U.S.DEPARTMENT OF HEALTH AND HUMAN SERVICES FOOD AND DRUG ADMINISTRATION CENTER FOR DRUG EVALUATION AND RESEARCH + + + + + COMMITTEE FOR ADVANCED SCIENTIFIC EDUCATION SEMINAR SERIES + + + + + "The Use of Placebos in Clinical Trials and the Ethics of the Use of Placebos." + + + + + WEDNESDAY APRIL 21, 1999 + + + + + The Panel met in Conference Room D, Third Floor at 5600 Fishers Lane, Rockville, Maryland, at 1:30 p.m., Janet Woodcock, M.D., moderating. PANEL MEMBERS: DR. ROBERT DELAP, FDA DR. PETER LURIE, Consumer Advocate DR. JAY SIEGEL, FDA DR. PAUL STOLLEY, University of Maryland DR. ROBERT TEMPLE, FDA DR. SIDNEY WOLFE, Consumer Advocate C-O-N-T-E-N-T-S PAGE Call to Order and Opening Remarks Dr. Kobayashi 3 Introduction Dr. Woodcock 6 Dr. Wolfe 11 Dr. Lurie 16 Dr. Stolley 24 Dr. Temple 28 Migraine Case Study 46 Hypertension Case Study 74 Major Depression Case Study 93 P-R-O-C-E-E-D-I-N-G-S (1:36 p.m.) Call to Order and Opening Remarks DR. KOBAYASHI: Good afternoon. I'm Ken Kobayashi, speaking for the Committee for Advanced Scientific Education, or CASE. I would like to welcome everyone here for this very special CDER Seminar. I know that we have a number of people in the audience who have come a long way, from as far away as Connecticut, and I would like to say that we appreciate your interest in this program. Please note that because of the amount of material to be covered, today's seminar has been extended and will conclude at 3:30. If people come in late, and there is lack of seating, we do have an overflow room up in 13-B37. CASE rarely mounts a program of this scale, and several people were instrumental in putting it altogether. I would particularly like to thank Karen Zawalick, who came through with the ball on many major last-minute requests and needs, Devota Herbert, Joy Bennett and Sandy Coffin. Today's seminar will be transcribed, videotaped, and video-conferenced to four locations. I thank See Lam and Dave Kausal, our technical virtuosos, for their invaluable assistance with these and other matters. I cannot thank Robin Huff and John Senior of the CASE Committee for all of their help, both in preparing for this seminar as well as others, and last, but not least, I'd like to welcome and thank our Panel members for their participation. I usually make the announcements at the end of the question and answer period following the talk, but I'll make an exception today, given the full schedule and make a few announcements now. Next week's scientific rounds will be given by the Office of Generic Drugs and will discuss the issue of bioequivalence for inhalational drug products. And we have an exciting line-up of seminars to bring this academic year to a close. Next month we have Dr. Jane Henney, Commissioner of the Food and Drug Administration, and Dr. Patrick Weir of Smith/Kline Beecham, who will speak on predicting drug class teratogenicity. In June, we have Dr. William Evans of St. Jude's Children Research Center who is internationally known for his work in childhood leukemia, speaking on his work in developing a highly innovative and successful treatment program for this deadly disease. Dr. Ezekiel Emanuel, who heads the Department of Clinical Bioethics at the NIH, will conclude the year with a talk entitled "What is Community Consent and Consultation?" Dr. Janet Woodcock, who I think is the Director of the Center for Drug Evaluation and Research, who needs little introduction to this audience -- (Laughter.) -- will moderate today's discussion. We have a distinguished Panel assembled here today, Dr. Sidney Wolfe, who is Director of the Public Citizen and Health Research Group. Dr. Peter Lurie is a medical researcher also at the Health Research Group. Dr. Paul Stolley, who is Professor and Chairman of the Department of Epidemiology and Preventive Medicine at the University of Maryland School of Medicine. Dr. Robert Temple wears many hats here. He is Director of the Office of Drug Evaluation I and of the Office of Medical Policy. Dr. Robert DeLap is Director of the Office of Drug Evaluation V. We are pleased to have with us a member of our sister center in FDA, Dr. Jay Siegel, Director of the Office of Therapeutics, Research, and Review at the Center for Biologics, Evaluation, and Research. Today's program will consist of two short didactic presentations, followed by discussion of the three cases. Each case will feature comments by the Panel members and will include an opportunity for comments from the audience. After all three cases have been discussed, Dr. Woodcock will conclude and wrap up the discussion. And now, I will turn the meeting over to Dr. Woodcock. Introduction DR. JANET WOODCOCK: Thanks, Ken, and thank you for all your efforts. This has been a very successful seminar series this year and I think we are all proud of it. I would like to welcome the member of the FDA community, our invited guests, and outside people who have come into our seminar. Our topic for discussion today is "The Use of Placebos in Clinical Trials and the Ethics of the Use of Placebos." One of the most interesting, I think, and challenging parts of our work here at FDA, is the extent to which judgment is involved, and the way the different values and perspectives of different groups influence their views of the issues that we deal with. The use of the placebo is one of these issues. It's something that has been the topic of debate ever since their use was invented, and currently, this is a very hot topic of debate in various academic circles, as many of you may know. FDA oversees the conduct of many clinical trials in the United States, and we are frequently drawn into this controversy. Our standards and our guidances often drive the design of clinical trials in specific disease areas, including the use of placebo. Since the ethical situation evolves with changes in scientific understanding and with development of alternative therapies, we have to pay constant attention to the ethical implications of our standards that we have developed, and our guidance. The purpose of the seminar today is educational, because in our review and regulation of clinical investigations, the FDA staff plays an important public protection role in making difficult judgements. It's extremely important for us to hear the different perspectives, and to understand all the points of view. Only with an appreciation of the variety of views, can we effectively and credibly carry out our public health protection role. Given this, I am going to ask everyone here today, both the audience and the participants, to agree on some principles for our discussion. First and foremost, let's all acknowledge that we all here work in the public health field for the benefit of the public. That's what we do. And that we in fact are united in our goal of helping people, no matter what our differences are, about the proper path to take. Therefore, please grant each other the respect of their conviction and courtesy, and this includes some ground rules. Let's have no interruptions, and let's try to keep to times, and I will assist everyone in these endeavors. Now, what are these issues around the use of placebo? First of all, I think we have to remember, although it's not a topic for discussion today, that these issues are raised within a larger context of concern that has always been present about experimentation on human subjects. There is an inherent imbalance of power that's perceived between the knowledgeable investigator, and the less-informed volunteer, or the dependent patient who has an illness. And this has long been a matter of concern for many people. Recently, we have to acknowledge that the industrialization, so to speak, of clinical trials, the reimbursement of investigators and their ties to the pharmaceutical industry, have increased some people's levels of concern. So, this is a background, I think, to this discussion. The use of placebo controls is a separate issue, but it is linked in many people's minds to the overall issues of ethics and clinical trials. Now, the goal of drug development is to rapidly and efficiently characterize candidate products, so the beneficial ones can be made available to the public. And I think we all agree upon this. From a strictly scientific point of view, the most efficient and informative way to do this is by using placebo controls, given the limitations of our understanding of sources of variability in people. Nevertheless, it is widely agreed that use of placebo is not ethical in cases where people would risk serious or irreversible harm by taking them. In these cases, we must turn to the use of alternative trial designs to get the information we need. In many cases, though, there is a gray zone, as was discussed in the e-mail that was sent out to everyone, and there are a couple of gray areas in this. What is harm? How much harm is it ethical to inflict? I mean, anyone who has ever vaccinated a toddler knows that you're inflicting some harm upon that person, that toddler, when you're sticking a needle in their arm, but that's obviously a beneficial thing to do. Drawing blood from experimental subjects inflicts a certain amount of pain and harm on them. So, the definition of harm is one area that I've observed, where people have a disagreement about how much, quote "harm" is tolerable. And then there is another area; there is disagreement on the existing data. And this gets to the idea of equipoise in clinical investigations. There are many therapies where people have an idea that they work, but it is not clearly known whether they are effective or not. And often, there are great ethical debates about whether these should be studied. Sometimes, such therapies have been studied and it has been found that they are ineffective, or even harmful. And in retrospect, the study seemed a very credible study to carry out; however, the idea of how much is known about the long- term effect, or the beneficial effect of an intervention, impacts on how ethical it is to study it and to omit it from people's care. This seminar today is designed to explore this frontier of the ambiguous areas, or the areas where there isn't agreement of what is ethical and what is not ethical in clinical trials. So, without further ado, I am going to introduce Dr. Sidney Wolfe, who will start out with one perspective on this issue. Dr. Wolfe. DR. SIDNEY WOLFE: Thank you. I would like to thank Dr. Woodcock, one of the I'm sure many geneses of this meeting was a discussion that Dr. Lurie and Dr. Sasick, who is on our staff, and Dr. Woodcock and I had last year, about problems in the area of placebos, and she obviously took it very seriously, and we are very hopeful that this seminar will be instructive. I have already learned a lot by reading some older papers that I hadn't seen before, and I think that inevitably, the influence of the seminar will be to move things forward. I can't imagine that what will be learned here will not affect FDA policy, and as Dr. Woodcock alluded to, define maybe, or refine some of the guidances that exist. One way of describing the purpose of the seminar is to focus on how to reduce harm to patients in clinical studies by a case-by-case elimination of certain placebo trial designs, which inflict gratuitous violence, pain, suffering, or permanent damage including death to patients. By gratuitous, I mean at least one of the dictionary definitions which is: unwarranted by the circumstances; unnecessary. The FDA is just one of a large number of decision-makers who have the potential to say no to an unethical clinical trial. The list includes the researchers; obviously, the funders, be they the drug industry, or device industry, or the government; the Institutional Review Boards, or the non-Institutional Review Boards, the for-profit Review Boards that are overseeing some of the non-institutionally-based trials conducted by for-profit clinical trial companies; the FDA, as another partner, as I mentioned; and then medical journals. There are at least some medical journals that have made a statement that they will not publish an article whose clinical design is unethical in their view, although at least two of them that I know of, having said that, have published studies that are pretty clearly unethical. Weak links in this chain of decision- makers reverberate, and frequently, when a study that is at least later, if not earlier, deemed to be unethical is done, the people in the linkage will say, Well, it must be okay, because the IRB approved it; or, It must be okay because the government funded it; or, It must be okay, because the FDA allowed it to go forth. I would like to mention a few examples that sort of crept through this entire chain. I do not know whether all of