U

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

8:37 a.m.

DR. BERGER: Good morning. I'd like to welcome everyone to the FDA CBER Allergenic Products Advisory Committee Meeting. My name is Mel Berger. I'm a Professor of Pediatrics and Pathology in the Allergy/Immunology Division at Case Western Reserve University in Cleveland. I think our first step will be to go around the table and have everyone introduce themselves, and I guess we start with Marsha.

Oh, sorry. The microphones, you push to turn the microphone on, and then you have to push it to turn it off.

DR. WILLS-KARP: Good morning. Marsha Wills-Karp, Professor of Pediatrics at Children's Hospital in Cincinnati. My area of interest is the immunobiology of asthma and allergies.

MR. HAUCK: Peter Hauck, Alk Abello. I'm the industry rep for the panel.

DR. WEISS: I'm Michael Weiss, Dr. Michael Weiss, Clinical Professor of Medicine, University of Washington in Seattle.

DR. GRANADY: Dr. Lynelle Granady. I am in private practice in New York, and affiliated with Mt. Sinai Medical Center.

MS. DAPOLITO: Gail Dapolito, acting executive secretary for the committee.

DR. MACDONALD: Susan MacDonald, Professor of Medicine, Johns Hopkins Division of Allergy and Clinical Immunology, and Associate Chair of the Department of Medicine.

DR. NELSON: Harold Nelson, allergist, National Jewish Medical Research Center in Denver.

MS. OLSON: Christy Olson. I'm the nurse educator for the Allergy & Asthma Network, Mothers of Asthmatics. And I'm a consumer representative.

DR. PORTNOY: Jay Portnoy, Professor of Pediatrics at the University of Missouri, Kansas City School of Medicine, and the Children's Mercy Hospital. And I'm also the director, the Chief of Allergy.

DR. BERGER: I think I'll ask Jay and Ron to introduce themselves also, please.

DR. RABIN: Ron Rabin, senior staff fellow in the Laboratory of Immunobiochemistry in CBER.

DR. SLATER: Jay Slater. I'm the chief of the Laboratory of Immunobiochemistry at CBER.

DR. BERGER: Hopefully you've all seen the proposed agenda, and perhaps looked at some of the materials. We'll have several presentations by Drs. Slater and Rabin. We will have an overview of this issue of reclassification of allergenic products which had previously been classified as IIIA.

This will be an informational presentation to give a context to what the committee -- to what Jay and his colleagues at the FDA will be doing, and what the committee will have to consider in the future. But this is not for the purpose of actually making any formal recommendations at this time. So I don't think there will be any formal recommendations or motions to propose nor to vote on. I would remind you that if the committee wants to make a recommendation, then that is a very formal process, and so we don't normally have motions and things that follow sort of Robert's Rules of Order. Either we have discussion or we have a formal recommendation process. And again, there doesn't seem to be any action item which would require the latter at this meeting.

As you know, an important part of the meeting is to allow time for public discussion. We have set aside an hour for that. We do have a peanut gallery here for members of the public. No one has formally requested to appear, but if someone appears then we will give them a chance to have their concerns entered into the record and addressed.

I think now I'll turn it over to Gail who has some housekeeping details.

MS. DAPOLITO: Thank you, Dr. Berger. I'd like to read the conflict of interest statement for the meeting.

The following announcement addresses conflict of interest issues associated with this meeting of the Allergenic Products Advisory Committee on April 7, 2005. Pursuant to the authority granted under the committee charter, the director of FDA's Center for Biologics Evaluation and Research appointed the following individuals as temporary voting members for the committee discussions: Drs. Jay Portnoy and Marsha Wills-Karp.

The Food and Drug Administration has prepared general matters waivers for the following special government employees: Dr. Harold Nelson and Dr. Jay Portnoy, who are participating in today's meeting of the Allergenic Products Advisory Committee for the discussion of a proposed strategy for the reclassification of Category IIIA allergenic products being held by the Center for Biologics Evaluation and Research.

Unlike issues before a committee in which a particular product is discussed, issues of broader applicability such as the topic of today's meeting, involve many industrial sponsors and academic institutions. The committee members have been screened for their financial interests as they may apply to the general topic at hand. Because general topics impact on so many institutions, it is not practical to recite all potential conflicts of interest as they apply to each member. FDA acknowledges that there may be potential conflicts of interest, but because of the general nature of the discussion before the committee, these potential conflicts are mitigated.

