that time.
While estimating protective levels using GUP data is relatively straightforward, as I will go through in just a few minutes, there are complexities that exist that make it very difficult to come up with a good protective level using PEP study design or passive immunization study design. I’ll make the case for that in just a few minutes. Before I do that, let me show you how you might estimate protective levels using GUP data.
In the GUP animal studies, animals are immunized and then the immune response is measured at relevant time points -- for example, perhaps peak titer and/or immediately before challenge. Then animals are challenged with spores, and survival is assessed. It’s possible then to determine the relationship between antibody levels and survival. You saw a number of these curves today when Ed gave his talk. You get a nice relationship between probability of survival and antibody titer in the animals.
Let’s switch over to the human response. You do your clinical studies, determine the population distribution of antibody titers in the human study population.
Then the question is, how do you link these two data sets together? We believe this could be done using an antibody bridge. In this case you would extrapolate protection in the animals to the human population. In this example the humans that have antibody titers in the purple region would be expected to have greater than 90 percent survival, humans having antibody titers in the blue region, between 70 and 90 percent survival, et cetera.
It’s important that when you make this extrapolation, antibodies are assessed at relevant and comparable time points in both the animals and the humans. For example, you might compare peak antibody response in animals to peak antibody response in humans. In doing this, it’s possible to assess whether the vaccine is likely to provide clinical benefit.
You could use these protective antibody levels to derive your immunogenicity endpoints in pivotal clinical trials. These levels might either be expressed as a threshold value -- for example, the proportion of subjects achieving the antibody level that protected a certain percentage of animals -- or, alternatively, you might have a clinical endpoint that predicted vaccine efficacy level. This was first proposed by Bob Kohberger at a workshop in 2007. I won’t go into exactly how you do that here, but basically you determine what the predicted vaccine efficacy would be for each individual and then take the average. You can then come up with a predicted vaccine efficacy.
That’s how you would do it using GUP animal data. How about using a PEP study design?
In the PEP study design, of course, animals are first challenged, then treated with antibiotics and administered vaccine, as appropriate. As you saw, in general, there are at least three groups: antibiotics only, animals that got antibiotics plus vaccine, and the control group that received no antibiotics and no vaccine. After a minimal amount of time needed for the vaccine to elicit an immune response, the antibiotic treatment is terminated and protection is assessed.
Here I’m showing again the data that Ed showed previously, the NIH data from three of their rabbit PEP studies. The black bar indicates the time at which antibiotics were administered; arrows indicate vaccine doses. The control animals that received no antibiotics and no vaccine died fairly rapidly after challenge. The animals that received antibiotics only -- about half of them that survived, in that they cleared essentially all the spores, and therefore survived. However, there were some animals that, after the antibiotics were terminated, had enough residual spores that germinated so that they died. The vaccine was able to protect against those residual spores.
In an ideal situation, we would like to estimate the protective antibody levels in animals using these PEP data just as we did with the GUP data and then compare those levels to those achieved by humans in the clinical studies. However, there are certain complexities that arise, and estimation of protective antibody levels is much more difficult than it was with the GUP model. Let me just go through a few of those complexities.
First of all, as you saw, the PEP model is a very dynamic situation, in which antibody levels are rapidly rising during the time period in which most of the animal deaths occur. Note that this is a log scale, so their antibody levels are going up around two orders of magnitude. So the question arises, what is the appropriate time point to estimate protective levels in the animals? Would it be day 7, the day that antibiotics are terminated? That’s probably likely to underestimate protective levels. Would it be better to use peak levels or the day that the last animal died in the antibiotics-only control group, which in this case is day 28? That has the possibility of significantly overestimating protective levels. What is the biological justification for picking any one of these time points over the other?
Secondly, there are technical limitations in the design of the animal studies that require that the animals receive antibiotics and vaccine on schedules that differ substantially from those of humans, as Ed went through in his talk. The question that comes up, then, is, what would be the relevant and comparable time point to assess antibody titers in humans?
An additional issue is that estimates for protective levels may be influenced by the methods used to impute missing antibody levels -- that is, those antibody levels for animals that actually died previous to the day that the protective levels are being assessed. This would add uncertainty to protective-level estimates.