them -- I know some of them -- did go through FDA. One, are these studies in which HIV- and PPD-positive people in Africa, after a time when it was clear that isoniazid, INH, would prevent the full- blown development of TB, were enrolled in a randomized placebo control clinical trial Another, which Dr. Lurie and I have focused on a lot in the last few years, which did at least in part go through FDA, were studies in which HIV-positive pregnant women were given placebos after a time when it was clear that AZT prevented it. And it was after a time when it was crystal clear that a shorter dose of therapy worked. Another example is SYST-EUR, a study done in Eastern Europe, in which, for a number of years after it was clear that isolated systolic hypertension was successfully treated in terms of clinical endpoints by diuretics and so forth, more people were enrolled into a placebo control trial. By definition, all of these studies cleared the funders, the researchers, the Institutional Review Boards, and in some cases, the FDA. And one question we have, which is not rhetorical, we will ultimately try and get answers to it is, How many times has the FDA refused to accept a placebo control trial because they believed, in those instances, that it was unethical? We must point out that we do believe that there are some placebo control trials that are okay, but I have no idea what the answer to the question is; we may be asking it at some point in the future. As mentioned before, there really isn't a specific, a drug-specific policy on what you do with respect to different therapeutic classes, as far as placebos are concerned. First and foremost, as I mentioned at the beginning, we need to protect patients. That is the primary concern in a clinical trial. Yes, immunization hurts, and that is a circumstance, assuming that the vaccine is effective, where you trade off some risk for some benefit, but a number of the cases that we have been concerned with are much more than just mild pain. I would just like to finish, and then turn it over to Dr. Lurie, by talking about the benefits and risks of the trial design itself. We are all familiar with looking at a drug in terms of its benefits and risks, but one of the problems with the benefits and risks of the trial design, is that the benefits may go to one party, and the risks may go to another, and there could be problems. Is there a difference between the group which benefits the most from a certain trial design, and the group which is at risk the most from the same kind of trial design? One example in terms of a trial design, is that in a number of instances, an equivalency study comparing a known effective treatment with another, will lower the risk for patients, as opposed to a placebo control trial, but it will increase the risk for the drug company. A drug company, if they can, would much rather conclude, as is usually the case, if not always for hypertensive agents and others, that their drug is better than a placebo, than risk the notion that their drug may not be as good as the standard existing therapy. So, that's a risk that the drug industry would take, which might, and would likely, benefit patients. I'm now going to just turn this over to Dr. Lurie for the middle part of our discussion, and then Dr. Stolley will make some concluding remarks, and then Bob will take over. DR. PETER LURIE: I'm going to start off by talking about what I hope are agreed-upon principles between both us and FDA, and what we are really going to be talking about I think is, not so much these principles, but the interpretation of them. So, let me go through the principles which would help one to decide whether the use of a placebo is acceptable. The first principle, or first question is, Do prior placebo control trials demonstrate effectiveness? And I think that we are dealing today with no examples of medications in which there is no drug proven effectiveness. So, as a general matter, the answer to the first question will be yes for all three examples today. If the answer is no, we would not object to the use of placebo, and I want to be very clear on that. We are really talking about situations where known effective therapy exists. Of course, what is effective is in the eye of the beholder, or the interpreter, and how convincing is that evidence, is the next question. And some evidence is more convincing than others; it comes from better or worse done studies, or case- controlled studies versus clinical trials, versus those with poor follow-up or good follow-up, and to the extent that the evidence is stronger, it makes the provision of placebos all the more unacceptable. What is the magnitude of the benefit that has been shown in prior studies, if the drug is minimally effective, to the point that one is not really that certain that it works at all? Well, to the extent that weighs in the direction of being able to provide a placebo, this is obviously tied into the first point. Now, how severe is the disease? If the disease is some trivial condition, unless one has one right now, like a cold, then we think those kinds of conditions can justifiably be studied with a placebo, but as the disease becomes more and more severe, as subjective symptoms become worse, or there is risk of irreversible morbidity or even mortality, it becomes more and more unacceptable to use a placebo. And importantly, Do alternative study designs exist? And it's incumbent, I believe, upon the researcher to really honestly go through all of the possible alternatives before concluding that the placebo is really the only way to go. There are arguably some conditions where the duration of the study is critical; studies where it might be acceptable to provide placebo for a short period of time, but not for a longer period of time. Will the most severe patients with a disease be enrolled? To the extent that the patients are sicker, I think that that argues against the use of placebo. And to the extent that there are good data on historical controls, that weighs in favor of not providing placebo. Is the placebo effect for the condition very variable? Conditions like mild pain, that are both not severe -- answering a previous question -- and which have a substantially variable placebo response, we think that providing placebos, even though obviously known effective medications exist, would be acceptable. Finally, is there equipoise over drug effectiveness in this population? And if there's reason to think that efficacy shown in another population may not extend to the present one, that may argue for a placebo control trial in the new population. Now, it is worth reviewing the legal status of the FDA's placebo practices, and some of this comes from regulation, and other parts from legislation. Active control trials are to be used, quote "When the condition treated is such that the administration of placebo, or no treatment" -- here is the critical language -- "would be contrary to the interests of the patient." And I suppose that is part of what we will be arguing about today. The Food, Drug, and Cosmetic Act does not, as is sometimes assumed, require placebo control trials. What is required is adequate and well- controlled investigations. And by regulation, the FDA has described these as: placebo control; dose- comparison studies; no treatment concurrent controls; active treatment concurrent controls; and even historical controls. So, there is a place for interpretation and a certain amount of flexibility. I think out there in the real world, the impression that people have is that the FDA is really insisting on the first of these, under almost all circumstances, and then people feel obligated to echo the FDA practice, something that Dr. Woodcock I think correctly pointed out is, in fact, what happens. Importantly, there is no formal FDA policy on placebo use at all. And what we hope will come out of this is the beginning of some kind of FDA policy, which is condition-specific, and I think that that's important. What we are glad about today's discussion is that it's not going to be, we hope, one of exchanging platitudes, but will rather be a discussion of condition-by-condition, because it's really condition-by-condition guidelines that are necessary. Now, this lack of FDA policy on placebo use has contributed to some confusion among FDA medical officers, and as some of you know, many of you were surveyed by us in the Fall of last year, and we asked people about the provision of placebos in situations where known effective therapy existed, and we got some of these comments back: "We had internal agreement that it was not proper to conduct the placebo studies, but the industry chose not to follow our advice." So, clearly, it's not only the FDA that is at fault, by a long shot; sometimes the industry will choose not to follow the advice of the Agency. Somebody else said, "Very difficult issue and very situationally-dependent." My point again, it is situation-by- situation, condition-by-condition guidelines that are necessary. There are situations, such as short duration, proof of concept studies, where placebo control would be appropriate, but in a long-term study, approved active control would be preferable. Now, there are four underlying issues here, I think. The first is the ethics of placebo use and the requirement under all existing ethical codes to provide maximum protection to patients. Clearly, what is in effect being said, or what is in part up for debate over here is, what degree of suffering is acceptable to patients or, what degree of risk to patients are we willing to tolerate? And that's part of what, that's one of the issues. So, the ethical protection of subjects is one element of it. The second issue is the clinical usefulness of the data that are generated, to doctors, to patients, and to payers. And we would argue that, as a general matter, active controlled trials in circumstances where known effective therapy already exists, will provide information that is more useful to doctors, patients, and payers. The average physician who has somebody coming in with hypertension, does not so much ask themselves, Do I provide Captopril or nothing? They ask themselves, Do they provide Captopril or hydrochlorothiazide, or any of the other 60 drugs that are on the market? And there is a dearth of this kind of information. Part of that, not only because of FDA, but part of it comes from the FDA's continuing efforts on the need for placebo control trials. Of course, studies that provide no useful data, are not useful and shouldn't be conducted, and so, there is a need to be able to draw useful inferences. And finally, there is a balance between the so-called needs of regulators in the pharmaceutical industry, which as Dr. Wolfe has pointed out, often weigh in favor of placebo control trials, versus the needs of patients, doctors, and payers, who would probably be better served by more active control trials. The issue here is the balancing of these clinical usefulness issues and the ethical issues against the needs of others, and we tend to put the ethical issues and the needs of the patients ahead of the needs of the pharmaceutical industry and regulators. The issue really, when we come down to it is, the differential incidence of harm. The question is, what is the difference in the incidence of some serious outcome, like a stroke or a cardiovascular event, or patient discomfort, like severe pain or migraine, between the placebo group and the treated group, in previous clinical trials? And from that, we should be able to, since there will in general be previous placebo control trials that precede the study that we will now be discussing, you will be able to estimate what the absolute value of the difference is between the treated and the untreated. What is needed then is a determination of the threshold for the differential incidence of serious outcome, or patient discomfort, above which the placebo becomes unacceptable. And in part what I think we need to do is start making an agreement upon what that threshold might be. And that I think will help us to develop disease-by-disease guidelines. Finally, as Dr. Wolfe pointed out, it is not enough to say that informed consent was obtained, or that IRB approval was obtained, and therefore, the study is ethical. The medical literature is littered with wholly unethical studies that were obtained with supposedly good informed consent, or that went through Institutional Review Boards. Sure, informed consent and Institutional Review Board approval are necessary, but it is the responsibility of the investigator and those in the regulatory community that help those investigators to design their studies, to act independently and to