In the 72nd meeting of the Oncologic Drugs Advisory Committee, Tuesday, Sept. 24, 2002, ODAC members recommended that AstraZeneca’s Iressa (gefitinib) receive accelerated approval.
The following meeting transcript, slide decks, administrative documents, and more are collected from the publicly-available FDA records and are made available on the Cancer History Project as part of a series marking the 20th anniversary of the publication of papers on the role of the EGFR mutation in lung cancer.
PARTICIPANTS
Donna Przepiorka, M.D., Ph.D., Chair
Karen M. Templeton-Somers, Ph.D., Executive Secretary Members
Douglas W. Blayney, M.D.
Otis W. Brawley, M.D.
John T. Carpenter, Jr., M.D.
Bruce D. Cheson, M.D.
Stephen L. George, Ph.D.
David P. Kelsen, M.D.
Silvana Martino, D.O.
Jody L. Pelusi, F.M.P., Ph.D.,
Consumer Representative
Gregory H. Reaman, M.D.
Bruce G. Redman, D.O.
Sarah A. Taylor, M.D.
Consultants (voting)
Thomas R. Fleming, Ph.D.
Claudette G. Varricchio, DSN, RN, FAAN
Patient Representative
Thomas G. Simon
Industry Representative
George H. Ohye
The full transcript of this meeting is available for download below. What follows is an excerpt:
Committee Discussion and Vote
DR. PRZEPIORKA: I think we are going to have actually more extensive discussion, but first I want to hear from Dr. Temple and Dr. Pazdur about anything else.
[No response]
Well, we have a number of issues that are set before us and the first question from the FDA is the FDA believes the relevance of the sum improvement data discussed above cannot be adequately evaluated with a randomized, blinded study with an adequate control arm. The two doses of ZD1839 show no difference in efficacy and are thus not adequate. Do you agree? Dr. Varricchio, do you have any opinions regarding this question in particular?
DR. VARRICCHIO: I think that it is a soft claim of efficacy of symptom management in this situation. Obviously something is going on but you don’t know why and what might have caused the improvement in symptom management. So, I feel to really have a strong basis for saying that this drug is the cause of the improvement in symptom management you do need a randomized, controlled trial with a real no drug arm, not this drug arm. Whether it is placebo or best standard care, that can be discussed.
DR. PRZEPIORKA: Mr. Ohye?
MR. OHYE: I would like to make first a general comment, if I may. I would like to compliment the sponsor for carrying out these trials because I know they are difficult, and also embarking on the expanded access. I think that is very humanitarian and the committee should go on record to say well done.
I also think the FDA has done a good job in finding every blemish, every freckle, every micro-pimple in the trial, and that is their job because they have to look at this with an abundance of caution.
My general comment is I think when we look at all these questions we have to look at the fifth question that was proposed, accelerated approval, and we have to think in terms of the requirements for accelerated approval that were so adequately or so well explained to us on page 3, third slide by Dr. Williams. We have to look at the standards there and we have to bear in mind that the evidence that we have to have is do the data show that it is reasonably likely to show a clinical benefit, not that it has to be proven beyond reasonable doubt. I think it goes without saying that we have met some of the other criteria that were dealing with a serious, life-threatening disease and that the question is whether it is reasonably likely to predict a clinical benefit. I think the data are very much in favor in that area. Thank you.
DR. PRZEPIORKA: Dr. Redman?
DR. REDMAN: I guess I will ask the biostatisticians, can you state with the information that is present that it did not have a negative impact, a negative aspect? Is that no?
DR. FLEMING: Well, there are many issues here that we are going to be discussing and, as has already been pointed out, there are serious limitations in interpreting these data in the absence of randomization. The data establish some level of plausibility that there could be a quality of life impact and, to the extent that it establishes that, it makes it more implausible that it is a negative effect. But, in fact, the risk for biases here, as we will articulate shortly, are so substantial that one would have to be very cautious about what you interpreted when you are saying is there adequate evidence of causal positive influence of intervention on these quality of life measures.
DR. PRZEPIORKA: Dr. George?
DR. GEORGE: Just a comment, I would like to come back to something that Grant mentioned in his slides that I think is very important and we can’t lose track of. That is, the evidence for accelerated approval is not different than for ordinary approval. That is, we have to look for substantial evidence, and that is what I keep coming back to. Because of these potentials for serious biases in these studies I have a real problem coming to the conclusion that it is substantial evidence. My feeling is that obviously something is really going on. It is just too bad that we didn’t have the direct randomized comparison so we can really know what the magnitude of that is. So, I just thought we ought to keep that in mind. I thought Grant had a very good point about that.
The evidence for accelerated approval is not different than for ordinary approval. That is, we have to look for substantial evidence, and that is what I keep coming back to.
– Stephen George
DR. PRZEPIORKA: Dr. Varricchio?
DR. VARRICCHIO: Just a thought that has occurred to me listening to this, if this drug did not have such a user-friendly toxicity profile we might not be considering some of the things we are considering. This is kind of low risk in terms of the toxicity profile, and maybe we are creeping out further out on the limb than we might if the drug had a more serious toxicity profile.
DR. PRZEPIORKA: Dr. Blayney?
DR. BLAYNEY: I would like to say two things. One, if this drug was more toxic, like in the ECOG study, it ought to be reflected in a decrement in the symptoms score. So, there is some reassurance, in my mind, that the lack of toxicity is reflected in the improvements here.