Another problem is that antibody response appears to be influenced by infection. There is evidence from some PEP studies that antibody concentrations can be higher in animals that receive vaccine and challenge than those that receive vaccine only, which could result in overestimation of protective antibody levels.
When taken together, it appears that estimation of protective antibody levels using data from the animal PEP studies would be associated with a high degree of uncertainty. You may either significantly underestimate protective levels, which would not be good because it might allow a subpotent vaccine out on the market, or you might overestimate protective levels, which is equally bad because you could keep a good vaccine that has the potential to save lives off the market.
Now let’s move on to passive immunization studies. Those have told us some very important information. Both data in the literature and the data you heard from NIH and CDC this morning indicate that antibodies alone can provide protection against exposure to Bacillus anthracis spores. In an ideal situation, we would like to use passive immunization -- that design -- to directly assess human antibody levels that protect animals from challenge. It would be a very easy and straightforward way to get protective levels. But in reality there are certain complexities that exist that limit the usefulness of these data, in our opinion.
First of all, passive immunization study design, like the PEP study design, represents a dynamic situation in which antibody titers after infusion are actually dropping off relatively rapidly -- over an order of magnitude from the day of infusion to the day that the last animal died. What I have shown here is the CDC passive immunization study that you saw earlier.
Again, what is the appropriate time point to assess protective antibody levels? What is the biological justification for picking any one day over another to assess those levels?
As you heard from both Ed and Conrad, it appears that circulating antibody levels at the time of challenge are not the only mechanism of protection afforded by active immunization. I have shown the CDC data, both from their active immunization study and from the passive immunization study. If you remember what Conrad said, the circulating antibody levels at the time of challenge, which was at 52 months, were actually very low, and lower than those needed for protection in the passive immunization study. Obviously, some other aspect of the immune system is also contributing protection. One possibility is that the anamnestic response, antibody response, may partially be contributing to protection. The kinetics of that response are shown a little bit better from some of the NIH studies, in which it’s very clear that you get a response to exposure to the spores very rapidly, and certainly a robust response in the first week after challenge.
If you are using passive study designs to estimate protective levels, the levels you get from the passive study may have to include both the levels that are present due to active vaccination at the time of challenge and, possibly, some of this antibody response that is due to exposure itself. So passive immunizations, we believe, may have the potential to significantly overestimate antibody levels needed for protection.
How should we move forward? CBER is proposing the following strategy. This is what we are going to like your comment on this afternoon.
We believe that appropriately designed GUP studies should be conducted, and data from these studies would be used to estimate protective antibody levels in animals and extrapolate antibody protection to humans via an antibody bridge. Importantly, these protective levels would be used as a basis for defining clinical immunogenicity endpoints in the pivotal clinical studies.
Animal PEP studies would serve as a proof-of-concept that the vaccine can protect in a post-exposure setting, in which an individual has been exposed. They would be considered supportive, but would not be used to estimate protective antibody levels.
Passive immunization studies also would serve as a proof-of-concept that antibodies generated by PA vaccines provide protection. Again, these would be considered supportive, but would not be used to estimate protective levels.
Which brings us to the discussion point that I’ll present now. We won’t discuss this per se right now. We’ll leave that to the afternoon. But we are going to ask you is to please discuss whether CBER’s proposed strategy would adequately bridge protection data to humans to support a post-exposure prophylaxis indication for a PA-based anthrax vaccine. Let me remind you again what we mean by a post-exposure prophylaxis indication for the purposes of our discussion today. That is for the prevention of disease caused by residual spores in exposed individuals who have received a full course of antibiotics.
With that, I will stop and be happy to take any questions where I can clarify anything.
DR. STAPLETON: Does anyone on the committee have questions for Dr. Burns? Dr. Ferrieri?
DR. FERRIERI: That was a great summary, Drusilla. Why is the premise that one would be using in the post-exposure scenario the vaccine at a point when antibiotics would have ceased, in order to prevent disease by germination of residual spores? Why can there not be the concept that you would have simultaneous administration of vaccine when starting a course of antibiotics?
DR. BURNS: When I think about it, I think it’s pretty clear that antibiotics are very, very effective against the disease, while they are being given. Would the vaccine provide anything beyond that? I’m not sure. But where it really is important and where the vaccine would be very, very important is once you get rid of those antibiotics. That would be the major use for the vaccine.