make sure that, not only is informed consent and IRB approval obtained, but that the studies themselves are ethically designed. DR. PAUL STOLLEY: Well, thank you for inviting me. I have sort of a bag of miscellaneous comments that I would like to bring to you. I think the goal of a study is probably the most important thing that drives the design, and the FDA and the Sponsor want to know, is the drug efficacious? Does it work? The practitioner, once that is established, really would like to know, How does this drug compare to others in its class, and should I choose it over its competitors for the same indication? And that of course would take a different kind of a design, a head-to-head design, usually. In terms of the use of placebo, when you get into the grey area, I think people of goodwill may disagree, and one way of trying to get over that disagreement is to try to quantitate the amount of suffering that somebody on the placebo might endure, and see if the different people who are arguing the case can agree as to the amount and type of suffering, and then whether it is acceptable. One of the things we haven't discussed is unblinding in trials. Unblinding occurs a lot more than we would like to think. The one of the few trials that showed that Vitamin C was effective in preventing the common cold was a trial in which the subjects had unblinded it by biting through the tablet and getting that acidic taste of Vitamin C, and all of the benefit, when they reanalyzed it, were in those that unblinded themselves, and knew they were taking Vitamin C. The trial was repeated, and that flavor was given to the placebo as well, and the benefit didn't show up on repeating it, so unblinding can be subtle, and very damaging. Bradford Hill tells the story of a crossover trial in which they went from Drug A to Drug B, and one of the subjects came up to him and said, "Dr. Hill, did you change the drugs on me last month?" And he said, "Why do you ask?" And the answer was, "Well, last month when I threw them in the toilet, they floated, and now they sink." (Laughter.) And going back and reading Bradford Hill is always edifying, because, as I was preparing for this, I was reading the defense of the use of the probability value of five, for efficacy, or for the decision rule, because it is just generally accepted by the scientific community, but in preparing an article on how Sir Ronald Fisher got to be on the wrong side of the lung cancer, cigarette-smoking controversy, I read a lot about him, and he is the one who sanctified P=.05, and you know, there is no rationale for that. He just selected it because it sounded reasonable. That's the entire theoretic rationale for that choice. There is no reason why that can't vary depending on circumstances having to do with the toxicity of the drug, need for therapy and so on. And a number of other issues. Perhaps even Sir Ronald Fisher, who never liked to change his mind, might agree with us. And then I guess my final remark is that, the balancing of statistical efficiency and the maximal informativeness of a trial, reducing human suffering, is a problem we face here. And at times, being able to do a quick, decisive trial using a placebo, might cause less suffering than a very lengthy trial where you don't, and the opposite of course is true. So, some way of quantitating the human suffering that is involved in placebo trials is needed, and must be discussed. I think perhaps patients could be brought in on this decision. I still treat patients and it's kind of amusing to me, a little amusing to me, to have a patient come in and describe a mild headache that they get in the front, but it disappears when they take one Tylenol, and now as I am listening to them, I have this migraine, you know. And it's just, can I finish the hour out? And you know, so there are headaches and there are headaches, and I think we have to try to get the patients involved. Thank you. DR. WOODCOCK: Bob? DR. ROBERT TEMPLE: Okay, I want to go quickly through a few important concepts: briefly discuss the ethical issue as I see it; address in more detail what's been considered so far, why this issue is important. Why do you need a placebo? Why can't you just do an active control? Discuss a few practical considerations, basically ways to design around the ethical problem, if there is one. Finally, how to make the case that an active control trial would be informative in a particular instance. I want to make it clear. The issue about placebo versus other designs is fundamental to demonstrating that a drug is effective. I will argue that certain designs, active controls, are not informative in certain situations. If they are not informative, that means if we were to rely on them, we would put ineffective drugs into the marketplace. Nobody wants that, I know, but that's what the consequences of this discussion are. It's not only about what would be nice; you can't do a trial that hurts people, but you do have to have a trial that is informative. So, sometimes if you can't do a placebo control trial, you just can't get there from here. We can't approve the drug at all. Anyway. Let me give a few more details. The ethical issue, as you have heard is, if there is known effective therapy for a condition, when is it ethical to deny this treatment to some patients in a clinical trial? You have heard about the Declaration of Helsinki, but the most extreme view expressed by Rothman and Michaels some time ago is that, "If there is effective therapy, you can never deny people that therapy in a clinical trial." Now, I understand there may have been a phone conversation where he might have changed his mind a little bit on that, but that is the -- slightly -- but that was the position. And he really meant it. I've been on stage with him. He meant, you can't do placebo control trials in baldness, because we have Rogaine. You can't do seasonal allergic rhinitis control trials, because we have lots of effective antihistamines, et cetera, etceteras. As I said, that is the most extreme view. I would say, my view is, and our view generally has been, that if we assume careful, informed consent, and appropriate review by an IRB/IEC, patients can be asked to participate in a placebo control trial, even when there is existing therapy, when the risk of lack of treatment is only discomfort, even fairly severe discomfort. We will get to this later, but a migraineur knows what a migraine feels like; who would be better prepared to describe whether they are willing to defer therapy for two hours, than someone with the disease? They know. If they are willing, and it has been explained to them, and if they know they can get out of the study if it gets too bad, I would allege that it is ethical. That doesn't mean that patients are going to be willing to enter the study. That's a different question. But I think it's ethical. For that reason I would consider most psychiatric conditions, outpatient depression, obsessive-compulsive disease, panic disorder, anxiety, angina, seasonal allergies, lots of other things, acceptable. This presumes -- I want to emphasize this -- that the patient has been properly informed, knows that there is existing therapy, and all that stuff is done properly. You have got to assume that; if that is in doubt, fix that. Now, it is completely clear; we do not disagree on this at all, that if there is adequate evidence that available treatment improves survival, or prevents irreversible morbidity in the intended study population, and the treatment is available -- We'll get to this, maybe a little bit later. I think the question of what you can do in another country that doesn't have therapy is a complicated one. Sid referred to that. I'm not going to go into that, now. But talking only about the United States, where presumably all the therapies that are available to people are available, if you know that there is improved survival, or something like that, you can't do placebo control trials. We agree, and I'm sure we don't have the lists, Sid, but I can guarantee, we have told people they cannot do certain trials they wanted to do, and have argued about them, gotten modifications, and so on. DR. WOLFE: How many times in the last five years? DR. TEMPLE: Yes, see I can't tell you that, but I know -- Sid wanted to know how many times in the last five years. But he was just kidding. (Laughter.) DR. WOLFE: No, he wasn't. (Laughter.) DR. TEMPLE: I don't know. A lot of times, you don't get an application like that, Sid, because they already know we wouldn't accept it. A lot of times they already know we wouldn't accept it, because we've told them. So, you can't do placebo control trials -- you can't do placebo control trials with thrombolytics, post-infarction beta-blockers, post- infarction aspirin, post-infarction ACE-inhibitors, at least in patients with ventricular dysfunction. You can't do antibiotic, placebo-controlled antibiotic trials in dirty surgery. You can't not treat patients with childhood, acute lymphocytic leukemia, testicular cancer. You can't do a long-term trial of moderate to severe hypertension anymore. We know there's a clear benefit. An interesting case is, we don't think you can actually do congestive heart failure trials anymore, because ACE-inhibitors have been shown to improve survival in virtually every phase, so maybe for a day or two, but beyond that, you are getting into territory where you can't do it. I need to remind you, however, that just because you can't do a placebo control trial, doesn't mean an active control trial is informative. An active controlled, beta-blocker, post- infarction trial is uninformative. Thirty out of the 35 trials, approximately, that have been done in that condition have failed to distinguish drug from placebo, So, what is it going to mean if you don't distinguish one drug from another? A point I will return to. So, in those cases, you just can't do it. There are a lot of situations in which the risk to patients could be debated. There is a current debate going on about whether you harm schizophrenics by taking them off active therapy, or by delaying therapy. That bears discussion. There are sophisticated people on each side. The data on the benefits of treating very mild hypertension, 90 to 95, are quite weak; some people might decide that could still be studied, some people probably would not. And there are many examples of that; I'm not going to try to do them all. Now, the next point is, Why does this matter? Why can't you just do an active control trial, show no difference, no significant difference between the two treatments, and say, Ah-hah. I win. I'm as good as the other drug? And that's what I want to explain. I should mention here that the crucial difference is between trials that show a difference between treatments, and trials that show no difference between treatments. The difference could be, again, placebo; it could be an active drug; it could be dose-response. Those are all informative. The problem arises when what your goal, is to show no difference. And the problem is, the question of assay- sensitivity. In order for failure to show a difference to mean anything, you must know that the trial had the capacity to distinguish between an active drug and an inert drug. That is, an active drug and placebo, if there had been a placebo there. That's called assay-sensitivity. The ability to detect differences. Unfortunately, for a wide variety of drugs that we know are effective, you cannot make that assumption about any given trial. I will give just a small number of examples. So, what is the problem with equivalence trials? There is the historical assumption of assay- sensitivity. When you do an active control trial, you don't have a placebo. You don't show that a drug beats placebo in that trial, you have to deduce it, from experience in the past, or something else. Now, that is unless you do a nice study called a three-arm trial, in which you use placebo, active control, and new drug, which gets you information about the comparison, and also does it in a way that is informative. That's a very good design. Very common in depression. So, one assumption is, you have to assume that there is assay-sensitivity. You don't measure it. I'll talk about that a little more. The second is, that there is a lack of incentives to study excellence. When you are trying to show difference between treatments, you have to be on good behavior. If you are trying not to show a difference, there are a variety of things you can do that would undermine the study, or fail to make the study sensitive to showing drug effects, and you don't have to be unduly cynical to think that that is a bad kind of incentive to give investigators and