Like Dr. George, I am convinced that there is a signal here. We have wrestled in the past with very similar sorts of problems. It looks like there is a signal. A lot of people tell me that there is a signal, and I think this drug ought to be available. I am wrestling, and perhaps I shouldn’t—that is your problem, wrestling with the regulatory and statutory burdens placed by the legislative branch on a government agency. So, that is where I have come down.
DR. PRZEPIORKA: Dr. Temple?
DR. TEMPLE: This question is the main one in which you come to grips with the plausibility of the symptom benefit. So I have to ask a question just in response to something that went before. There isn’t any doubt that there is an effect. We don’t have any doubt that there is an effect on a surrogate, that surrogate being tumor shrinkage. Nobody has challenged that; we don’t disagree with the numbers. The whole question is whether that is reasonably likely to predict clinical benefit.
We have been offered by the company the option to believe that the evidence they have of symptomatic improvement is at least a basis for thinking it might correspond. So I have to ask my question. We are very accustomed to being suspicious of unblinded quality of life assessments as a general matter. One of the points that was made, although I must say not as strongly as I would have made it if I were the company, is that much of the symptom benefit is apparent before anybody knows whether there has been a tumor response. That is comparatively unusual, and we would be very interested in knowing whether that affects anybody’s thinking about this or not because we are going to have to come to grips with it.
DR. PRZEPIORKA: If I could just paraphrase what you just stated, what we need to actually ask ourselves is, looking at the symptom improvement data in isolation, there may not be enough there to say this is really real. But in terms of looking at the symptom improvement data together with the response data, does that make a more plausible story?
DR. TEMPLE: Yes, and just one more point. Usually in refractory disease we have a history of accepting tumor response data as reasonably likely to correspond. The problem here is that in the first-line therapy where that question was tested–nothing; no hint; no nothing. So, that makes it a somewhat unusual problem and the potential that you might believe a little bit in the symptom data. The question is, and Tom has been raising this, how should one believe in that. I was particularly interested in one feature of it which is that the symptom response appears to be reported before anybody knows there has been a tumor response and I wondered if that made people think one way or the other about it.
DR. PRZEPIORKA: Dr. Reaman?
DR. REAMAN: That certainly was very important for me, but equally important was the fact that there was a cessation of collecting symptom response data when patients progressed. So I don’t know how you balance out those methodologic shortcomings.
DR. PAZDUR: Let me just follow-up to Bob’s question. Many of the drugs that we have gotten in for accelerated approval on a single-arm basis do not even have any attempt to look at quality of life data or symptom control. How would the committee view this in a refractory disease setting if we were simply dealing with a 10 percent response rate here, with the given confidence intervals, and that is all we had, period? Is that reasonably likely to predict clinical benefit if we didn’t have the first-line data?
We have approved drugs, for example CPT11, as Grant pointed out, with a very similar response rate. I am not going to cut hairs here between a 10 percent response rate and a 12 percent response rate or a 15 percent response rate. Obviously, we felt comfortable enough giving accelerated approval to that drug. We didn’t have the first-line data of CPT11 at the time of approval but, nevertheless, we felt comfortable enough to go ahead with that approval.
We have approved drugs, for example CPT11, as Grant pointed out, with a very similar response rate. I am not going to cut hairs here between a 10 percent response rate and a 12 percent response rate or a 15 percent response rate. Obviously, we felt comfortable enough giving accelerated approval to that drug. We didn’t have the first-line data of CPT11 at the time of approval but, nevertheless, we felt comfortable enough to go ahead with that approval.
– Richard Pazdur
So, pretend we didn’t even have the symptom benefit analysis, would a 10 percent response rate be predictive, or reasonably likely to predict clinical benefit? And, then a question for you that we frequently get from sponsors is how low can you go?
[Laughter]
DR. PRZEPIORKA: Is that question number two, if I read question number two correctly? Let’s finish one first and immediately go there. If there are no other questions, let’s go ahead and take the vote. The question is, to rephrase this, the symptom improvement data discussed above cannot be adequately evaluated without a randomized trial. Do you agree? We will start with Dr. Martino. Please state your name before giving your vote.
DR. MARTINO: Silvana Martino. Yes, I do agree.
DR. PAZDUR: Could you repeat that?
DR. PRZEPIORKA: The data cannot be evaluated adequately without a randomized trial. Do you agree?
DR. MARTINO: Yes, I agree that it cannot be evaluated. I understood you correctly. Thank you.
DR. BLAYNEY: Doug Blayney. No, I disagree.
DR. VARRICCHIO: Claudette Varricchio. I agree.
DR. BRAWLEY: Otis Brawley. I disagree.
DR. PELUSI: Jody Pelusi. I disagree but I would put a little caveat in here. It would be really helpful to see those patients who then went off study, to really show us their quality of life data as well to help us balance that.
DR. REAMAN: Gregory Reaman. I agree.
DR. PRZEPIORKA: Donna Przepiorka. I agree.
DR. FLEMING: Tom Fleming. I agree but I would like to actually expand a little bit on the rationale. I think interpreting symptomatic response data in this type of a setting is treacherous. It has long been recognized to be treacherous. The FDA has given a very clear articulation of all of the reasons.
I would like to expand briefly on some of them. It is an open-label trial. Placebo effects clearly exist. The fact that there are some immediate improvements may well be due in part to therapy; undoubtedly are due in part to a placebo effect and not necessarily a true treatment effect. We don’t know what the contributions are by ancillary care. I am bothered by the fact that we are not really doing an intention-to-treat analysis. We haven’t truly looked at what is the symptomatic response across the board in all patients. Why is it not relevant to know what the symptom response is when someone progresses? Why do we stop assessing it at time to progression? To really understand this we should be fully assessing this.