Mr. Peter Hauck will be participating as the non-voting industry representative acting on behalf of regulated industry for this meeting. Mr. Hauck's appointment is not subject to 18 U.S.C. ? 208. He is employed by Alk Abello, and thus has a financial interest in his employer. FDA participants are aware of the need to exclude themselves from the discussions involving specific products or firms for which they have not been screened for conflicts of interest. Their exclusion will be noted for the public record.

With respect to all other meeting participants, we ask in the interest of fairness that you state your name, affiliation, and address any current or previous financial involvement with any firm whose products you wish to comment upon. The waivers are available by written request under the Freedom of Information Act. Thank you, Dr. Berger.

DR. BERGER: Sorry. Again, push the button to speak and push the button when you're done. And Dr. Slater will begin with an introduction to the FDA and the laboratory.

DR. SLATER: Thank you, Dr. Berger, and thank you all for coming. I think we have an interesting day in store for you today. We're going to be covering a lot of different topics. As Dr. Berger indicated, the core we think of the discussion is going to be the discussion later on this morning and early this afternoon about the reclassification of the IIIA products. But we do have other items of interest that we are going to cover today. And this is sort of an overview of the agenda. We will be giving a very brief lab overview. That'll take about five to 10 minutes. Then I will give the committee an update on the Cockroach Allergen Standardization Project, which is ongoing. Ron Rabin will talk to you for about a half hour, giving you an overview of our research activities in the lab. Then we'll talk about the reclassification of the IIIA products. Then finally this afternoon, after our discussions, Drs. Carbone and Foulkes will join us to give us a brief discussion of the Critical Path research activities in the Center for Biologics.

So let's talk about the lab overview. We'll talk about staffing, lot release, and reference maintenance. This is just an introduction to the people in the lab. And those of you who are here from my lab, when I mention your name just stand up and wave, please. I'm Jay Slater. I've been the lab chief now for six years. Ron Rabin, whom you'll be hearing from a little bit later, has been with us for four years. Our postdoctoral fellows, Bo Chi, Jinsong Zhang, and Nicki deVore have been with us for three years, two years, and two years respectively. Our research technicians in the lab are Mona Febus, Marc Alston, Cherry Valerio, and Katia Dobrovolskaia.

This is a cartoon or a graph that I've been showing since I came here. And the basic message is that after a few years of instability at the beginning, we've really had a fairly stable staffing of LIB. The only people counted on this graph are the research technicians. And the reason for that is that they're the ones who really bear almost the full brunt of the ongoing regulatory activity, the lot release, the reference maintenance activities. And we're really at a level now -- we lost Al Gam this past summer. He retired. But we really are at a level in which I think we're in very good shape in terms of handling the manufacturer's workloads. And I think our performance has reflected that.

What are our routine regulatory activities? Well, lot release is the main regulatory activity, but we also spent a lot of time on reference distribution and reference maintenance. And the reference maintenance includes semiannual checks to make sure that our references are maintaining their potency and their complexity. And in addition, when references either run out or clearly are not being maintained, we replace them.

Lot release activities. This past year we reviewed 417 protocols that were submitted by our manufacturers. In terms of reference distribution we sent out over 1400 vials and 83 separate shipments to our manufacturers this past year. Just to give you some context, over the last five years these are the lot release protocols that have been submitted to us. And you can see that 417 is really very much consistent with the previous submissions. And so we neither seem to be on an upswing nor a downswing in that. In terms of reference distribution, these are the number of shipments and these are the number of vials shipped. You can see that there was a significant drop-off between 2002 and 2003. In 2002 I made an appeal to the manufacturers to be more economical in their use of the references that we sent them. And they've obviously heeded my request. We're dispensing much less in the way of references than we had before.

I'd be happy to take any questions. If not, then I'd like to proceed to the next item of business, which is an update on our German cockroach standardization activities. Now, I spoke to you last year about our initial in vitro activities with German roach. I've given you in past years the rationale for standardizing German cockroaches. By the way, for those of you who are wondering and haven't seen these pictures before, these are Madagascar Hissing Cockroaches, in fact as large as this piece of birthday cake. And we are not working with them, thank goodness. These are the more standard American, German, Oriental roaches.