drug companies. We will leave the third one out; that's not really critical. Okay, so, in active control, equivalence trial where the goal is to show no difference, there is an often unstated, not always recognized assumption, which is that the active drug, the control, was effective in the particular study in question. That is not necessarily true, and it's not testable, and it means that these trials are historically controlled trials, in a sense, and we know what we think of them. We put this in regulations -- let me give you one example. There are an infinite number one could do. These are the results of six controlled studies, three-armed trials, of a new antidepressant, Nomifensine, no longer with us because it causes hemolytic anemia, but it is an effective agent; Imipramine, standard antidepressant; and placebo. You don't see the placebo there. The trials measured the HAM-D. They were all at four weeks, and they did an analysis using common baseline, we don't do that very much anymore. And what you see under New and Imipramine, is the Ham- D at the end of four weeks, and you can see there was a change of about ten points on the HAM-D. A typical response for an active drug. And you can also see that the new drug and Imipramine are right on top of each other, every time; 13.4 versus 12.8; 13.0 versus 13.4; they are all essentially identical. If you were to believe the equivalence trials, this would be the case you would believe in. Now you see the placebo arm. In five out of these six trials, the trials lack any ability to even hint at a difference between an active drug and a placebo. The only trial that was able to distinguish between drug and placebo was the fourth one down, a trial with seven people per group. I mean, talk about ridiculous. No one would take that seriously. But it was the only trial that had assay- sensitivity, and readily distinguished placebo from the two new active drugs. The other trials couldn't tell anything from anything. If we had developed this drug, I mean, you could take a completely inert substance, like the placebo, and win by comparing it with active drugs, in trials of this kind. As far as we know, there was nothing wrong with these trials. They were a little small, although 30 is not that small, but it makes the point. This is the experience that has been observed in depression, and other illnesses, over and over again; it's not unique. I'm not going to go through this, but I went through the last three years of trials in depression, in psychosis, in OCD, and in panic; about a quarter to a third of the trials, and they are decent-sized, 40 to 80, couldn't distinguish drug from placebo. That is the nature of the drugs we have. There is no doubt that the drugs work. They are more effective than placebo, far more than the one in 20 that you would get at a P-value of .05; it's close to a third. They work. They just don't work all the time. We don't even know if it's the drug's fault. In some cases, people just got better by themselves. There are a lot of reasons for a study to fail. Well, they are not small. They are 40 -- these -- I'll go back. Some of these are very large trials, in fact. I guess you can't go home again. One of them on Quetiapine at the mid-dose had 200 per group. That's the largest trial I have ever heard of in psychosis. Still couldn't tell anything. The Olanzepine trials were 80 and 50. Those are absolutely typical. You don't know that a larger trial would do it. You can't tell that, until somebody goes and does larger trials and proves it, and nobody has the incentive to bother to do that. Anyway, if you can't be very certain that a positive control in the study would have beaten the placebo group, had one been there. The fundamental assumption of the positive control study can't be made, and you can't use that design to reach an effectiveness conclusion. And if you can't reach an effectiveness conclusion, we can't rely on it. Just as a throw-in. If you want to know how one drug compares with another, you have got the same problem. A study that can't tell drug from placebo can't accurately compare two drugs. If you want to know whether the two drugs compare, you need a placebo, also. And that's a very good design. We like those a lot, even though we don't insist on them. Briefly, about the incentives. Certain kinds of sloppiness widen the confidence interval, and will not lead you to a false conclusion about whether a drug is equivalent to another. For example, if you have a lot of measurement error, you might not reach a false conclusion. But there are some things you can do that will not increase variance, but that will obscure differences. For example, you can mix up the drugs. Now, you may think that doesn't happen often, but ask people at a drug company, and they know it does. You can have very poor compliance. You can have people who all get better spontaneously. You can have people who don't have the disease, because the diagnostic criteria weren't very good. All of those changes guarantee that you will be equivalent to the drug you are comparing yourself to. That is not a good set of circumstances to give people, and you don't have to be unduly cynical; I'm not that cynical, but -- to think anybody would do that on purpose, but you might be. (Laughter.) Anyway. It's not a good situation. When you are trying to, all of these examples, the example I showed you and the many others I could, of where effective drugs don't beat placebo, all were in situations where people were trying as hard as they knew how. They were having wash-out periods, they were assuring compliance. They were yelling at people to take the drug and try not to mix up the drug. And they still couldn't do it all the time. Now, if you wanted to do an active control trial, because you think the circumstances are right, here is what you have to do. You have to show that placebo control trials regularly allow a detection of a difference between drug and placebo. We have never defined what regularly is; it means, almost always. Then you have got to use a design, endpoints, sample size, all those kinds of things, that is pretty close to that, so you know that it is relevant. You can't -- well, you can't take a different endpoint and think you know anything. And I will leave the third one out, because that goes to how you actually do equivalence trials. So, you have to make this statement sort of historically. Now, can active control trials be credible? Of course they can. Most bacterial infections are studied in active control trials. We accept them with thrombolytics on the basis of an analysis that Jay can tell you about. The treatment of deep vein thrombosis, it's easy to tell drug from placebo. Many stages of HIV infection, highly responsive tumors, anaesthetic agents; most people stay awake until they get the anesthetic, there are exceptions. (Laughter.) Beta agonists and bronchospasms, you get a prompt, you get an immediate response. It's pretty easy to tell, and we do utilize active control trials, so they are appropriate. Do I have another two minutes? Okay. Just a couple of practical considerations. I think this is of interest to everybody. Sometimes, you just can't do a placebo control trial. What can you do, in that case? One of the best things you can do is you do an add-on trial. Everybody gets the standard therapy, and you test the new therapy, added onto the old therapy. That only works when the drugs are different pharmacologically; that's how anti-epileptic drugs are done now, that's the only way. You don't have drug versus placebo; people are uncomfortable doing that; they don't want to. Heart failure, that's how you do it. You take the standard therapy, diuretic, ACE-inhibitor and so on, and you add something new. Now, you're probably going to have to take a beta-blocker in addition. This gives no data on monotherapy, but at least it shows you the drug works in some environment, and that's better than nothing. You can beat the standard, tPA, b- streptokinase. Winning is always good. It could even be an imperfect dose of the standard and still be informative. That might be ethically questionable. Dose-response studies, if ethically acceptable, are also informative. Now, you can't cheat. You can't give a trivial dose and pretend you're not giving the placebo. That's no good. That doesn't solve any ethical problems. You can study a population subset not known to benefit. I think that was mentioned before. It's interesting to think about the SHEP Study. That was a study in isolated systolic hypertension in the elderly. What we had known for 20 years by then, that treating hypertension was good for you and improved survival in other things. Why was that okay? Because no one had ever done a study in isolated systolic hypertension in the elderly. The epidemiology says it looked at the other situations, but everybody is pretty comfortable with that, and I totally agree with Sid. You couldn't do SYST-EUR in this Country. Whether you can do it somewhere else is a different debate. Yes. Well, we would -- if someone came to us with that, which no one did, we would have said, No, you can't do that study. It's okay to study nonresponders to standard treatment; then you're not depriving them of anything. Clozapine was developed that way; Deprenyl was developed that way. One thing you can do if the duration is of interest, is provide early escape. If you are doing a short-term hypertension study -- we'll talk about those later -- and people's blood pressures start going up, and you're worried about it, get them out. Maybe that makes you comfortable doing a short-term hypertension study when you wouldn't have been, otherwise. Sometimes you can treat people for a long time with active drug, do a randomized placebo control withdrawal at the end, and as soon as their disease, whatever it is, comes back, you take them out of it, count them as a failure, and note the difference between the two. That may be ethically comfortable, whereas a long-term, you know, a six-month study in hypertension is out of the question. You can't do that. But you can treat people with an active control. Take the drug away, and then they're a little hypertensive for two weeks. Many people would accept that; we'll probably talk about that later. Okay, well, you've heard the conclusions. We'll pass that one. DR. WOODCOCK: Thank you, Bob. All right, we have heard people's take on a wide variety of the issues relating to the placebo. From one hand, the imperative not to do harm to people, but from Dr. Stolley we heard the issue of, you know, how much harm is harm? And we're always drawing blood from people and harming them in different ways. What is tolerable in studies? To the other side, which you heard mainly from Dr. Temple which is, well, we can't -- it's also unethical to studies that are completely uninformative, because then you are exposing human subjects for basically no purpose. And so, how do we wind our way between these? And to illustrate some of these issues, we want to talk about three case studies, and is it all right with the Panel if we start with the migraine? We can dispense with that fairly quickly, perhaps, and I think it does illustrate some of the issues? Migraine Case Study DR. WOODCOCK: All right. We wondered how people got their e-mails. Let me just read briefly. "There's an element that causes symptoms, but little likelihood of causing long-term harm. Existing treatments are effective in treating symptoms, but don't have an impact on reducing the possibility of long-term harm or effects." Migraine is going to be the first one of these we discuss. And what I would like to do is start with the questions. The first question for migraines is, "Can a placebo control study design be used ethically? If so, under what circumstances, and with what limitations?" And any people who want to contribute on what would be the best design. Who would like to start off on this? Yes. All right. Dr. Lurie DR. LURIE: I understood the process slightly differently, but that's fine. As you said, it's an example of a condition where a patient's symptoms are all that's really relevant. There's no long-term risk, as there is with hypertension, for example. So, the question, as I said earlier, becomes How much patient suffering is acceptable? And we have said a certain amount is acceptable for mild pain, cold symptoms, allergy medications, but if anybody has suffered a migraine in this room, and I suspect there are many, perhaps from hearing us speak before, the two hours of intense suffering, as with migraine, we think is a very substantial discomfort to patients, and an early escape, as FDA seems to be doing these days, after two hours, does not seem acceptable