Then we get into these analyses about trying to provide credibility of response by looking at correlation with symptoms and vice versa and with survival. We have been down these pathways for 25 years in oncology research, recognizing how fallacious this reasoning may be. Response may simply be a marker for intrinsically better patients who would have had a better symptomatic outcome even without treatment.
Essentially, when we look at a marker, let’s say it is tumor response and a clinical endpoint, as I believe symptom improvement would be if we showed it, looking at whether there is a correlation is a necessary condition but not a sufficient condition for validity as a surrogate, meaning that if there isn’t a correlation then one really does have a difficulty in being able to interpret whether or not, in fact, one basically would say it can’t be a valid surrogate. But if there is a correlation between response and symptomatic improvement it just gets you in the door. It just means that there is now at least a possibility and, in fact, maybe only a limited possibility that it truly is representing treatment effect.
Essentially what has long been recognized is when you have a class of agents and you are looking at validating a surrogate, you really need to have studies that look at the treatment effect on the marker and the treatment effect on the clinical endpoint to really understand, for that class of agents, if the effect on the marker is reliably telling you about the effect on the clinical endpoint.
As we will say shortly, the sponsor did a remarkable job on the INTACT trials. I am very confused if they didn’t have an intention of using those trials as the basis for establishing clinical benefit why the reliance in accelerated approval was just basically on the 39 trial. Why didn’t we have a control arm, in essence, to be able to validate that there truly is a treatment effect on the symptom response?
So, the bottom line is these data may well provide clues and encouragement for doing a properly controlled trial, certainly though, they can’t be interpreted as providing adequate evidence to establish that treatment is actually influencing symptom relief, or even, the words the FDA has used, providing substantial evidence for such.
DR. REDMAN: Scientifically, I think the answer to the question is yes, but also with the caveat based on the toxicity profile that the drug surely didn’t have a negative impact on quality of life.
DR. KELSEN: David Kelsen. I agree.
DR. CARPENTER: John Carpenter. I think I disagree but that is only a partial disagreement. I don’t think these numbers are really adequate but I am very influenced by the degree of improvement in symptoms without any knowledge of the response data.
DR. CHESON: I am Bruce Cheson. I still have this feeling that when patients are off their chemotherapy they start feeling better and some of them just don’t start feeling better within a month but they feel better over a longer period of time. With this study design, I think that the number of patients influenced by that is not discernible given the information we have. So I agree.
DR. TAYLOR: Sarah Taylor. I agree.
MR. SIMON: Tom Simon. I disagree.
DR. PRZEPIORKA: Of the 14 votes, 9 agreed and 5 disagreed.
I think Dr. Pazdur has already adequately stated the second question for us. Given the lack of clinical benefit in 2 large studies of ZD1839 in combination with standard first-line chemotherapy, is study 39 response rate of 10 percent in 139 patients with resistant or refractory non-small cell lung cancer reasonably likely to predict ZD1839 clinical benefit? Dr. Carpenter, do you have comments?
DR. CARPENTER: John Carpenter. I am not too influenced by the Phase III results here because you are giving an anti-proliferative drug with 2 drugs that depend on cell proliferation for their efficacy. It is very much like giving hormonal therapy and chemotherapy for breast cancer where the response rate doesn’t go up, and if you give them together they will change the outcomes. I think the model may be fairly true here so I would negate those 2 Phase III studies and say that what they tell us is that we didn’t know how to study this drug, and I think everybody will agree that the most robust evidence all afternoon is that when you give this along with chemotherapy people don’t do any better. There was no serious discussion of that.
I think if we are going to undertake this right, we are going to have to look at the early studies just based on their own as a single agent. It seems to me that it is clear there is some prediction. What we don’t get, because of the inadequacy of the trial data we have here, is how strong that signal is. But it is very clear that some people are getting better and some people are getting clinical benefit and at least nobody, so far, has picked out a way to identify those.
I would submit to the company that if you look at who gets better on this drug, besides doing all the fancy gene analysis, if you took a simple immunoperoxidase stain and picked out slow-growing tumors everybody who gets better on this drug has a long and natural history in a tumor that grows slowly, and if you hypothesis were that people in the lowest quartile do better on this drug, you might figure out who gets better on this drug pretty quickly, and without doing too many elegant genetic studies. So, I think there is a signal. I think that 10 percent is hard as far as the response rate, and I think it is reasonably likely that that predicts clinical benefit. Given the inadequacy of the data we have, I would say yes.
DR. PRZEPIORKA: Dr. George?
DR. GEORGE: Maybe I can delay some of my comments until item 4 when I guess we are going to discuss additional trial designs because that may provide some information about what we think about this first-line trial as providing evidence in this case. But I want to point out that it is certainly possible that an agent such as this, with a low toxicity profile could, indeed, provide clinical benefit as measured by symptom improvement or quality of life and have absolutely no effect on survival. If that were true then, of course, that is a clinical benefit. It is not survival benefit but I think sometimes the question is phrased as if we always mean survival and it is clear in the regulations and the way things are stated that it is not just overall survival. In fact, in this trial I asked the question earlier whether we had any information about quality of life or symptom improvement even in that first-line therapy and we have no evidence of that whatsoever at this point. So, it is clear that there is no survival benefit so it is certainly possible.