I gave you the rationale behind doing this. We've actually discussed this at length in previous advisory committee meetings. Just in a nutshell, roaches appear to be associated with inner-city asthma. A review that we did of available commercial German roach extracts indicated that they tended to be of very low potency, and of variable quality. And we decided to proceed with the next stage in this, which was to make a concerted effort to determine the biological potency of several German roach extracts using the ideal testing method. And I'm actually going to spend a good portion of this presentation reviewing the ideal testing method with you because it's not something that people are typically very familiar with. This is the method that was used in the 1990s and earlier to standardize grass pollen allergenic extracts. It is a method that utilizes highly allergic individuals. It is a quantitative method using serial three-fold dilutions in which the major effort is to establish the dilution at which the erythema response is 50 millimeters. And that dilution is called the D50.

When you do an intradermal skin test, you get both a wheal and a flare. For statistical reasons, this method uses the erythema response rather than the wheal response. The sigma E that we're measuring is the sum of the longest diameter, indicated here as Diameter A, and the midpoint orthogonal diameter, which is indicated as B. If you add those two up, you get the sigma E and those are the data that are utilized.

So the testing approach in this first of all is to identify those highly allergic individuals. And in order to do that you do a screening test, a straightforward puncture test using a bifurcated needle with a concentrate. You then go ahead and do the serial threefold dilutions, record the wheal and the erythema size, and calculate the sigma E at each concentration. What you get then is a plot, and this is a plot from a paper from the 1970s using histamine. You plot the mean sum of the erythema diameters of the sigma E on the ordinate. These are the serial threefold dilutions. The shorthand that is used here is that these are actually in negative log units. So a dilution of 4 is a 3^-4dilution, and so on down the line. You plot them out. Then going from the ordinate at the 50 millimeter line. You then figure out what the D50 was, and in this case the D50 was 5.56. Which means that a dilution of 3^-5.56this individual tested would have been predicted to have a sigma E of 50 millimeters.

Just to wrap up the background discussion on this. When testing was done previously, the normative response to highly potent extracts was found to be a D50 of 14. Therefore, a D50 of 14 was defined as 100,000 BAUs per ml. Working from that you can calculate the BAUs per ml by this simple formula from the D50. So when you do this method, you generate a D50 for the extract, and then you calculate the BAUs per ml based on this empiric relationship and the consequent formula.

Some of the rules that you follow when you do this, and I'm going to take you through a highly idealized set of D50 testing, and then we'll actually look at the real data and you'll see why you need rules to help you work this through. The skin test response, the sigma E, should fall within the limits of 0 and 125 millimeters. Each more concentrated dilution should produce a graded erythema response. In other words, as you give more and more concentrated extract, you should get larger responses. And the dilutions, at a minimum we want to have two dilutions on either side of 50 millimeters. In other words, we want to bracket the 50-millimeter point.

So I'm actually going to walk you through a theoretical set of skin tests at this point. Typically we start with a dilution of 3^-17or what's called the 17 vial. And in this study subject, we're testing three allergen extracts. One is the yellow one. The other is the blue one, and the last one is pink. The first series of tests indicated by the green arrow are five tests, one for each of the extracts, one for histamine, and one for saline. And what we see here is that the saline and histamine responses confirm that the subject is neither dermagraphic nor hyporesponsive. None of the three extracts appears to be eliciting any response. We plot it out here at the -17 log dilution. All of them are at zero. Therefore, in our protocol, in order to be economical in terms of the number of injections that we do for these study subjects who undergo a very large number of injections, we then if we get no response we skip three log dilutions rather than just one log dilution. So the next one for all three of these will be at -14. And here are the responses that we get. Both the yellow and the blue extracts continue to yield minimal or no response. But we can see with the pink one that we have a small response developing, which we can plot out on our graph. Therefore, the next injection for the pink one will be at 10^-13and the next injection for the yellow and blue will be at -11.

And here are those responses. Well, things are going well for both the blue and the pink extracts. The next injections for both of these will be at single dilution increases, going from 11 to 10 for blue, and from 13 to 12 for pink. However, as we plot out the yellow response, we see that we've made a mistake with yellow in that we now are already above the 50 millimeter point for the yellow extract. Since we need to bracket, we need to have at least two on each side of 50 millimeters, we're actually going to backtrack next with the yellow extract and go to dilution of -12.

And here's the next series of responses. Now we seem to be on track for all three of the extracts. We will now proceed with a single dilution increase for both the blue and the pink. However, the yellow one, if we went up a single dilution we would just cover ground we've already covered. So we can go to the next dilution for the yellow one. Here are the responses that we get for that. The good news is that we're now above 50 millimeters for all three of these extracts. However, we want to have at least two valid data points on either side of the 50 millimeters, so we're going to do another higher dilution. Which we can plot out this way.