to us. I went back and looked at some of the recent migraine studies, and noticed several things; one was the absence of any attempt for a meta-analysis or any kind of review, the way one finds in some of these other conditions we'll discuss. And the studies that I looked at, at least, were numbingly and consistently positive, and again, just the recent ones. But, the placebo response was reasonably uniform. It works out to be about a third of people who respond, if you define responding as moving from a moderate or a severe headache to one that is mild or not present at all. What is interesting in looking at these is that alternative designs have been used. One was a study in the Academic Emergency Medicine from 1998, which was an equivalency or a positive control study, one can't quite tell, but compared to IV Ketorolac to IV Prochlorperazine. It showed that Prochlorperazine was better, and that I think is very useful clinically for physicians. The problem is, of course, it's a risk to the drug company, and if IV Ketorolac's sponsor was the sponsor of this trial, they would probably be quite disappointed by the results. Patients, on the other hand, would benefit greatly. There were also some interesting dosing studies that I think are also alternatives, for example, an article by Klausen in Headache in 1997 looked at Naratriptan at a number of different doses, and at the highest dose, there was a 60% response rate, followed by a 50% response rate at a little bit lower dose, but the two lowest doses had 35 and 32% response rates, and the placebo was 34%. I think that that's the kind of dosing study that would be helpful, because anybody at the top three doses would have said, 60% versus 50%, well, that's a lot better than what we assumed the placebo response rate to be, about a third, which in fact is what it turned out to be, in this particular study, and even if they hadn't made that assumption, they would have seen a drop-off at the third dose, from 50% down to 35%, and they said, we wouldn't have used that. So, I think that, again, for us, the important element here is to not downplay the severity of two hours of unnecessary migraine that some one- third of people will experience, and to look at the fact that there are some alternative designs that can be informative. DR. WOLFE: Just a little quick follow-up on that. I mean, I agree with what Bob said before, was that one should not mask in what appears to be just a dosing study, what may be de facto, a placebo arm. I think that once you have been through at least a few drugs in this category, you have a better idea of how to design a dosing study. I mean, unlike the hypertension example which we will get to, where you use a surrogate, here you have got intense pain, and you should be able, with a dosing study that is honest and based on what you knew from the pharmacokinetics of the drug and from the other drug in its class, you can design one that can get you all the information you need, without resorting to a placebo. DR. WOODCOCK: Dr. Stolley. DR. STOLLEY: Well, I've been an authority on migraines since age 13, and the reason I think a placebo trial is permissible is that you can put people to sleep after two hours. You can give them a hypnotic, and often that's the only way to help somebody with a very severe migraine, is to give them a hypnotic so they can sleep. So, I think that that protects patients. DR. WOODCOCK: Dr. DeLap? DR. DELAP: Yes, I think this is a very interesting discussion. I think, one, I would like to pick up on just a couple of things from the introductory discussions by Dr. Stolley and Lurie, pertaining to bringing patients into the process as much as possible, and the fact that something that passes an IRB and an informed consent screen, that's not necessarily reassurance that it is the right thing to do. I think investigators of course have different -- well, I would say, two primary motivations in getting involved in clinical research. One is to provide treatment for their patients, and the other is to study the new treatments as they are administering them. Those are not necessarily the same goals, sometimes, and there can be some conflicts. Patients come to the experiment with, on the one hand, a hope of benefit. They really, I don't think anyone really wants to get an inactive treatment for something that is causing them serious problems, but patients also I think bring some understanding of the research process, and certainly, some, what I would describe as, beneficence or altruism to the process, as well. And their understanding of the necessity for clinical research to advance the field. Now, of course, the investigators and companies who are developing products oftentimes know 'way more about the underlying disease process and about the drugs that are being tested. The patients know a lot more about symptoms that they are experiencing. And again, I think that's one of the really relevant points in the migraine case. I think if the patient has the sense, that understanding that they are going into a placebo control trial and they know their migraine, they know their symptoms, they know what they are up against, and if they want to agree to participate in a trial, as opposed to taking some medication that they may have taken in the past successfully, I think there, they really know pretty much what they're getting into. I mean, there aren't a lot of other secrets. They know that they might get a treatment that's not going to work, and they might have whatever discomfort they typically have, for that period of time. And so, I would agree with Dr. Stolley's comment, but I think as long as you have ways to, in the end, ensure that the patient is comfortable, if they're willing to endure that, then I think the design is acceptable. Now, I think there are also other designs that may be acceptable, and I think that's another issue. If someone wishes to set forth a design that has an active control, and take into consideration all of the historical information about placebos, and what they do, and how many patients you might need to study, and what the confidence intervals need to be of the relief, to satisfy yourself that you have fairly tested the new drug. You know, I think that kind of a design can also be acceptable. It may take more patients, it may take longer to do, it may be somewhat less definitive, unless adequate numbers of patients are studied. DR. WOODCOCK: Okay, so I think I'll call on the rest of the people in a second, but let's sort of review the bidding here. With migraine, what we have heard is, there is some opinion that the pain that people, to which I can personally attest is occasionally severe pain that people are being asked to endure, and I think Dr. Lurie expressed the view that, in fact, given that in his point of view there are alternative trial designs, it could be advanced that that would be too much to ask people -- am I paraphrasing you? That that would not be appropriate to ask people to endure those two hours of pain, because there might be alternative ways to get this information. We also heard, though, that perhaps people know what their pain is, patients can be involved in these decisions, and an alternative view was, it's not too much to ask people, since they really know the whole story for their migraine. They know what they are getting into, and they are capable of perhaps, in that view, deciding whether they would enter into this trial or not. And I think this is illustrating a sort of fundamental difference here which is, why subject people to any pain, when it could be avoided by some other trial design, versus well, people can make these decisions. And I think underlying that issue is the fact that, although these alternative trial designs may work in some cases, they also have a much higher chance of not being successful, and not getting the information. So, if that's fair to all the participants, that's where we are right now. Do we have more comments? Jay? DR. SIEGEL: Well, first I'd say, I would agree largely with Dr. DeLap's comments. I think though that some attention has to be paid to the, and probably there isn't time to address the critical issues here, but to the questions that were raised about the quality of the informed consent process, because if you ask a patient to undergo this sort of pain, the critical issue is, are you asking under any coercive, or otherwise manipulative conditions? And I think that if you're not, then there are many patients who really wish to be part of medical science and a medical study, and ought to have that ability, if all is done openly and fairly. So, I think significant attention does need to be paid to that process, how it's done. I think it is a little bit, if you can ensure good informed consent, then it's paternalistic to say that patients shouldn't be allowed to make the determination as to whether that two hours of pain is worth enduring. A second issue that this and all these questions raise in my mind, that I'm wondering about is, are, the way we framed this question, can you ask the patient to potentially endure this pain, is based on a strong presumption that the active control will work. Because the active control design is also a design in which significant numbers of patients are not receiving proven effective therapy. And indeed, if you know that the active control works, then you ought not be doing the trial. Well, at least, you might be doing it for other reasons, but not to prove efficacy, if you know efficacy. Presumably, if you are doing the trial to prove efficacy, you don't know if the study works, so you are in fact asking patients to take the risk of having an ineffective therapy in either design. And then the active control does raise some additional issues about the informed consent process, because one of the concerns about informed consent, in general, is that many patients in many trials, after informed consent, despite perhaps being told that it's unknown if the drug works, sign onto the trial quite convinced that the drug works. And if you are to obtain true informed consent, and tell the patients in fact that even when you do the active control trial, what you are telling the patient is, we have a proven effective therapy here, but we want to enroll you in a trial in which you may receive a therapy that doesn't work. And I think there are important differences, but some of the ethical issues are the same, and we need to guard against that presumption, that all experimental therapies work. DR. WOODCOCK: As we well know, to bitter experience, is not true. DR. WOLFE: Just a quick comment on that which is, I think that the implication was not simply that the patient needs to be involved, as a patient, in the informed consent decision, as to whether she or he participates in the trial. I think that if we took a roomful this size of people who had had one or more, or ten or more migraine headaches, and explained to them that the only alternative was not a placebo versus a new active drug, but that an alternative design was the active control design, assuming that the active control was something that had previously been found to work, versus the placebo, which we were at a point, years ago with these kinds of drugs -- And fully inform them. I agree with Jay that you need to explain that in the placebo control trial, there's a certainty that the placebo won't work, and that there is at least some hope that the other will work, whereas in the active control, you know that one of them has worked, and you don't know yet whether the other one works. But, I think you might get a different kind of decision as to which of the two trial designs would be preferable, if you involved people who have actually experienced a headache. Now, this is possible because it is the effect, or the effect of not being treated is right there, very immediate, measurable on their part. One can get some quantitative measurement and compare them for themselves when they start using the drug an hour or two later and so forth. So, I think that we might get different design trials, particularly now, and maybe ten years ago it was different, but there is an enormous amount of data which Dr. Lurie mentioned is just extraordinarily repetitive when you look at it, trial after trial, with the same general kinds of findings, that when the next one of the drugs in this family comes up, I think that a group of patients might choose to have an active control trial, if they were allowed to. DR. WOODCOCK: Yes. Could I just comment on, just from an outsider's viewpoint, and not a member of the Panel? But, Dr. Lurie, didn't you say that in signing up for a placebo, you diminish your chances of 