Now, the question of whether it is reasonably likely to predict such things, then we run into an area of metaphysics and I can’t give a probability of what reasonably likely is, but it is clearly possible that it could.
DR. PRZEPIORKA: I will just weigh in with my two cents here. I don’t know of anyone with non-small cell lung carcinoma whose cancer went away by itself or developed a PR by itself. So, very clearly there is activity here, and very clearly 10 percent is substantial as a third-line agent. I don’t know that you could say for all non-small cell lung cancer. I think that is where some difficulty may arise, but very clearly there are patients who have derived clinical benefit from treatment with IRESSA. Dr. Cheson?
DR. CHESON: The response troubles me I guess. You can take the hematology out of the doctor but not the doctor–you know, we deal with tumors that respond so it is a bit problematic. But just looking as someone who has done clinical research for a number of years, you would be hard-pressed to find a study in which the Phase III data, particularly response rates, were better than the Phase II data. So, this 10 percent, with confidence intervals going down to 5.6, may actually be optimistic in the grand scheme of things. So, that gives me significant room for pause about substantial clinical benefit. I thoroughly agree that some people really have had astounding benefit but what is that going to be when the drug is used to treat tens of thousands of patients?
DR. PRZEPIORKA: Dr. Kelsen?
DR. KELSEN: I agree with Dr. Pazdur’s original comments, or I will try to answer his original comments. A 10 percent activity in third-line therapy in a variety of cell tumors, and colorectal is a really good example, is meaningful. Whether it is a surrogate for a higher level of benefit is difficult to say today, but it is a surrogate for activity and this drug does have activity, and the risk/benefit profile for this agent is substantially better than the risk/benefit profile for irinotecan and oxaliplatin.
DR. PRZEPIORKA: Dr. Fleming?
DR. FLEMING: Certainly there is evidence here that there is biologic activity. The question is how impressive is that evidence and how reliable is it to indicate true clinical benefit? We have two studies. We have the 16 study and we have the 39 trial, and the 16 study—just going back to what the team had intended, they were trying to discern between a 5 ruling out a 5 as an inadequate response rate, against a 20 percent response rate. So, according to their statistical criteria they needed to see a 13.3 percent response rate for success. They had less than that in the trial looking in the non-Japanese patients.
As we had discussed earlier, if you focus on those people that, in fact, truly were first- and second-line progressors there was only 1 responder. So, there is very limited information in that 16 trial about whether there is really an interesting or impressive response rate in third-line. We are left with the data in the 39 trial and essentially we are looking at the basis of that as 22 responses with, as Bruce was pointing out, confidence intervals that go down to 5 percent. In fact, if you follow the protocol it is 97.5 percent confidence intervals that go below that.
Then, as the FDA has pointed out, there are a lot of favorable characteristics in these people who responded. So, how impressive in the context of this trial are these responses? It appears, as has been presented to us, that the strategy that was in place here several months ago, prior to the release of the INTACT trials, was that the INTACT trials were going to give us the truth. Essentially, they were marvelous studies, absolutely fabulous studies conducted by the sponsor randomizing a thousand people per trial, getting 750 deaths per trial giving, in pair-wise comparisons, 500 deaths.
Essentially, these types of studies are incredibly informative and reliable in their assessments allowing us to detect, with reliability, even a 25 percent reduction or what would correspond to an improvement in survival from 10 months to 13.3 months.
What do the data show? The data show estimates that are slightly unfavorable, about a 10 percent increase in death rate in the 14 trial and no difference in death rate in the 17 trial. If you put this data together from these two studies, what you get is a remarkable level of precision. You have a thousand deaths per pairwise comparison to assess.
So this is not just a matter of, we didn’t achieve statistical significance. This is a matter of these studies nailed with great precision exactly what the true effect is going to be in first-line. Essentially, what that true effect is is a loss of one week in overall survival on a ten-month median for the control arm, a loss of one week at 500, a loss of two weeks at 250, with a precision two standard errors of within five weeks.
What that means is these agents could be as unfavorable as reducing survival by six weeks and as favorable, best-case scenario, three-week improvement which is one-quarter the level of benefit that would correspond to what the study had sensitivity to detect.
So I am left with a philosophical debate. These studies clearly conclusively establish providing consistent and compelling evidence that there is no effect on survival, not just that we have failed to achieve significance. There is no effect in the global population on survival. Response rates and TTP were also negative.
What is the relevance? Clearly, it is always true that if you are really interested in survival effect in third-line that first-line results could give either false-positive or false-negative conclusions, false-positive being maybe you have an effect earlier on on survival if you deliver IRESSA earlier in time. If, in fact, the studies had been positive, would the sponsor be asking us to step back and reexamine their relevance to third line because you had agreed they would be relevant up-front.
Yes; in fact, they could also be false negatives and that is the question that is validly being asked now. But hindsight is twenty-twenty. To say, in retrospect, gee, we can explain now why these results are providing compelling evidence of no-survival effect, aren’t relevant to the third line, I would argue we should be incredibly cautious about that interpretation. That is clearly a data-driven interpretation.
Yes; it may be true, but the bottom line is there clearly is relevance to those data and those data have to be factored in significantly. With that being factored in, it provides a great influence, I would argue on whether response rates in twenty-two responders in the uncontrolled 39 trial is compelling.
DR. PRZEPIORKA: Dr. Carpenter?
DR. CARPENTER: I think I would look at the Phase III slightly differently but only from an interpretation standpoint. They show, as well as I can imagine, that if you add these two drugs together, the way they did, that there is simply a vanishingly small chance that there is any real improvement. Would you agree with that?