Examining the data, we conclude that we now have clearly sufficient data for the pink extract. However, we have problems with both the yellow and the blue extracts, both of which are rectifiable. Blue for some reason, the next concentration, we seem to have plateaued. Therefore we're not getting a uniformly increasing curve, and we really need to go at least one concentration higher to get more valid data. With the yellow, looking back on the data, we see that we really only have one data point below 50 millimeters. And so we're going to go back and do one of these more dilute concentrations for the yellow. And what we can now see is that we actually have valid data for all three of them. And we can proceed to calculate the D50s for these.

Now the purpose of this exercise was to take you through an ideal case. I'm now going to take you into the project itself. We'll show you some of the data, how we tried to analyze it, how we decided alternative methods of analyzing it that might be better. Let's look at what our objectives were here. Our objectives were to take three commercially available allergenic extracts for German cockroach and really check in a good statistically designed study what the potency would be biologically. It was in our interest in doing this to pick three extracts that were different from each other, so we would get some idea of the spread of potency and try to relate this to some of the in vitro measures. So we picked three, and these are readily commercially available extracts from manufacturers A, B and C. They in fact were from two different sources. The source material preparation methods are outlined here. And the extraction methods are outlined here. Basically, there were no really great differences between these. They all started from German roaches. Some of them were used in a source material that was ground to powder before they extracted it. Others were ground in a blender during the extraction process. But fundamentally they were starting with the same materials and using very similar methods for extraction.

There are certain things that it's easy to do with allergen extracts and that is to try to measure known allergens, and to try to measure the relative potencies. That's what we did. INDOOR Biotech has a two-side ELISA method for measuring both Bla g 1, Bla g 2, and Bla g 5. We also used our standard competition ELISA technology to determine the relative potency of these three extracts compared to a highly potent extract that we had from some years back. We did this using pooled human serum, as well as rabbit sera that we made by injecting rabbits either with German roach or with recombinant Bla g's 1, 2, 4, and 5.

The specific allergen data are indicated here. E2CG is the reference standard that we have been using provisionally for some years in the lab that was made especially for CBER in the 1990s. As you can see here, all three of the commercial extracts contained Bla g 1. All three of them contained detectible Bla g 2, although there were some variations. Notably, only extract B contained Bla g 5. The other two really had almost non-measurable amounts of Bla g 5.

When we measured the potency using the three methods, RP1, relative potency 1, is the one using pooled monospecific rabbit sera against Bla g's 1, 2, 4, and 5. RP2 using rabbit anti-German roach serum. And RP3 using the human serum pool which we've called S1CR. Again, you can see that extract B appears to be the most potent by all three methods. This is consistent with the Bla g 1, 2, and 5 data that we had. Lots A and C appear to be somewhat equivalent in potency by using the pooled monospecific anti-sera. But when we use the human serum pool, Lot A appears to be almost devoid of specific activity, whereas Lots B and C have some detectible potency.

The clinical study to determine the potency of these three extracts could not have been done without the very active interest of the Division of Allergy, Immunology and Transplantation at NIAID. This was an IND that they put together. There were four sites. This is through the Inner-City Asthma Consortium. The four sites were in Baltimore, Washington, D.C., Chicago and Denver. And people often save thanks for the end of talks. I think the thanks here should go smack in the middle of this talk. This study really could not have been done without the real interest and commitment of the Inner-City Asthma Consortium. Individuals who were extremely helpful included Robert James and Herman Mitchell at Rho, which is a company that does the organizational work as well as the statistical analyses for the Inner-City Asthma Consortium. Peyton Eggleston was the PI in Baltimore, Andy Liu the PI in Denver, Jacqueline Pongracic in Chicago, and Sampson Sarpong in Washington, D.C. These folks all worked very hard to make this project work.

The purpose of the IND was to determine the biological potency of three commercially available German roach extracts and to test their bio-equivalents. The patient population was the patient population of the Inner-City Asthma Consortium: adults with a history of allergenic disease or asthma and a demonstrated sensitivity to German roach allergen. It was a multi-center open label trial. We planned to enroll 61 study subjects. We actually enrolled 62. This was an adult group. This was not a pediatric trial. It was predominantly female, and the ethnic background of the study subjects is precisely what you would expect from the Inner-City Asthma Consortium.