DR. FERRIERI: True, but the antibiotic courses last typically 60 days in humans. During that time, one would conceivably want to immunize, prior to the last antibiotic. None of the scenarios and models that have been presented really deal explicitly with what the human situation would be -- the unpredictability of mass exposures, for example.
DR. BURNS: I think from the data in the literature from other animal studies, it’s pretty clear that vaccine by itself is not going to protect. That is probably a given. So what we are talking about is perhaps shortening the course of antibiotics. But I think what we want to do is leave the possibility of shortening the course of antibiotics as a discussion for another day. That’s a slightly different topic. It also involves our colleagues from the Center for Drugs. Today we’ll use this as the definition we’ll use for discussion.
DR. STAPLETON: Yes, Dr. McInnes?
DR. MCINNES: Drusilla, in terms of thinking about antibody, how much discussion has there been around qualitative differences in antibodies that are induced? How much work has been done in perhaps looking at isotype induction, avidity, those sorts of issues? I recognize very well that you have a binding assay and you have your functional assay and very good correlation. But I’m not sure how much work has been done, really, in trying to characterize the qualitative aspects.
DR. BURNS: I think that’s a very good point. Clearly one of the premises of going from the animals to the humans is that the antibody quality needs to be similar, if we are going to make that direct antibody bridge. How do you do that? What are the characteristics of that antibody response that are important in capturing quality? Certainly I think one of the beliefs of the community as a whole -- and certainly I think I can speak for CBER -- is that we felt it was very important to look at neutralizing antibodies, at least to begin with. If you get a correlation with total antibodies, then perhaps you can go from there. But that, we felt, was absolutely essential. If you compare neutralization capability of the antibodies from animals to neutralization for humans, you at least have that comfort that at least that aspect is similar.
What other characteristics are important is a little more difficult to get a handle on. Certainly from a research point of view, I think you want to look at as many aspects as you can. From a practical standpoint, where do you draw the line?
DR. STAPLETON: Dr. Debold?
DR. DEBOLD: Can you explain why the course of antibiotics is limited to 60 days?
DR. BURNS: I’m not an expert on the antibiotics and the duration of antibiotics, but I believe it was based on the data available at the time, certainly for how long spores may reside -- there were some animal data for that ‑‑ to give confidence that you at least get rid of the majority of the spores. There were also some human data that gave some indication of how long you may need to be on antibiotics.
I don’t know if any of our colleagues from CDER are here today who could elaborate more on that.
DR. STAPLETON: And I’m not sure that plays a role in decisions about vaccine, from that standpoint.
DR. FERRIERI: I’m not sure if Vicky’s question has to do with whether we need a vaccine at all, for example. Is that where you are heading? If you gave antibiotics much longer, would it completely eliminate spores?
I wasn’t sure of the basis of your question. But slightly tangential to it is that my memory is that in that small outbreak that we had in New York City several years ago, those who were exposed who were on antibiotics did not have a breakthrough and did not demonstrate that residual spores were going to germinate and kill them.
Do you remember those points specifically, Dr. Burns?
DR. BURNS: In 2001, especially the people who got the huge exposures -- I think a lot of this depends on the amount of exposure that you get. Those on Capitol Hill, who probably got a huge whiff of spores compared to some of the other people who were away from the primary source, actually were given the opportunity to take antibiotics for up to 100 days.
Also, to address that question, there are animal data out there. Again, we don’t know the lifetime of spores in humans compared to animals, but certainly if you look at the data available, there are animals that have died 60 days after original exposure, presumably due to -- I mean further out than 60 days -- presumably due to residual spores.
There is also a study that was done a long time ago -- I think in the 1950s -- by Henderson, where they actually looked at spores in nonhuman primates and they found them out to 100 days. Those animals only got eight LD50s. So if you are getting a much bigger exposure, I think there is the potential for residual spores germinating and causing disease after antibiotics.
DR. STAPLETON: Dr. Gellin?
DR. GELLIN: Just one other comment, for which I don’t have the data. But not everybody who is given a prescription for something takes it as prescribed, for a variety of reasons. I don’t know what the experience was in the anthrax response, but not everybody took what they were supposed to take, and some people, for a number of reasons, dropped out.
DR. STAPLETON: Dr. Moulton?
DR. MOULTON: I would like to go back to the previous question about this indication and the timing. Could you go to slide 16, please?