DR. FLEMING: I agree.
DR. CARPENTER: I think the question is limited by the way we asked it and so I don’t think they tell us, necessarily, what the drug does in front-line therapy. They just tell us that the way we asked the question, or the way they asked the question, that we feel confident that, if you do it that way, there is no advantage to adding this drug. But I wouldn’t, for a minute, conclude that if you asked the question differently with a different type of study that you would, necessarily, get the same answer.
DR. FLEMING: You mean adding IRESSA first-line to a different combination; is that what you mean?
DR. CARPENTER: I am talking about separating the IRESSA from the chemotherapy. I think what they did was to combine them as if it were a third chemotherapy drug and, like other third chemotherapy drugs, it didn’t work and we feel pretty confident that it didn’t. What I am saying is that answers that we have are limited by the way the questions were asked.
DR. FLEMING: Just to clarify. Yes; I do agree that, in essence, we are limited to the interpretation of what these studies were designed to address. When the result is negative, to step back and try to assess why is very relevant. But to say they are not relevant to third-line is stepping back too far.
DR. CARPENTER: I may have not stated my question right. Excuse me.
DR. PRZEPIORKA: Dr. Brawley.
DR. BRAWLEY: One possible design of the trial would be, for example, to take individuals treated with first-line chemotherapy who respond who have a very high likelihood of relapse and then giving them IRESSA at that point to see if it sort of prevents relapse or prevents recurrence. That is a question that has not been addressed and it may—there is some scientific reason to believe it actually might work.
But they really have not—I mean, the drug’s clinical benefit, in my mind, can be very different from does the tumor respond. So I think I have given you what my answer to No. 2 is going to be.
DR. PRZEPIORKA: Dr. Pazdur.
DR. PAZDUR: I wanted to kind of go over this question before you vote on it because I think there are several things here that the committee has to understand. First of all, when we are asking this question, this is tantamount to should the drug be approved on accelerated approval. It is Question No. 5 of the sponsor. So you could just skip that Question No. 5. This is it.
We were rolling around on our merry way here in our Division with this 10 percent response rate and the symptom benefit. We believed that basically we weren’t going to be taking a look at the symptom-benefit work per our previous discussion but we did look at this 10 percent response rate.
Then we were kind of floored when the two large studies came into play. This is really what Tom was getting at here, that this is the point of the question. We are not asking about a 10 percent response rate. As I told Dave Kelsen, we approved a drug with a 12, 15 percent response rate. We already have that history of doing it. That is not the question here.
The question is in the context of these two other trials. If we didn’t have these trials, we probably wouldn’t even be here. We would have already approved the drug on our merry way. We have this data here. We can’t just ignore it. We have to take a look at the whole data package when we look at the approval of the drug.
The question here is not the 10 percent response rate. It is in the context of these two other trials that are front-line trials. The observation that this drug does not work with chemotherapy is an observation. It is not an explanation, and I have not heard from the sponsor a viable explanation of why these trials have failed.
If they would like to get up now and give it, I would like to hear it. George?
DR. BLACKLEDGE: Well, Dr. Pazdur, I can hypothesize as well as anyone else. It is very clear that whatever effects you are seeing with doublet chemotherapy, you cannot, it appears, add to. That appears to be the case, whether it is another chemotherapy agent or whether it is a novel agent of this kind.
I don’t have the explanation yet, and I don’t think anyone else does. All I can say is that it does seem to be an emerging pattern for both chemotherapy agents added as a triplet and also for novel agents added as a triplet.
Whilst I don’t think we can ignore the data, I do think that it looks an extraordinarily different situation from where we have clearly seen agents, noncytotoxic agents, giving real benefit as monotherapy in various different situations when they haven’t shown any additional benefit in combination.
So I am not sure I agree with Dr. Fleming completely. I think that, whilst you must take these data into effect, use as a monotherapy for clinical benefit and for response which leads to clinical benefit is a very, very different situation.
DR. PAZDUR: One of the problems that I see with that answer, George, when you take a look at this drug when it is favorable to your situation, you may look at it as a chemotherapy drug. When it is not favorable to your situation, you take a look at it as a special agent here which is somewhat perplexing to me.
I am fully aware of the doublet, triplet information in lung cancer. The question here, we have no other situation that I know in medical oncology, and I am having a tremendously difficult time trying to figure out why, in a first-line setting, you would not have some effect here.
Obviously, AstraZeneca has gone on an extensive development program not only in lung cancer but in a myriad of diseases with chemotherapy and this agent, I assume, based on some preclinical rationale, we got the results of this trial. It has thrown a tremendous monkey-wrench here. What is the explanation? That is the essence of this question and that is what needs to be discussed here, not the 10 percent response rate.
We have approved drugs. I don’t know how low is low and how low you can go, but this is the issue here of the 10 percent rate, albeit it might even go down to 5 percent. I don’t care. The question is these first-line data being there and that is the issue and that is the crux of why we brought this drug to the committee.
DR. FLEMING: Could I add just briefly to Dr. Pazdur’s question for the sponsor to respond to. If the sponsor didn’t anticipate that the two first-line studies would, in fact, provide the validation of the surrogate effects, justifying the accelerated approval, what was the strategy since, essentially, the accelerated-approval strategy indicates that postmarketing studies would usually be under way.