Once we accumulated the data, which actually this group completed patient enrollment and data collection within five months. It was really quite impressive how quickly it happened. We then were confronted with how to go about analyzing the data. Now, linear regression is the standard method that was really included in the protocol. This is what was used in the ideal protocol in the 1990s and before. The idea here is to draw a best-fit line through the data, determine the D50 in that manner. And the problem was, as we looked at the data, it was quite clear there were a number of study subjects for whom there would be extremely poor correlation coefficients. This was not surprising. In previous studies using this protocol, poor correlation coefficients were a major source of data invalidation. And in the current study it was clear that several of our study subjects were non-linear. And this is just data from eight study subjects. What you're looking at here in the black is the actual data points, either a solid line or a dotted line or a dashed line. Each of those is one of the allergen extracts, and each panel is one study subject. And you can see that in spite of the fact that all of these study subjects were screened with puncture skin tests, some of them were essentially flat lines. They really never got above the 50-millimeter point at all. They were hyporesponsive. And obviously, regardless of the analytical method you use, you can't include these study subjects in your analysis.

Some study subjects had curves that were -- it seemed pretty reasonable to do linear regression analyses. But others it seemed almost impossible to attempt. So Dr. James, the statistician at Rho, thought up another approach, and that was just a linear interpolation approach. In the linear interpolation approach, rather than assuming a mathematical relationship, the simpler method would be simply to take the serial dilution at which the plots cross the 50-millimeter line. And basically what you're working with in that situation is two points: the point below 50 millimeters and the point above 50 millimeters. And you use that to determine your D50. Now, if you remember some of the images of the actual data, you can see that that can be problematic as well. And so we established a set of rules by which one could use this method in an organized way. And then we applied it to the data to see how the data would look.

The first rule is that you use the first crossing of the 50-millimeter line. So study subjects that would go up and down, you would look at the first crossing that was followed by at least two consecutive non-missing dilutions above. So, if you went up and down, you would use the first crossing that was followed by a consistent staying above the line. Now, if the above criterion doesn't apply, and the data's most concentrated dilution is above 50 millimeters, we assume that any more concentrated dilutions, had they been collected, would have remained above 50 millimeters. In that case, we would be using the last crossing as the actual crossing that we would record for the D50 purpose. And if the extract for a subject does not cross the 50-millimeter line at any serial dilution tested, or whenever the line is crossed the first two criteria don't apply, then the D50 simply can't be calculated. And that makes intuitive sense.

So in this case, I've taken the same eight study subjects and we've simply plotted. And I know it's very hard to see, but maybe you can see it on your handouts, the points that were actually chosen, these are indicated in blue spots. Obviously, this one, there is no crossing, and so there are no D50 to determine. But what you can see here, and just by a direct comparison of the regression versus the interpolation method for two study subjects, you can easily see how the regression method generates a result that doesn't make nearly as much sense as the interpolation method does. Now, even in this study subject whose data is very hard to interpret, the interpolation method gives you something that certainly makes more intuitive sense than a projected value out here, or this value back here.

But there is another method that can be used, and I'm sorry, this is the one slide that didn't quite come out in translation to a different computer, and that's the four-parameter logistic model. The four parameters being the minimum, the maximum, the EC50 and the D or Hill slope. And this is also known as a sigmoidal calculation. You can see back here under these words the sigmoidal curve calculating the D50 in that manner. This has the intuitive advantage of using the most data. Remember, the linear regression uses all of the data, but the data at the extremes tends to dominate the calculations more than the data in the middle. The linear interpolation method uses the data right around the D50 the most and kind of ignores the data elsewhere. The four-parameter model is the most intuitively attractive because it uses all of the data. And you can either plot it out, or you can do it algebraically. When you're doing it algebraically you calculate the concentration at which a 50-millimeter point occurs. The D50 is then the log of this concentration divided by the log of three.

So, needless to say, what we did in this case, and you can see here again that several of the curves really look sigmoidal, and the projected or calculated curves which are in color really look very close to the actual curves themselves. So needless to say we did the data all three ways, and compared them. And this is the D50 data from manufacturers A, B and C by the interpolation method, the linear regression method, and the four-parameter fit method.