50% being better at two hours, to about 30%? DR. LURIE: Yes, 50 to 60%, down to about 30%. DR. WOODCOCK: Okay. So, about half the people may not get any relief, anyway? DR. LURIE: Bob is saying -- Bob is saying, Big deal. And I think that actually is a big deal. I personally am not a migraine sufferer, and I'm happy for it, and I would feel differently -- I mean, an informed consent form would state not only the things that Sid has said, but it would state, Because we are using a placebo control trial instead of an active control trial, which very often can provide useful information, you will with certainty, if assigned to the control group, have a 30% greater likelihood of having migraine for two hours. And that's what a truly informed consent form would say. Now, as a general matter, they don't say that, because everybody knows that if you really disclosed that, some people would sign up, I think that's true. But if you really said that, I think people wouldn't. And especially with the direction in which clinical trials are going these days, where there is sort of a decentralization of clinical trials in doctors' offices, and very often it's the doctor, himself, who is enroll -- or herself, excuse me -- who is enrolling the patient, and acting as the investigator, it is very difficult for a patient to say no in a circumstance like that. DR. TEMPLE: May I? DR. WOODCOCK: Okay, let me go in order here. Bob, first. DR. TEMPLE: There are a few misconceptions that need to be corrected. DR. WOODCOCK: Speak into the microphone. DR. TEMPLE: There are a few misconceptions that need to be corrected. The example of the Ketorolac-Prochlorperazine study was successful, because one drug was better than the other. If that study had shown no difference between the two drugs, it would have been completely uninformative. You wouldn't have learned a thing. It's not true, you should know this, that sponsors do not particularly prefer to do placebo control trials. One of the reasons we told people that they had to do them was that they were eager not to. It's very easy to win in an active control trial. And people do active control trials in Europe all the time. Europe is finally getting the idea that they need to add a placebo group to make them informative, but these trials happen all the time. In the example you gave, the very results you described would have been uninformative about the value of the low doses, completely uninformative if you didn't have a placebo to know that the placebo was doing about the same as the low doses. It is of interest to know what the lowest effective dose is. It's particularly of interest in these drugs, because there is a little concern about coronary artery spasm, and you want to use the lowest dose you can. So, it's not trivial to want to know what the dose response is. Some people might want to start on a lower dose, if they are -- DR. WOLFE: But you can do that without a placebo. You can do a dose response curve without involving the placebo. DR. TEMPLE: But you won't learn whether the low dose works. If you've got a 40% response rate on the low dose, and no placebo, what would you know? Did it work or didn't it work? You don't know, because you don't have a placebo. DR. WOLFE: But you would have a good idea from the previous trials. DR. TEMPLE: No, I don't -- DR. WOODCOCK: Okay, Jay -- DR. TEMPLE: -- agree at all. Finally, last point. I have got to say, migraine is the purest possible case where the patient knows exactly what she or he is getting into, and I agree with Jay's use of the word, paternalistic. I'm perfectly content to let Paul decide whether he wants to be in this trial, and you don't need to take a vote of this room to do it. He can figure it out. You know what a migraine is. You have had them. DR. STOLLEY: I want to be in the trial and I want the placebo. (Laughter.) DR. WOODCOCK: You've had the treatments, right? DR. WOLFE: So you need to have your head examined. All right. (Laughter.) DR. WOODCOCK: We'll finish up with Jay, and then we're going to take some questions from the floor, or comments, on this particular example which is a fairly pure, I think, balancing of the issue of the interpretation of the amount of harm that comes to subjects in a very pure case, for a short period of time and so on, versus the issues of what we can learn and how informative other trial designs might be. Jay. DR. SIEGEL: Yes. I'd just like to comment on a substantial note of agreement that I heard, that may have been missed by some in the face of disagreement. Not so much on the ethical part, but on the informational part, or the ability to draw inferences, which is that Dr. Temple noted that the times when you can draw inference from an active control trial, are times when your drug would reliably perform better than placebo by a determinable amount. And Dr. Lurie noted that, in this case, and I'm not arguing specifically the merits of this case, but in this case, what the informed consent should say, the word to use is, with certainty. That, with certainty, if you are randomized to placebo, you will have an outcome that is 30%, or one-third, or whatever the number is, inferior to the outcome you would get had you received an active control. And I think we probably do have some level of agreement that these standards of, that the active control is better than placebo reliably, or with certainty -- with certainty may be a somewhat higher standard -- but are similar. So that if in fact we -- you know, and not arguing the merits of migraine per se. But if in fact we have a situation where we can say with certainty, that under the conditions of the trial, the patient would do worse with placebo than with active control, we have a situation where we can draw inferences from an active control trial. DR. WOODCOCK: Agreed. I think there's agreement on that. All right. Well, why don't we move to -- There's not? Okay, why don't we move to questions from the audience. Come up to the microphone, or if you have a comment. Maybe we have some migraine sufferers in the audience, who would like to give some editorial comments. Dr. Johnson. DR. JOHNSON: I'm not a migraine sufferer, but I am interested in the quantitation of this notion that, if there is an alternative design and an active control design, and let's say you determine that it requires ten times the number of patients, is somebody looking at the global ethical perspective here? Is that more unethical if you have to enroll ten times the number of patients with an active design, rather than one-tenth the patients with a placebo control design? DR. WOODCOCK: I'll have Dr. Lurie answer -- DR. LURIE: Well, there's -- that sort of ratio of active-controlled to placebo-controlled sample size requirement is a kind that is frequently bandied about, but generally not supported by what is actually necessary. Generally, the increment in the sample size requirement is nothing like ten times greater. It's generally some much more modest amount; 50% greater, or what have you. In the Africa studies for example that we looked at, it was about a 20% incremented studies, and I think -- in sample size. In the others, it depends on the particular tolerance that you are willing to take in an equivalency study, et cetera, et cetera. But, as a general matter, it will not be enormously greater. Let's put all of this in the context of the drug approval process, which as you all know, is long. And you know, you have to put it in the overall context of that, and particularly, in the context of a situation which we're talking about here, which is there already are effective drugs on the market. I think that, you know, where there is no effective drug on the market, well, it's a straightforward shot, and there's a rush to the market and you can use the placebo control. But I think that, again, just to emphasize, I don't think that you very frequently are going to run into a ten times greater sample size number. DR. WOODCOCK: Okay. Two more questions - - DR. TEMPLE: May I comment on that one? DR. WOODCOCK: Briefly. DR. TEMPLE: In an active control trial, you have to presume, you have to choose as your estimated effect, the lowest values you encounter. And that will lead you to get larger sample sizes, as a rule. I need to mention one other thing. We're all presuming that people entering these trials are true migraineurs. It's perfectly possible to screw that up and put in people who have different kinds of headaches, and your active control trial then becomes utterly spurious. You have got to remember that. There are a lot of problems when your obligation is to show no difference. DR. WOODCOCK: Two more questions, or comments. Go ahead. DR. YAES: Yes, Bob Yaes. There's another ethical issue which I think has been sort of touched on, but not really gone into, and that is, suppose you have a new drug, which is more effective than placebo, but less effective than currently available treatment. Is it ethical to bring such a drug to market? Now, one would hope that a drug company, you know, if they had an inkling that this was what the situation was, that they would throw the thing in the trashcan and that would be that. They'd go onto other things. But, it's conceivable that they might decide, you know, well, we can get it approved because we can show it's better than placebo, and then maybe we can effectively market it. If they don't do a comparative trial, then there is no way to show that that drug is less effective than drugs that are already on the market. Now, with migraine, I mean, that's a slight problem. I mean, it would mean that, if patients don't know that this new drug is less effective, then they'll suffer more pain, because they won't know that the other thing is better. But when you get into more serious, life- threatening diseases, like cancer and AIDS, then I think this consideration reinforces the idea that you don't want to use placebo. You don't want to use placebo because it would be unfair to the patients in the study, but you also don't want to use placebo because if you are dealing with an illness like AIDS or cancer, which kills you, I mean, then you don't want to let out any drug that is not at least as effective as drugs that are already available. DR. WOODCOCK: Okay. Well, let me just use the Chairman's prerogative and answer this, or the Chair's prerogative, or whatever. I think we don't have placebo controls, ordinarily, in life-threatening diseases that have successful therapy available. The issue of less effective drugs for the other range of treatments is a complicated one, which I don't think we can fully address here, today, and maybe we could take it up at another seminar, because there are things like a lower toxicity and so on that might make a less effective drug acceptable, or to a niche of the population. Final question here. DR. ANDREA SEGAL: I'd like somebody to talk about the influence of variability of placebo rate, and how this affects. It appears to me that I recall reading a study by a man named Beecher, and I believe this is where the original one-third placebo rate was drawn from, where the placebo variability I believe was from 15 to 58%, and there are many studies of sham procedures, such as cutting the internal mammary arteries to treat angina, that show that placebo rates can be as high as 70%. Maybe there is literature that shows that it goes even higher. And I would like somebody to comment on how that possible variability in placebo response can make an impact on non-placebo-based trials. DR. WOODCOCK: Dr. Wolfe. DR. WOLFE: This -- I'm not meaning to defer your question, but this issue is really right in the middle of the whole discussion of the third case that we have, which is depression, because clearly, as a function of whether it is three-day old reactive depression to a death or something like that, or two- year old serious, endogenous depression, there is a huge difference in the placebo response rate. One of the confounders in looking at all of the literature on antidepressants is the failure to clear out the different degrees of depression and so forth and so on. But I mean it's a very, very good point, and I was, as I read through that literature on depression, I was astounded at how long they have sort of realized this, and it hasn't filtered down into consciousness. DR. WOODCOCK: Dr. Stolley and then Dr. Temple. DR. STOLLEY: Well, I'm glad you mentioned the internal mammary artery ligation experiment that was done back around 1958, where everybody woke up thinking that -- they had an incision, and thinking they had had the procedure, but when they opened the envelope, it was ligate or not ligate, on a random basis. And the placebo rate was not as high as 