To me, this looks like simultaneous administration of the vaccine and the antibiotic. Is that right?
DR. BURNS: That’s right. That’s what would be the case with humans.
DR. MOULTON: Okay. Because your response made me think -- and also the way that indication was phrased, it sounds like it’s afterwards.
DR. BURNS: The vaccine would be given so it could be protective after the antibiotics. But you have to give enough time for the vaccine to elicit --
DR. MOULTON: Right, we are talking about giving vaccines during the course of antibiotics.
DR. BURNS: Absolutely, yes.
DR. FERRIERI: Thank you very much. That was my point. It wasn’t so clear to me, Drusilla.
DR. BURNS: I’m sorry.
DR. FERRIERI: I think this is critical. I’m a very pro-vaccine person. This is what I would want to receive right up front.
DR. BURNS: Yes. I should make that very, very clear. The vaccine would be given as soon as possible. The regimen would be started as soon as possible -- first antibiotics and then the vaccine, as soon as possible.
DR. STAPLETON: Are there any other questions from the committee?
If not, we will take a break to clear the room and have our closed session.
MR. JEHN: At this time, if we could just have the committee and Emergent BioSolutions and FDA stand fast, and everybody else clear the room.
(Whereupon, the open session of the meeting was adjourned, to reconvene at 2:05 p.m., the same day.)
DR. STAPLETON: The next item on our agenda is the open public hearing. Mr. Jehn will read us the information about open meetings.
MR. JEHN: As part of the FDA advisory committee meeting procedure, we are required to hold an open public hearing for those members of the public who are not on the agenda and would like to make a statement concerning matters pending before the committee. At this point we have received one written statement. Nobody has requested to speak.
Is there anyone in the audience that would like to make an open statement, any of the public who would like to make a public comment?
If not, we can move on.
DR. STAPLETON: If no one in the audience wishes to make a comment, we will move on to discuss and make recommendations about the topics today.
I believe Dr. Burns is going to proceed from here.
DR. BURNS: We’re about ready to get a slide up that will remind you of the discussion point. We would like you to discuss whether CBER’s proposed strategy would adequately bridge animal protection data to humans to support a PEP indication for a PA-based anthrax vaccine. For your reference, I have put two things on the right side of the slide. First, just to remind you, the definition of a post-exposure prophylaxis indication is, for prevention of disease caused by residual B. anthracis spores in exposed individuals who have received a full course of antibiotics. Also I have the proposed strategy listed out for you.
I don’t know if you want me to go through that again. It’s just up there for your reference.
DR. STAPLETON: Thank you very much, Dr. Burns.
I would like to point out again that any topic discussed in the closed session should be kept closed and not brought up in this open session.
Why don’t we start with Dr. Gellin and go around the table and ask if there are specific aspects of CBER’s proposed strategy that you have questions about?
DR. GELLIN: I think we’ll hear some of these. Some of the issues that have come up before are about estimations of efficacy in different sectors of the population, recognizing that there are specific age indications and not all ages may respond the same. Then maybe a little bit more discussion about, as Dr. McInnes brought up before, not just the level of antibody, but characteristics of antibody that might actually help to inform how well they work.
DR. STAPLETON: Dr. Romero?
DR. ROMERO: Back to my comment originally, is the proposed level of immunity adequate, given the lethality of the agent we are trying to prophylax against?
DR. STAPLETON: Dr. Baylor?
DR. BAYLOR: I just wanted to intervene here. I think we really want to keep the focus on our approach, and not get into any of that. This is really not --
DR. STAPLETON: Not specifics.
DR. BAYLOR: Right.
DR. STAPLETON: Which makes this a little difficult.
Dr. Rennels, do you have anything specifically to add?
DR. RENNELS: No. I believe the approach is sound. The big question is, how good is good enough? But we’re not going to go there.
DR. STAPLETON: Dr. DeStefano?
DR. STEFANO: I guess I have more of a general question about exactly what it is that is planned for this. Is this approach being proposed for the vaccines that already exist and data that are already collected or are you anticipating proposing more research to get at this question?
DR. STAPLETON: My understanding -- and Dr. Baylor can correct me if I’m mistaken -- is that what we are really trying to discuss is, is this approach of taking animal data that is proposed for this vaccine appropriate for additional studies?