Looking very quickly at it, you can see that the data are not all that different in D50 values. But remember, when we turn them into BAUs they will be somewhat different. What also you can see is that the interpolation and the regression method have the advantage of using more study subjects data. Out of the 62 study subjects, you can see that nowhere near all of them had analyze-able data. If you're below the 50-millimeter line, the data can't be used no matter what you've done. But using the four-parameter method, we could only use between 42 and 43 study subjects, whereas by both the interpolation and regression methods, we could use between 48 and 55 study subjects. So that gives us an advantage for those two methods.

Looking over at the standard deviations, you can see that the interpolation and the four-parameter fit methods had substantially smaller standard deviations than the regression method. Which gives an advantage to the interpolation and the four-parameter fit over the linear regression. We went back and forth about this a number of different times, and came to the conclusion that the interpolation method really had the advantage over the other two. But I'm going to show you all three of the data.

So again, the potency data, calculated BAUs straight from the D50s by that calculation that I gave you before. You can see that by the interpolation method, the BAUs were 1,700 to 8,500. The regression method 2,100 up to 12,000. Four-parameter fit, 1,300 up to 5,000 or so. And these are the 95 percent upper and lower limits.

So the conclusions were that the D50s calculated using the interpolation of four-parameter fits had slightly tighter standard deviations than the regression method. The interpolation and regression methods maximized the number of study subjects whose data could be analyzed. Interestingly, bioequivalence which we had set at the beginning of the study as a delta of less than 20 percent among all three of the D50s, using all the methods, the three extracts were essentially bioequivalent. But, there were statistically significant differences between A and B, and between B and C. In other words, B was from a statistical point of view the most potent of the three, although by our preconceived bioequivalence limit they were all bioequivalent. And A and C were not statistically different from each other.

So the next relevant question is what in vitro test best reflects in vivo potency. Now, we are not there yet. We have not determined this yet. But we do have enough data to begin to try to address these issues. And what I'm putting here is all of the data put together in one place. These are extracts A, B and C. This is the Bla g 1 content, Bla g 2, Bla g 5. These are the relative potencies by three methods. And these are the BAUs, and we just used the interpolation data here.

One way to look at this in the aggregate is to simply normalize all of the data to product B, which was the most potent. So in this analysis you just take product B and you set it as 1, and then product A and C for each of these is just a ratio of that value to product B. And just by looking at this you can see that on first blush the BAUs and the Bla g 2 level seem to be the most closely related in aggregate. And this is an important caveat here. In the aggregate, the Bla g 2 data appear to best reflect the overall potency, but in order to do this we actually need to look at this study subject by study subject. And that's what we're going to do. We will be examining the IgE specificities of each of the individual 62 subject sera and correlated to their D50s.

But even though we're not finished with the process, we can start to look at German roach extracts and see where this fits in the constellation of other extracts that we know about their potencies. And just for review, the grass pollen extracts, with the exception of Bermuda grass, are 100,000 BAUs per ml. Ragweed extracts, based on an estimate of 350 Amb a 1 units per ml, equaling 100,000 BAUs per ml, are typically somewhere between 30,000 and 100,000 per ml. Bermuda grass, 10,000. Cat, between 5,000 and 10,000. And the geometric mean of those three extracts that we tested was about 3,300 BAUs per ml. So you can see here that when tested in a highly allergic population, the potency of the German roach extract is clearly lower than the others. Then the question is well it's lower, but is it still going to be useful. And the answer is probably yes. Skin testing doses -- this is from a practice parameter from 1995 -- have been estimated to be certainly 100 to 1,000 BAUs per ml. This can certainly be achieved with most of the German roach extracts. Immunotherapy doses are a little hard to find. This is from a review that Paul Turkeltaub did in 1999. This is from another practice parameter. If in fact we're shooting for 2,000 to 4,000 BAUs for dosing, that's going to be difficult to achieve with these existing extracts, and in fact will probably not be readily achievable with most of them. However, if we can get away with lower doses, then we probably can achieve that. This is something that will need to be the subject of future studies.