So, if you prospectively, before you saw the results of the first-line trial, already knew that those first-line results weren’t going to be relevant to efficacy in third-line, what third-line comparative randomized studies were already under way as the basis for validating this accelerated approval in third-line?
DR. BLACKLEDGE: We had no randomized third-line studies underway. The reason for that is that when we planned Study 39, we discussed extensively with our investigators about possibilities of randomization. They advised us that, certainly within the context of the United States, that would be extremely difficult to carry out.
In addition to that, the studies that we carried out in third-line were validation of the Phase I data that we saw where we unexpectedly saw responses and it seemed logical to carry out a Phase II in that setting. That is the basis that we went forward with accelerated approval. Now, we are clearly faced with a difficult situation. I don’t think any one of us expected the results that we saw in the INTACT studies although, in fact, if you look at other data that has emerged since we started them, maybe you might have expected that we wouldn’t see that effect.
I don’t believe it invalidates, however, the response that we see and the strong suggestive evidence, but not proven evidence, of a clinical benefit linked to those responses.
DR. FLEMING: Just in completing the response, then, to this response, your approach in the INTACT trials was remarkable. You did a remarkable effort to come forward with outstanding trials to establish whether or not there were effects on survival and other clinical endpoints in first line.
As a result, it seems to be a paradox that you have mounted the accelerated approval in third-line based on the 39 trial without any backup plan for how you were going to, in fact, be able to validate as accelerated approval requires, it surely leaves me to think you actually were anticipating a favorable result in INTACT that would serve as a basis for validating and, in which case, if we then took that logic to the limit, we would say, you did view that there would be relevance to what you see in third-line in the first-line trials to the third-line indication.
DR. BLACKLEDGE: We have never linked the third-line submission with the first-line submission. Clearly, if there was a positive result, we would have been very pleased with that and so would the patients. But this is not the only clinical-trial program that we are carrying out. We are carrying out trials with monotherapy in adjuvant situation.
We have maintenance studies going on and we would be more than happy, as we have described earlier, and as our investigators have described earlier, to attempt to validate the data that you have seen today from Trial 39 in a randomized setting.
DR. PRZEPIORKA: Dr. Carpenter?
DR. CARPENTER: John Carpenter. Since we seem to be pressing in on this question of should we approve this drug or not, I wanted to get a couple of comments in if it is okay. I think that we could all speculate on the reason that the third-line therapies don’t—I mean, the first-line therapies don’t validate the results seen as a third-line drug. But that may be because of the mechanisms of action of the drugs and they would be counterproliferative.
That is an hypothesis but it certainly is a testable one. It seems very clear that slowing indolent tumors are the ones that get better here.
If this drug—and I will be very favorable towards this drug with the limited evidence that we have, partially because there is no viable competitor in this situation. But I think there is a whole flood of studies that could be done to elicit out the way to use this drug. You could study performance-status-2 patients and test this versus anything since nothing else works very well.
I think there is a study of vinoralbine in older patients. That is certainly doable. You can use it in a short period before front-line chemotherapy with a crossover. You could use it as an adjuvant with a placebo control after frontline chemotherapy and sort this out. All these things are easily doable.
You can do the proliferation assay and see if you can predict who gets better on this. All those are easily doable studies. So I think I am going to come out in favor of making this available with the proviso that a bunch of studies about just how to use this drug need to be done and should be done.
DR. PRZEPIORKA: So, Dr. Pazdur, I think your question is not just is there any evidence for clinical benefit for this drug, but do the results of the two randomized trials actually suggest more or less that the results that we see in the single-arm study makes a difference.
I have to actually weigh in with Dr. Carpenter. I think it is very clear there is clinical benefit in the single-arm study, but I think the questions being asked in the randomized studies are completely different questions. Although we don’t know why, I am not sure we actually know that the inhibition of the kinase is actually the mechanism of action that this drug uses because there doesn’t seem to be any correlation with EGFR expression.
I think it is very clear there is clinical benefit in the single-arm study, but I think the questions being asked in the randomized studies are completely different questions. Although we don’t know why, I am not sure we actually know that the inhibition of the kinase is actually the mechanism of action that this drug uses because there doesn’t seem to be any correlation with EGFR expression.
– Donna Przepiorka
I don’t know that anybody right now could actually answer your question about why the combination does not work because I don’t think we have enough information available.
DR. TEMPLE: Maybe everybody said this and maybe it is clear to everybody, but we need to be quite sure. As Rick said, if all we had was the response rate in third-line, we might not even have brought it here because we have been approving drugs with reasonable response rates in refractory disease on the basis of that right along.
We were, however, surprised by a negative result in first-line which, as everybody has noted, was a complete surprise to everyone or they wouldn’t have done the trials. The fundamental question is does that shake your belief that a 10 percent response rate is reasonably likely to predict clinical benefit, just to be clear.
That is why we are asking you, because we would ordinarily have been comfortable with that response rate. Should we not be comfortable anymore because of those trials? That is really the question.
DR. PRZEPIORKA: With all due respect to the statisticians, I think what we have heard from the clinicians, both from the sponsor and on the committee discussing is we don’t think the results of the randomized trials are that clinically relevant to our opinion of the single-arm trial.
But I could take a vote and let you know for sure.
DR. PAZDUR: That’s appropriate.
DR. PRZEPIORKA: So the question is, given the lack of clinical benefit in the randomized trial, do you still think that the response rate of 10 percent using ZD1839 is likely to predict clinical benefit? Dr. Martino.
DR. MARTINO: Yes; I do. I am discouraged, but not totally devastated.