DR. BURNS: Right. We want to make this very general. This would be for any PA-based vaccine, whether it even doesn’t exist now or not. Would this approach be appropriate? That is, if these studies were done and the results were evaluated, would the approach be such that you would feel that it would give you confidence that the vaccine would work in humans -- that is, extrapolating using GUP study data for a PEP indication? Are you convinced that that is something that would be appropriate?
That’s really where we want to focus today.
DR. STAPLETON: With that in mind, Dr. Gotschlich?
DR. GOTSCHLICH: I think the approach is sound. What, however, seems to be missing is an appreciation by people who present the data to really make every effort to present this in units that are understandable to each other. At this point in time, people use units that are different for each individual study. There are data that could tie them together, but it has not been done at this meeting.
DR. STAPLETON: Dr. Tacket, would you like to add anything?
DR. TACKET: Nothing specific, except this seems to me to be a logical and almost an intuitive way to do the bridging studies that we are talking about.
DR. STAPLETON: Dr. Durbin?
DR. DURBIN: I don’t have any further concerns or anything to add.
DR. STAPLETON: Dr. Cheung?
DR. CHEUNG: The same.
DR. STAPLETON: Dr. Lyons?
DR. LYONS: No. I think it’s very sound.
DR. STAPLETON: Dr. McInnes?
DR. MCINNES: I, first of all, have to say that I think this was the most spectacular briefing document that I have ever seen. It was just phenomenal. It was clear. It was thoughtful. It just was fabulous.
It struck me in reading it that I thought it was somewhat conservative in claiming that we had a surrogate. I think I felt a little bit more confident about that than the briefing document suggested. I think you are requiring with this approach, which I think is extremely well thought out and clever, under very difficult circumstances, to find a path here that makes sense. I think the burden of proof in demonstrating that antibody protects is pretty clear. More appears to be better -- dose response. I’m not sure that we are totally closed on the issue that quantitative assessment of immunogenicity is not alone in predicting efficacy. I think this move to look at functional antibody in the framework of a neutralization assay is good. I think some thought should be given to what ultimately for manufacturers will be a workhorse assay or whether they are going to run both of them or whether some are going to go with an ELISA showing correlation or some are going to go with TNA. I think it would be useful to have some guidance on that.
I think it’s important that they demonstrate an ability to boost at some time remote from the priming. I’m not convinced that this full booster dose really demonstrates induction of memory. I think there’s a very large amount of antigen being delivered -- which is the full dose -- being delivered in the boosting. I’m not sure that’s totally reflective of measuring memory induction and what might be simulated with somebody who has lower levels of circulating antibody that then encounters an organism. What are they going to do? I think it would be important to understand those kinetics. I think that needs to be demonstrated.
I have a question. If we accept that the passive immunization study -- it’s sort of a proof of concept that antibodies of certain characterization and amounts are associated with protection -- is it really necessary for everybody to continue to demonstrate that with their candidate? I don’t know the answer; I just raise the question. For other vaccines, passive protection has been demonstrated that showed that antibody could be protective, and people have moved on, not having to constantly do passive protection studies. So I ask whether it’s going to be necessary for each manufacturer to do that or whether we will get to the point of accepting that as a principle and then moving forward.
In general, I really like this proposed strategy of saying that for post-exposure you would be immunizing people concomitantly with administering the antibiotics so that, in fact, you can use a general-use protocol and all the principles that accompany that to see at what immunogenicity you might be sitting with and what level of antibody you might be sitting with two months remote from that time of immunization. I think that makes a lot of sense. I think it’s extremely thoughtful and it provides a pathway for people to move through.
I have probably talked way too long. Thank you.
DR. STAPLETON: That’s very helpful. Thank you. I think we’ll probably get back to a few of those points as well.
DR. WHARTON: I have to say, it was very nice to see the work presented this morning by NIH and by my colleagues at CDC. I think that impressive body of work really did lay the groundwork for a very impressive and elegant synthesis by you all in the briefing document. It’s a hard question: How do you find a licensure pathway for something that you can’t do clinical studies for? You can’t do the kinds of studies we usually do for vaccines. Thinking about how to implement the Animal Rule for licensure pathway is a difficult question. I think the approach that is outlined is a sound approach that does show a way forward by which vaccines for this indication could be licensed.
DR. STAPLETON: Thank you. Dr. Sanchez?