Another aspect in which it's ? it's just interesting to look at where this fits. I don't have to tell the people on this committee what my thinking is about the value of using major allergens on allergen extracts. But it is useful to look at where this fits in the context of major allergens. And also perhaps gives us some insight into what the important allergens are in Bla g 1. If you re-analyze the data that Dr. Nelson presented in 2004, looking at some common standardized extracts in the United States, and you recalculate the data as major allergen content per 10,000 BAUs, what you find is that the range in these extracts: cat, Timothy looking at Phl p 5, fescue looking at Fes p 5, D. farinae and D. pteronyssinus looking at the group 1 allergens, you find a range of between 15 and 172 micrograms per 10,000 BAUs of extract. When we look at the data that we have from the roach extracts, first of all we can't really compare the Bla g 1 data to these data because Bla g 1 is in arbitrary units. It's not in micrograms. But looking at the two allergens that we've measured in these extracts using the same ratio, micrograms per 10,000 BAUs, we see that the Bla g 2 content is right dead set in the middle of this range from Dr. Nelson's work. So this is actually I think a little bit more of an argument that Bla g 2 may in fact be a good surrogate for the skin testing. What we also see here is pretty good evidence that Bla g 5 will probably not be, at least in most study subjects. But again, I have to warn you that our aggregate data are not necessarily going to guide us. If, for instance, we find that for three quarters of the study subjects Bla g 2 is the major determinant, but for another quarter Bla g 1 is, we will have to account for the content of both of those in our standardized extracts when that happens.

So the next steps. We do in fact have sera from all of these study subjects. We will be determining their IgE specificities using Westerns, ELISAs and ELISA inhibition. Based on both the aggregate data and the individual data, we need to determine the appropriate surrogate test for standardization. And finally we will need to determine an appropriate set of reference standards for this.

And I'm happy to take questions about German roach. Yes, Dr. Nelson.

DR. NELSON: It used to be that the skin tests were done in triplicate. Were these done in triplicate?

DR. SLATER: Actually they were not done in triplicate. We were faced with that decision, but -- and in fact, when they -- occasionally the skin test would be repeated. About five percent of the time, a skin test at an individual dose had to be repeated. When that happened we actually analyzed the data completely separately for first skin test doses and second skin test doses. We found that there was no real difference between repeated doses, at least in those that we could look at. But we did not have it worked into our protocol to do it in triplicate. Our problem was that we really needed to get data on at least three extracts in order to try and do the analyses that we were hoping to do. And I suspect had we done it in triplicate, we probably would have gotten -- been able to use more of the study subjects. Fortunately, the data that we got was significant. Dr. Portnoy.

DR. PORTNOY: Yes, two questions. First of all, what happened if you ran out of space on the back? I mean, did you ever do so many that there wasn't enough room, and how far apart did these have to be so that one test did not interfere with another? And the last question I have is were there difficulties in interpreting the results based on the degree of pigmentation in different people's skin making it difficult to actually see the results?

DR. SLATER: Three very good questions. No, we never ran out of space on anyone's back. But think about the reason for that. Basically, you started a dilution of 3^-17. At maximum you get to 3^-5. We never went above 3^-5 on anybody. So if you get no reaction, which is the subject in which you're going to get the maximum number of injections, you go from 17 to 14 to 11 to 8 and to 5 and that's it. And there's certainly more than enough room to do that on any adult's back.

The distance between each of the injections had to be 2.5 to 3 centimeters between each of them. And we did not run into trouble with overlapping. I think it's an indication of the low potency of these extracts that we didn't run into trouble. I think had we done this test with grasses we probably would have had to space them more widely.

Pigmentation was actually -- it was an exclusion criterion. If either from very deep pigmentation or from other skin conditions we felt that we would not be able to read erythema, that patient would be excluded. The fact is, you can read erythema on people even with fairly deep pigmentation. It really can be done. So I'm not aware that we actually excluded anybody for that reason.

DR. PORTNOY: And one additional question. Were there differences in skin reactivity with different races? Or different degrees of pigmentation?

DR. SLATER: We haven't analyzed that yet. I thought you were going to ask if there were differences among the four centers geographically. And we haven't answered that yet either. We haven't looked at that. We may not be able to. We may not get -- you know, with only 15 study subjects at each site, we may actually not be able to learn that yet.

DR. NELSON: You always start at the top and work down. Has it been looked to be sure there's not a gradient over the back and erythema?

DR. SLATER: We actually didn't always start at the top. In the example that I showed, just for simplicity of showing it, we started at the top. The actual protocol called for sort of going back and forth starting. We always started on one end and worked to the other. But we didn't always start on the top. That was randomized. Dr. Granady?

DR. GRANADY: Can you comment on using the erythema versus the wheal size?