DR. BLAYNEY: Doug Blayney. Yes.
DR. VARRICCHIO: Claudette Varricchio. Yes.
DR. BRAWLEY: Otis Brawley. Yes.
DR. PELUSI: Jody Pelusi. Yes.
DR. REAMAN: Gregory Reaman. Yes.
DR. PRZEPIORKA: Donna Przepiorka. Yes.
DR. FLEMING: Thomas Fleming. No.
DR. REDMAN: Bruce Redman. Yes.
DR. KELSEN: Dave Kelsen. Yes.
DR. CARPENTER: John Carpenter. Yes.
DR. CHESON: Bruce Cheson. No.
DR. TAYLOR: Sarah Taylor. No.
MR. SIMON: Tom Simon. Yes.
DR. PRZEPIORKA: The vote is 11 yes and 3 no.
The third question regards the expanded-access protocol. We had some discussion regarding expanded access and single-patient exemptions and single-patient protocols. So, Dr. Temple, Dr. Pazdur, if you just want to set the stage regarding specifically what that means, especially for the new members, of the committee, that might be helpful.
DR. WILLIAMS: This question becomes a little more difficult after your answer to No. 2. We will have to decide what to do with your answer to No. 2. Theoretically, supposing the drug is not approved, there is the issue of expanded access. I think the company has mentioned that they would have to reexamine to determine whether they thought it was ethical.
We had two ODACs where we discussed expanded access and we talked about the level of activity that should be seen, et cetera. If this were to be of such a level of activity that we would not approve it for accelerated approval, the question would be would it still be indicated to allow widespread expanded access.
I guess we could go ahead and have that discussion as you wish.
DR. PRZEPIORKA: Comments? Dr. Kelsen?
DR. KELSEN: Actually, I was initially heartened and then a little bit taken aback by two contradictory comments from the sponsor. First, I thought I heard that it would be very straightforward to perform a confirmatory trial either before accelerated approval, while the expanded-access program was open, or after accelerated approval, that, in either case, it would be quick to achieve definitive opinions to clear the air because patients would readily approve randomized to a study in which they initially did not receive the drug and then, at sometime, they received the drug.
That is what actually swayed me quite a bit because I could see you doing that before accelerated approval with the expanded-access program wide open.
Then I was a little taken aback by the answer that we got that the reason that we didn’t see Phase III data right up front today, which would have clearly answered the question, is that it was felt that patients wouldn’t accept such a randomization.
Maybe it would help if the sponsor could clarify their position.
DR. BLACKLEDGE: What I was referring to, Dr. Kelsen, was the situation two years ago when we began our trial program and the expanded-access program. We have been in regular discussions with both the FDA, with NORD, with medical ethicists and patient advocates to review the expanded-access program as we move along.
We are certainly not going to make any instant decisions and we would want to discuss the situation extremely carefully with all those people, particularly the FDA. Of course, we would want to be in a situation where we were able to carry out a confirmatory study.
We didn’t speak to the patients about the original randomization. We spoke to investigators at that time. But now you are hearing from investigators who would take part in those studies that they feel, in view of the changed environment and changed need, that it would be possible to carry that out.
Now, we want to work together with all the involved stakeholders to make sure that we come to a happy resolution of this. We began the expanded-access program in the anticipation that one day we would stop it. Clearly, we would like to do that as soon as possible because we would like to see the drug approved with accelerated approval.
We would also like to do the confirmatory study as quickly as possible and, therefore, we would like to achieve some kind of balance so that we could do the study and yet not deprive patients of the benefit that they are gaining.
DR. PRZEPIORKA: Just a comment. I recall, in our discussions, that the program should be set up so that patients who are eligible for any studies would not be allowed in the expanded-access program. I think that is a very valid way to continue if this program does stay open. But I think it is also very important, now that we have seen the response rates and that there may be certain subsets of patients in whom the response rate is highest and others in whom there is no response, and especially there is no valid reason to use it with chemotherapy, that that is something that absolutely has to be given to the patients prior to their making a decision to take this drug.
DR. PAZDUR: Donna, we have corrected the informed consents to reflect these recent INTACT trials.
I do want to mention one thing. There are 12,000 patients on expanded-access trials which is a huge number of questions. There are a huge number of patients and there are a huge number of questions that are unanswered about this drug.
I want to step back. This is something that we are trying to encourage sponsors to do is to start looking at other trials that could be done within the context of a more expanded research program with a particular drug. Even if they were not meant for registration or to fulfill a new indication, or even to just give us more idea about what would be a better use of this drug, specific populations, populations that this drug would not work in, rather than just giving everyone, 12,000 patients, this drug.
The best way to get information for everyone in society is for patients to be going on clinical trials. We would like to work with sponsors more to design other trials. For example, in the third-line setting, could we have used a point analysis of time to progression where patients could have gone on a chemotherapy regimen that their doctors agreed to versus this drug and cross over at the time of progression to IRESSA looking at a time-to-progression endpoint.
Everybody would get the drug. Some people would get it a little later than the others, but one would have access and get more information on this drug.
We really would like to emphasize the best way for us to get information is randomized trials and is to look at trials in a more close way. I think this would give us much more confidence in approving drugs in the long term.
DR. PRZEPIORKA: Any other comments on Question 3? Hearing none, let’s move on to No. 4. Regardless of whether ZD1839 is granted accelerated approval for treating non-small-cell lung carcinoma, additional trials may be needed. Please discuss any potential study designs that will demonstrated ZD1839 provides clinical benefit to non-small-cell lung cancer patients.