70%, but it was on the order of 30 or 40%, and this showed that you could have a fairly striking placebo effect, when somebody thought that they had the surgical procedure, even though they didn't. It was partly this that led to the study that has been designed for Parkinson's patients, where they get a hole drilled in the skull, I believe, but only some of them get the procedure, and there, I would have to say that that's making too much of a placebo effect, because when you have a disease like Parkinson's Disease, where you may have diurnal variability, and you use something as crude as what they call the YAR Score, rigidity and ability to do things quickly, and so on, which is kind of a crude measure -- Nevertheless, it's a downward course for everybody. It's an unremitting disease. And I had a lot of trouble justifying a sham operation in that, and was wondering how that got approved. (Laughter.) DR. WOODCOCK: Okay, we'll move on. Dr. Temple, did you have a comment about this? Can you try and keep it brief? We have about 35 minutes left, okay, for the next two cases. DR. TEMPLE: The only point I would make is that describing the change in the placebo group as the placebo effect is a misnomer. It includes all of the changes due to natural history. However, a variable response in the placebo group is one of the hallmark features of trials of situations where active control trials are not very reliable, because the results are so different. DR. WOODCOCK: Right. Bob, did you have any -- DR. DELAP: Yes. I think we have sometimes a false sense of security, too, from the active control trials. People think that if you do an active control trial, and Drug A looks like Drug B, then they must be about the same, when in fact, all you're really saying is that you have established within a certain confidence interval margin, that you can't tell them apart. Now, you know, oftentimes what that means in practice is that Drug B could in fact be 30% worse on whatever outcome they're measuring, and it still turned out the same in this particular experiment. That's the simple statistics in the matter. DR. LURIE: Yes, but that's really no different than what you are seeing in a placebo control trial. You're saying, with some probability, this drug is superior to placebo. So, that argument -- what you are saying is true, but it really cuts both ways. In the end, one has to establish parameters of alpha values of tolerance levels and the like, and go from there. And if you understand the statistical interpretation correctly, you understand it. It's not certain that a drug that is statistically significantly better than a placebo, is better than placebo, it's 95% or more likely that that's the case. DR. DELAP: Well, yes, this is true. I think, again, a lot of our drugs are not terribly effective to begin with sometimes. I mean, you may -- migraine is an example of a situation where we do have drugs that do have effect for the majority of patients, but they're not always effective. So, there is a treatment effect of both placebo and if you are giving away 30% of the effect, you may be giving away most or all of what was there. So, you know, again, if you don't have assurance that you haven't lost the entire effect, then you know, I'm concerned that you open the door to having a drug out there that you would say from the study, looks the same. The practitioner will say, well, this drug worked the same as the other drug in the study, and can be used interchangeably, and yet, it really isn't doing anything. DR. WOODCOCK: All right. Well, will the Panel -- we have 35 minutes left, approximately. Would the Panelists like to move to hypertension as the second example? Yes. Let me just give the -- I think this was a very illuminating discussion. We see we have different estimations of the degree of harm from migraine, and that's one of the -- And we also have different estimations of the degree of informativeness of the different studies that might be done, although there is more agreement upon that. I like to use a simple-minded approach to understanding part of this equivalence issue which is, for the highly variable drugs with the small effects that Bob was talking about, and you do an equivalence trial without a placebo, the way to tell the lay public about this, I think is, you don't know whether both of them worked, or neither of them worked, because in many trials, the active drug doesn't work. It doesn't distinguish itself from placebo. I think that's a sort of easier way to understand it. Hypertension Case Study DR. WOODCOCK: Now, we are going to move to hypertension, as in the -- right. Hypertension. That's the one we want to do next? Okay. As an example of an illness, obviously, that can cause long-term harm if left untreated or inadequately treated for a substantial period of time. Of course, effective treatments are available. Limited periods of inadequate treatment commonly occur in medical practice -- nobody is saying that's you know -- For example, when a hypertensive patient comes in, presents originally, you may recommend diet, exercise, what have you, salt restriction, and other things, and not treat them with medication, and this may not cause harm, but no one has really proven the absence of harm. So, we are going to discuss hypertension and why don't we start with trying to go through these questions, telling whether or not they think the placebo-controlled study design can be used ethically. As you saw in your handout, these are typically short-term, 8 to 14 week, moderately short- term, placebo-controlled trials, looking at control of blood pressure. Who would like to start on this one? Dr. Temple, we'll let him go first. DR. TEMPLE: All right, let's -- 8 to 14 weeks is actually unusual. They are actually shorter than that, more like four to six. DR. WOODCOCK: That's what I thought. DR. TEMPLE: I need to tell you that hypertension trials occur that do not distinguish drug from placebo. There is a very large response in the placebo group, 5 up to 10 mm Hg. It could be digit- preference, we don't really know. We are exploring whether automated cuff designs might not have that, but the change in placebo is relatively large, compared to the added effect of the drug, which might be as little as 3 or 4 mm Hg. In addition, it is desirable here, as it is in other cases, to know what the lowest effective dose is, and the effect there can be quite small; it could be 2 mm Hg, and I will allege that you cannot properly learn these things from a trial that doesn't have a placebo. It is totally clear that nobody is going to, in this Country anyway, countenance long-term trials with outcomes using a placebo, so let it be clear, there is absolutely no disagreement about that, but a trial of four to six weeks is consistent, as you just said, with practice, and we think that there is little evidence, little possibility, there could be harm. We are in fact carrying out a study internally of something like 100,000 people who have been in trials, and will be back to tell you whether there is any possibility of harm, but at least at present, it seems entirely compatible with ordinary practice to have a period of, say, four weeks or six weeks. This is in people who are 95 to 105. You don't take your diastolic 130 and put them in a trial like this. That would never be done. Never be allowed. DR. WOODCOCK: Okay, Dr. Wolfe. DR. WOLFE: Yes. I disagree. But beyond those two words, Dr. Lurie will just briefly mention some of the evidence that the problems, as in the worsened cardiovascular outcomes, are pretty linear, and seem to go back to the beginning of these trials. So, that there is worry, even in a month or two. But I think, unlike the situation with depression, or migraine, where we have a clinical and measurable effect right away, here this is clearly a surrogate effect. We are looking at lowering of blood pressure, and hoping to get people into a, quote, "acceptable" range of blood pressure, compared with what they were before, or what they were when they were on another drug, and so forth. And the literature is filled with examples of studies, randomized controlled studies, where there wasn't any relationship between the extent of lowering the blood pressure in these trials, and the eventual outcome. A number of them have been published in the last couple of years. There's FACET, in which amlodipine, which lowered the blood pressure more than an ACE inhibitor, and had a much worse outcome in terms of cardiovascular problems. On the other end of the spectrum, there is the lowering too far of blood pressure, and there's a study published in the JAMA, I guess a couple of years ago, from the Group Health Cooperative of Puget Sound, where they showed that, as you start inching down to 87, 75, 70 in the blood pressure, you have increased risk of primary cardiac arrest. That's a case control study, not a randomized study, but I think the difference here is that, if one can agree that it is worthwhile getting someone's blood pressure into whatever range is thought to be okay for them, is a function somewhat of how old they are, because of the systolic blood pressure, then yes, there are some examples where, in randomized control trials, the antihypertensive agent wasn't distinguishable from the placebo. But I think that in the overwhelming majority, the drug, particularly when we are talking with mainly classes that we already have examples, whether it's ACE inhibitors or diuretics, or calcium channel blockers, or whatever, we have multiple examples already in each of those classes, where the drug is more, and statistically significantly more effective than a placebo. Peter, why don't you talk a little bit about the linear outcome? DR. LURIE: Well, earlier on, I made the argument that what we really need is some collective agreement about the amount of risk that patients can endure, and there's the subjective elements, as we discussed in the migraine case. Here, patients are in a sense asymptomatic, and therefore in a worse position to be able to decide for themselves what it would be like to have a heart attack in the future. If you look at the MRC, the SHEP Study, and the SYST-EUR Study, three of the leading antihypertensive studies, one thing that you notice is that, the best one can tell, the divide between the placebo and the treated arm begins right at the origin, right at the beginning, in a sense, but it goes more or less linearly out for as long as people are followed. I make that point because I've also said that we need to determine for these kinds of studies, an incidence of adverse effects, an excess incidence of adverse effects, which would become unacceptable for the conducting of a placebo control trial. So, the diversion extends all the 'way back to the first couple of months, and I went back to the MRC, SHEP and SYST-EUR studies, and calculated the excess risk of having a stroke or a major cardiovascular event, during, say, the first two months of therapy, and in the MRC, it's about .5 per 1000 strokes, or .7 per 1000 major cardiovascular events, in that two-month period. In both the SHEP and the SYST-EUR, you get about 1 in a 1000 additional strokes and 1.8 in additional major cardiovascular events. Now, the question is, is 1 in 1000 -- let's take that as an outcome -- is a 1 in 1000 chance of having a stroke or a major cardiovascular event unacceptable? And obviously, in the great majority of such studies, there will be no such event, or no such excess event, because the studies will be considerably smaller than 1000. But let's contrast that to levels of risk that are otherwise accepted in society. The Supreme Court has ruled that a lifetime risk of 1 in 1000 of occupational cancer is sufficient to mandate the Occupational Safety and Health Administration to act and regulate. That's 1 in 1000 in a lifetime to exposure to a potential carcinogen. We're talking about a 1 in 1000 in a two-month period. So, I think that this is certainly of the order of risk that leads to regulation in the occupational area, and I think it should lead to regulation, in this case, the preventing of placebo control trials in this area. DR. WOLFE: Just to comment on what Bob said, which is that there is quite a literature, again, something I learned by reading for this seminar this afternoon, on comparing the placebo effect in hypertensive trials, as measured by clinical blood pressure monitoring and ambulatory. You sort of alluded to that, and it looks as though there is still some, but it is considerably less if you do the 24-hour ambulatory monitoring. And that actually looks sort of hopeful, because if you can in some way diminish what appears to be a larger placebo effect, then even more