DR. SLATER: I can. It's for statistical reasons. The line that you get -- and you can think this through sort of intuitively. The line that you tend to get with the wheal size tends to be much flatter. And therefore, smaller errors in measurement lead to greater errors in the calculation of the D50. So when these studies were initially done I understand. The original papers on this were from the 1970s. When this was originally done, they did both with sigma E and sigma W for wheal size. And you can get valid data both ways. But the error of the measurement is greater with the wheal than it is with the erythema. Dr. Weiss?

DR. WEISS: Do you recall back 20 or 30 years ago when Paul Turkeltaub sort of designed this protocol for the measurements, did he have some of these unusual sort of curves that you showed where it goes up and down?

DR. SLATER: Oh yes. But when these original studies were done, those study subjects were excluded out of hand because they didn't actually meet the -- we were trying to look for a way to include as many subjects as we could. Those individuals, for instance Dr. Eggleston had been part of an earlier study in the 1990s, an ideal study. And one of the things that he said was they ended up excluding more than half of the subjects after they were skin tested because of just this. And we were looking for ways, statistically valid ways that we could include as many of the subjects as possible.

Remember those rules that I gave you before, that it had to be uniformly increasing skin test responses, two on either side of 50 millimeters. Many of those subjects would have just been excluded out of hand. And using the regression analysis was not very satisfying either, because the regression curves that you would get, while they would be valid best fit linear regression curves, didn't seem to make much sense in terms of what was going on. You're absolutely right. Mr. Hauck.

MR. HAUCK: Jay, did you analyze the puncture testing results? Were there any subjects that you included that were skin test, puncture test positive to one extract and negative to the other two? Did they, all three, puncture test positive?

DR. SLATER: We did have a preconceived notion from our prior testing that extract B was going to be the most potent, and extract A was going to be the least potent, simply from our in vitro testing. So we picked extract C, the intermediate potency one, as our screening test. We did not screen for all three. We only screened for one of them. I was afraid that if we screened for the most potent one, we would have a lot of flat lines for the other two. So we picked the intermediate one.

We also went back and analyzed at great length how we could have done the puncture testing better so that we wouldn't get hyporesponsive study subjects. And we had several of them who just had no responses at all. And we went back, we looked over their puncture tests. There was no way we could have picked these people out. It was almost as though the screen test was unrelated to how they ultimately did on the IV testing. It was very surprising.

DR. MACDONALD: Jay, in general, what is the incidence of a puncture positive/skin test negative in other extracts? Do you know?

DR. SLATER: I don't know the answer. Dr. Portnoy?

DR. PORTNOY: Did you ever in any of these patients who had the strange up and down pattern, did you ever repeat that to see how reproducible the curves were in individual patients? I mean, are these reproducible or are they just different each time they happen?

DR. SLATER: That's a great idea. I don't think anybody in the study even conceived of the notion of asking somebody to come back and do this again. But it would be a very good thing to do. We just didn't do it.

DR. PORTNOY: And any idea why there were differences in the potencies of these extracts? Is it the way they were manufactured, or do you have any clue about specific processes that are used to produce extracts that affect the potency?

DR. SLATER: Well, extract B was started with a different source material. And so that's one thing that we are looking into. But aside from that, I don't really see much in the way of clues. I do think that source material, you know it's an obvious statement, but I think source material is very important. And I think that it's one of the things that we need to look at as we move along with this. Dr. Berger?

DR. BERGER: So, in keeping with Jay's last question and your answer, do you think that the presence of glycerol in extract B made a difference? Do you have any way to look at that?

DR. SLATER: So you picked up on that? Yes, what Dr. Berger's referring to is that in extract B the glycerol was actually in the extracting solution, whereas in extracts A and C the glycerol was added after the extraction was completed. It may in fact be a real difference for this particular extract. I don't know that it's been shown to make a difference for other extracts.

DR. BERGER: Another question, which you could determine with the same kind of assays, which may have implications for the use of these subsequently in immunotherapy. Have you done any studies of concentrating the extract, say by ultra-filtration or lyopholization, or something like that?

DR. SLATER: We have not. We have not done that.

DR. BERGER: So it would be very interesting to take a concentrated extract tenfold.

DR. SLATER: Yes.

DR. BERGER: And then see if you got a tenfold higher reading from your method. That would also be a --

DR. SLATER: Yes. It would be very interesting. You should remember, though, that cockroach is one of the extracts that is fairly protease-rich. And so it's certainly possible that as we concentrate it, we're going to cause more degradation to occur. I don't know how to handle that.

DR. BERGER: If you would come to the microphone and identify yourself, please.