I think we had some initial ones. If anyone wants to underscore their favorite design or indication, now is the time to discuss it.
Dr. Blayney?
DR. BLAYNEY: Given what we have seen and our experience, I think it would be reasonable to ask the question, front-line in untreated patients, does this drug offer a benefit over–pick your chemotherapy. I would probably pick a good performance status, relatively good prognosis, patient population.
But I think, from my point of view, that is an ethical question to ask.
DR. PRZEPIORKA: Dr. Martino?
DR. MARTINO: I want to simply follow up on that. There is a design in front-line that has been mentioned around the table and that is actually looking at a new agent–this would be the new agent–in place of chemotherapy. I think that would allow us to, first of all, look at this agent up-front but, also, if it turns out to be as good as and, perhaps, better than chemotherapy, would be a nice alternative.
DR. PRZEPIORKA: Dr. George?
DR. GEORGE: Just a comment about the crossover designs that have been mentioned or the early and late IRESSA, say, in the third-line setting, there are a lot of problems with that. I just wanted to make sure everybody is aware of that, that it might be easy to get–it might be a practical issue of getting patients on the study but there is a real serious problem with–you, first of all, have to be very careful about the time to progression, making sure the blinding is working properly so you are not biased in deciding when somebody has progressed so that they can get IRESSA or something.
But just the whole issue of how soon you do that is going to cause you to be very limited in your kinds of conclusions. If that were the kind of follow-up study, I would be kind of skeptical of whether whatever came out of that would really prove or validate the clinical benefit of IRESSA.
So that just kind of bothers me. It is possible to do but it would need to be carefully thought out.
DR. PRZEPIORKA: Dr. Varricchio?
DR. VARRICCHIO: I would to suggest that, given the hints that have come from the trials that we have been hearing about, that any subsequent trial be designed so that it could look at stratifications in the analysis that might let you be able to begin to look at predicting which subset of people this drug would be most effective to be used in.
DR. PRZEPIORKA: Dr. Fleming?
DR. FLEMING: This is a difficult issue because whether accelerated approval is granted or not, the efficacy has not been established. The issue on the table here was not whether full approval should occur. It is whether accelerated approval should occur.
So whether or not FDA grants accelerated approval, it is recognized that efficacy has not been established and there needs to be a timely conduct of adequate and well-controlled studies to achieve that evidence.
Many of us have argued for a long time that accelerated approval has many advantages and disadvantages, advantages if there is adequate evidence to establish plausibility of benefit, it provides earlier access. However, at the risk of providing earlier access to interventions that haven’t been truly proven to have a favorable benefit-to-risk profile and also, at the significant risk of being able to do a timely assessment—i.e., if there is accelerated approval, we had already heard from the sponsor that they had reservations about mounting the Phase III comparative or Phase IV postmarketing comparator trials without accelerated approval.
Now, if there is accelerated approval with wide access, not access—an expanded access limited to those people who wouldn’t be eligible for a study but to anyone who would choose to get access, is it logical to think that we are going to be able to mount a truly informative randomized comparative study to reliably assess effects on clinical endpoints.
It could be argued, well, sure we will do a crossover study. As Dr. George has pointed out, they can be very problematic in interpretation, particularly if you offer crossover at a relatively early point in time. So let’s suppose, based on what I am hearing, that people are not so willing to take a failure in first-line and, in any way, argue that that should give us less confidence in third-line.
I think I would ask, then, under that logic if accelerated approval is granted, give me a third-line trial that is a randomized comparative study that will truly establish in a timely way efficacy on clinical endpoints. That is going to take several hundred patients properly randomized and followed for an adequate duration of time to be able to meaningfully establish the benefits at least on symptoms and, potentially, even on survival.
– Thomas Fleming
I think I would ask, then, under that logic if accelerated approval is granted, give me a third-line trial that is a randomized comparative study that will truly establish in a timely way efficacy on clinical endpoints. That is going to take several hundred patients properly randomized and followed for an adequate duration of time to be able to meaningfully establish the benefits at least on symptoms and, potentially, even on survival.
I have serious concerns as to how that is really, truly going to be something that can be done in a timely way as the regulations require if, in fact, accelerated approval is granted.
DR. PRZEPIORKA: Mr. Ohye?
MR. OHYE: When accelerated approval is granted, it is only granted in the United States. So there is Canada and other countries where this study can be carried out. I would also like to add that, before I retired, I was a competitor of AstraZeneca. As a competitor, I was very much impressed with the level of research that they are able to carry out because they have research centers in Sweden, which is a major scientific center, and they have research centers in the U.K. and in the United States.
So, with this worldwide capability and the brain power that they have available, I have every confidence that they would be able to come to an agreement with the agency on the appropriate study and carry it out in a timely fashion.
DR. FLEMING: And you are confident that the results that would be done in other settings would be truly relevant? We have already seen the differences in ethnicity with the Japanese and U.S. That doesn’t cause you any concern?
MR. OHYE: I think the challenge has been set down before the company and, while I can’t speak for the company, I have competed against this company and I know that their capabilities are formidable.
DR. PAZDUR: Tom, the drug is approved in Japan, so doing the trial in Japan would not be an option. The EU community; we have accepted lung-cancer trials from Western Europe and even Eastern Europe. So we feel confident in that aspect of quality of data and comparable results to the U.S.
DR. PRZEPIORKA: Hearing no further comments, I will adjourn the meeting. Thank you.