DR. SANCHEZ: I just want to agree. I very strongly support the proposed strategy. I think it sounds good.
DR. STAPLETON: Dr. Moulton?
DR. MOULTON: I’m just a little bit confused about the second criterion of the Animal Rule and what our relationship is to that. It’s really talking about whether there is one animal which is a good enough model or we need more than one species. I'm not sure what the sense is on that. I didn’t see anything in the briefing document about that. If there is no one model that everyone is completely comfortable with, then using two animals -- I think it would be useful to use one animal to bridge to the other and see how that works. If you can do that, then it will give you more confidence in bridging from a nonhuman animal to a human animal.
DR. STAPLETON: Dr. Ferrieri?
DR. FERRIERI: I’m reasonably comfortable with the strategy. I would say that I’m moderately happy about it. I think this is an extraordinarily difficult disease to develop a vaccine for, given that you can’t test it out in the field with thousands of kids or other population groups.
I agree very much with Dr. Gotschlich about the expression of protective data. I still have difficulty in accepting the extrapolation of animal protection and arriving at an arbitrary figure and predictability, based on well-established models of what the protective efficacy would be in humans. There I’m taking it more on faith rather than my being able to deal with concrete data.
I would add to Dr. Moulton’s comment about the choice of the animals. I really like having data from both, even if you are not going to apply the nonhuman primate data directly, depending on differences in approach by different organizations. I like the rabbit data, and I love knowing that there is nonhuman primate data. That helps me enormously in this whole concept of faith and bridging from one to the other.
The passive immunization studies we had recommended many years ago, before the release of the Institute of Medicine report in 2000. Some of us, including Dennis Kasper, on one of the two committees, argued vociferously for passive protection studies. I’m really glad they have been done. I can even imagine a practical approach to using high-titer immunoglobulin from vaccinated individuals in a clinical setting. I think this is extraordinarily important data.
That’s about all I have to say. I congratulate all of you on the work you have done in an extremely difficult area.
DR. STAPLETON: Thank you. Dr. Debold?
DR. DEBOLD: Clearly a lot of very good, hard work has been done. I think my primary concern here has to do with the differences in responsiveness across the two animal species that we have looked at. I’m a bit troubled by the nonhuman primate data. I’m concerned. That model, I think, needs to work, because the potential for possibly getting the dose wrong is critical. If there is some way to improve that aspect of the model for the PEP work, it would be great.
DR. STAPLETON: Thank you.
I too would like to congratulate FDA and CDC and NIH. They really did a great job in putting this together and in doing very elegant work.
I guess my comments are similar to what has been said by others. My take from this -- and it was pointed out in the briefing document -- is that the GUP studies -- this is a little different than other vaccines. For hepatitis A, passive immunization correlates very strongly with the type of antibody levels that you see in vaccinees, within a threefold range or so. That system, therefore, gives you some clue as to what a real protective level of antibody is, because you can protect 90 percent of people with post-exposure prophylaxis. That hasn’t held true in these.
So I think the post-exposure prophylaxis studies are very difficult, given the severity of the infection. Therefore, while they are helpful to show that it doesn’t interfere with the antibiotics and that there may be some additive benefit, I don’t know that they are as helpful as we would have liked. I think the passive immunization data also help us, in the sense that we know that the antibodies will work by themselves, but they clearly are not as big a part of the story for anthrax as it is for many other infections. So as far as bridging, I think use of antibody level per se from a passive immunization is not a good predictor. That has been shown by the different studies that have been presented this morning.
So I think it’s nice that GUP is really the goal here. Even though we are giving the vaccine post-exposure, we are really doing a GUP-type study. I’m very comfortable with that and have no problem with the approach and don’t see anything that’s going to blindside you on this, from my perspective.
That’s my comment.
Would others like to discuss specific issues that the committee or FDA or perhaps manufacturers should be considering as they move forward? Dr. Durbin?
DR. DURBIN: I just have a clarification and then a question. I think for the GUP approach, there are two acceptable models, both the rabbit and the monkey model. Would manufacturers then be required to use both of those in going forward with GUP as the primary pathway?
DR. BURNS: In the previous discussions we have had at the public workshops, I think people felt that there really are two animal models that are reasonable to use. One of the reasons that the Animal Rule is written like it is -- saying that in most cases you would use more than one animal model -- is because you are doing that extrapolation from an animal to humans, and by adding another species, it gives you more confidence that you aren’t making some big mistake in making that extrapolation. Currently we are thinking that you would need both models, just the GUP models.
DR. STAPLETON: I would agree with Dr. Moulton that that is an area that could be more directly analyzed as far as the comparison between the rabbit model and nonhuman primates and humans. The rabbit and nonhuman primate should also be compared.
DR. BURNS: I think it was alluded to today. One of the advantages of the rabbits is the numbers you can get, so you get statistical significance, whereas with the nonhuman primates it’s harder to get those larger numbers that give you very good confidence that your estimates for protection are as tight as with the rabbit.
But we hear you.
DR. STAPLETON: Dr. Lyons?
DR. LYONS: I just want to make sure people understand. I come from an animal model background. Some of the models presented today are spectacular -- I mean really spectacular. For trying to do this in the preclinical setting and clinical setting, the progress made in the last four years has been outstanding.
I would like every model to work. It would be great. But trying to get certain things to behave like we would like them to work is sometimes impossible, for a lot of different reasons.
I just want to congratulate everybody on that. The models are spectacular.
DR. STAPLETON: Dr. Gellin?
DR. GELLIN: We have talked a little bit about the age indications. If the setting for the use of this vaccine is post-exposure prophylaxis, I guess I would like to get some sense of where this goes next as far as trying to understand how these vaccines would perform in those younger or older than the age indications. I’m guessing this is some extrapolation with human immunologic, rather than younger or older animals. But since we are in an animal model discussion, I wanted to bring that up.
DR. STAPLETON: I see Dr. Burns is nodding her head. My take on this -- and if the FDA disagrees, I’m sure I’ll hear -- is that you are exactly right; this is basing an immunologic solid endpoint, like an antibody -- and there was discussion of quality of antibody. I think the fact that you are using a functional assay, the neutralization, really abolishes a lot of the -- there will be differences in affinity and maturation, depending on when you measure your antibodies, but I think the fact that we are basing this on a neutralization assay and that it correlates so nicely with the ELISA -- I would certainly be comfortable. I assume that’s how immunizations in different age groups, different patient populations, different doses, that sort of question about how you would apply this -- that it will be based on antibody in animal models.
I’m not hearing anything, so I guess they agree at the FDA.
DR. LYONS: I think you are raising an interesting question. Typically the experiments we have done in different animals have been not so -- the bar hasn’t been set so high. But what you really talking about is comparative immunology among species to get a better handle on what we can extrapolate and, more importantly, what we can’t extrapolate. That’s really what you come down to. I think those are potentially the future of research related to this kind of work. It’s going to be a critical issue down the road.
DR. STAPLETON: One kind of interesting immunologic question is, in the setting where you are using it post-exposure, what are the T cell responses? I think that really is very different for this vaccine. Are we really preventing infection, and are the spores becoming vegetative, or are we just -- presumably, we aren’t, but I don’t know that we know that. And T cells should tell you that.
DR. FERRIERI: As we move forward, I would advise FDA to be monitoring the gender distribution in human studies, because there is some information to suggest that the immune responses -- and this is nothing presented today ‑‑ may be altered in women. This may be a function of estrogen, hormonal levels, et cetera. I think we should be very sensitive to this in terms of judging immune response and efficacy in men versus women.
DR. STAPLETON: Are there any additional comments, questions?
This will be amazing. We may finish very early.
Dr. Baylor, are there any questions that you feel we should discuss that we haven’t?
DR. BAYLOR: My team says we have pretty much gotten what we came for. Really what we wanted to do was to try to bring this to a public discussion, be very transparent, so you know what type of approach we are looking at for these complicated vaccines, where you can’t do human efficacy trials. We really wanted to do this in an open, public forum.
We appreciate the comments you have given us. It sounds like we’re pretty much on track. We’ll keep you posted as we move forward on these vaccines. Thank you.
DR. STAPLETON: Thank you. Again, I think part of the reason this has been relatively easy and quick is that you have provided us with all this great information, and none of us have found any holes.
Thank you. We will adjourn and start again tomorrow morning at 8:30 a.m.
(Whereupon, the meeting was adjourned, to reconvene at 8:30 a.m., the following day.)