reason not to do a placebo. DR. TEMPLE: We agree. And we also have a study going on about that. Peter, could you say exactly how you made your calculation? Did you assume a linear effect over time? DR. LURIE: Yes, that's what I said. I assumed a linear effect over time, because that's -- I mean, you know, the data are pretty much linear, since we have these -- DR. TEMPLE: Okay, well, lucky for you. We really will have 100,000 people studied in four to eight week trials, and we ought to be able to get an answer on that. I can tell you, as the trials go by, and they typically do have a couple thousand people in them, you don't see anything like that. But that's not big enough to be fully reassured. I mean, for any given application with 2000 people randomized -- DR. WOODCOCK: Yes, Bob, you have to -- yes. DR. TEMPLE: In any given application with a couple of thousand people in it, you cannot see an excess of those events in the placebo-treated group, but those are very small numbers, as Peter said. You're not going to see very much. That's why we have to do the pooled analysis. DR. WOLFE: Are you trying to encourage more use of ambulatory monitoring, though, in these trials? DR. TEMPLE: Were trying to analyze the data we have, to see if we can do that. Yes. DR. LURIE: I just want to really compliment you on doing that analysis, because that is the kind of analysis that's necessary. I mean, by and large, we do need to get to the point of quantifying excess risk, and coming to an agreement about how much is acceptable, and that's a very useful step in that direction. DR. WOODCOCK: Yes. If we had more resources, we could do more retrospective analyses of these sorts. DR. WOLFE: We've all supported larger budgets for you. (Laughter.) DR. WOODCOCK: You know, it sounds like what we're getting to here is the issue, a different issue. Some risk is clearly unacceptable, we all agree with that, and the issue here is, quantitation of the risks that people are subject to, and Bob DeLap and then Dr. Stolley. DR. DELAP: Well, I have to confess to a certain ambivalence about this. I think it really gets down to the details. If you're talking about one day, I don't think people would get too concerned about one day of lack of treatment. If you're talking about, you know, 12 weeks, that's a different matter. If you're talking about a blood pressure of 105 versus 95 versus 115, obviously, the degree of concern is different. One of the other -- well, there are two fundamental things that really concern me though, here, which is, one is, I do think that it's harder for the patient to really appreciate what they're signing onto when they sign onto one of these trials. It isn't like the migraine situation, where the person has a headache and they know what they've got, and they know what they are up against, and they can fully appreciate what they may be signing onto. Here, it's, of course, it's hard enough for us to figure it out, I guess, but it's I guess impossible for the patient to really fully appreciate what risk they're assuming. And the other thing, in the real world, I think we know how these trials are actually done, with new blood pressure medicines. There are people who have clinical trial units around the Country, and they have groups of patients who participate in trials, and there is an aspect of kind of, in at least some study settings, where people get washed out from their therapy, and then they try the new drug. And then the next new drug comes along and they get washed out from their therapy, and they try the next new drug. So, it's not necessarily, you know, four weeks we're talking about here, but it may be four weeks now and four weeks there months from now, and four more weeks. So, over the course of time, for a particular individual who has signed on with one of these clinical research units, they are going to have quite a number of weeks of being washed out and off of treatment. DR. WOODCOCK: Thank you. Dr. Stolley? DR. STOLLEY: One of the things it says in the description is that, "Candidate drugs are not required to show reductions in mortality or cardiovascular events." And that's the really important question, a public health question, and the question the doctor and the patient want to know is not, Will you lower the number but, Will I live longer? Will there be fewer sequelae? And when the calcium channel blocker controversy arose, nobody knew whether calcium channel blockers actually were going to protect you from stroke, and you would live longer, and with newer, expensive antihypertensives coming on the market, and no drug insurance under Medicare, and you have to take them for the rest of your life, that's a key question. I don't know what you can do about it, but that's the question we need answered. DR. WOODCOCK: Dr. Stolley, I hear you focusing in a couple of these examples, and your original discussion, on really outcomes of medical treatment, which is an even broader issue, and choice within the armamentarium of the proper treatment, and maybe we can discuss this at the end, is that the proper role at the very beginning of drug development when you're really trying to select candidates that work? Or, is this something that we just all need to concentrate on getting done more, and sometimes -- DR. STOLLEY: Of course, that's not the question you ask when you don't even know that the drug has any effect on blood pressure, granted, but that is the ultimately important question -- DR. WOODCOCK: Amen. DR. STOLLEY: And it's the one that the NIH usually winds up answering and funding. DR. WOODCOCK: Dr. Temple. DR. TEMPLE: Which fortunately they're sort of doing in this case with the ALHAD Study. DR. STOLLEY: Oh, good. DR. TEMPLE: Which will be more information we have, than we've had up to now, on whether you can get an answer to those questions, which I don't think is completely obvious. I want to give you a brief anecdote, which I found sobering. This is about a drug called Carvedilol. It's a beta-blocker with some alpha- blocking components that you take twice a day. The reason you take it twice a day is that in a study done in the United States, it was perfectly clear that taking 25 mg, the largest dose you could tolerate once a day, did not lead to adequate control at the end of the day. And we don't approve drugs that don't give you 24-hour control. They had studied over 8,000 people in comparative trials in Europe, because Europe likes comparative trials, that did not include a placebo. And the dose they studied was 25 mg once a day. They showed in those 8,000 patients, no difference between known effective once a day therapies and their drug, which we know from the U.S. study doesn't work. It's a bad incentive to have to show no difference. People can't handle it. You wouldn't expect them to, it's not in their interest to, and you don't want to approve drugs that don't work in hypertension. You wouldn't want a drug that didn't work for the full 24 hours. Doctors don't necessarily measure that. They wouldn't know about it. You have got to be very careful before you start accepting active control trials in settings where you are not quite sure that they can show the difference. Just a reminder. Eight thousand patients and the wrong answer. DR. WOODCOCK: Dr. DeLap and then we'll sum this part up. DR. DELAP: I just wonder if we might further consider the possibility of add-on kinds of designs. I mean, if you are studying these kinds of drugs, there are plenty of patients out there that have not quite adequate control of their blood pressure with one pill, and it's a reasonable thing to think of what to add next, and what people think about taking a person who is on a therapy, their blood pressure is already under some control, maybe not quite enough, and now you're going to add another drug, or a placebo, but you're going to keep them on the drug they're already on. DR. WOLFE: I think that's a good point. I mean, in SYST-EUR because the drug company wanted it that way, they started out with a new calcium channel blocker, it was Isradipine, I think is what it was. DR. TEMPLE: Nitrendipine. DR. WOLFE: Nitrendipine, and then -- DR. TEMPLE: A new old one. DR. WOLFE: -- only added on diuretics and beta-blockers for those people that didn't respond. I think that the reverse kind of study would be very useful, which is what you're talking about, which is to have, for protection and so forth, people taking now as we know effective low-dose diuretic and then seeing whether these additional ones are going to confer any benefit. I think that the place that one might be going in terms of -- even though it's a longer-term outcome with hypertension may be, as Bob was describing before for seizure disorders. I mean, given that we have something known to be effective, which happens to be extremely inexpensive, and we really should be asking, can we outdo that, rather than risking people even for four weeks or six weeks or whatever, with no treatment in the placebo -- DR. TEMPLE: Sid, can you make the design here clear? If you wanted to do a trial in people who don't respond adequately to diuretics, what would you then do? DR. WOLFE: Well, I was saying, take people who do respond to diuretics. I'm knocking the design of the Nitrendipine study. I'm just saying that, if the question is, can we do better than diuretics? You might have people taking diuretics, and see whether you were going to protect them anymore, long-term, by adding another drug. DR. TEMPLE: The trouble is, all hypertension outcome trials don't study a single drug. They study an approach, that starts with one drug, adds another, adds another, until they get to some degree of control. DR. WOODCOCK: Bob, get to the microphone. DR. WOLFE: But the sequence of adding is sort of drug company-driven, as opposed to -- DR. TEMPLE: Well, that's true. DR. WOLFE: That's the problem. DR. WOODCOCK: Okay. Well, I think we've heard a variety of opinions about this issue of hypertension studies. Number one, we heard that there is difficulty in quantitating, or extrapolating the amount of risk people are signing onto, when they would enroll in one of these trials. And explaining it to them. Because we don't really understand that. But we're doing some studies, fortunately, to try to quantitate that. Number two, the ethical issue is added to because, as Bob DeLap said, we are pretty sure that patients wouldn't know what risk they were signing onto, even if they were signing onto some risk. We can't quantitate it, but we're sure, in people who are healthy and taking a preventative, that they really can't imagine what a risk of stroke or MI might be, a remote risk, and so that complicates it. We've heard from Bob Temple that head-to- head trials of certain drugs have been quite uninformative in hypertension, about their failure to be effective in a situation, and that's very concerning, because putting ineffective drugs on the market could be a problem. And we've heard of some alternative trial designs, either short-term for effectiveness, or long- term for outcomes, that people feel could be employed. So, do we have any questions from the floor? We're going to have to move right along. DR. JOHNSON: Quick technical question for Peter and also for Bob Temple. Have you, or can you, or could you, or would you develop a confidence interval for that 1 out of 1000 figure that you calculated? And what you guys are doing, Bob, will it yield confidence intervals? DR. LURIE: So, is it yes, no, yes, and no? (Laughter.) DR. JOHNSON: Have you, and if so, what is it, and if not, will you? DR. LURIE: I didn't do that, but I imagine it could be done. I took, you know, my calculation came out of the estimates of the differential risk at the end of five years in those particular studies, and there are confidence intervals around them, at least at the five-year point, and I suppose one could bring it back down to one year. I'd need some help, but it could be done. DR. TEMPLE: We'll put a confidence interval on it. The number of events is like, even with 100,000 people exposed, the number of events altogether is like to be quite small. DR. WOODCOCK: If any. DR. TEMPLE: So, you know, you won't have more than you have. Major Depression Case Study DR. WOODCOCK: All right. Well, I'd like to move on to the example, if that's okay with the Panelists, and talk about major depression. This is a very interesting example. We've categorized in, "The illness that causes symptoms, but has little likelihood of causing long-term harm." That might be a subject of debate, actually, amongst the Panel. "Existing treatments are effective, but we don't think they have impact, really, on reducing the possibility of long-term harm." So, why don't we go ahead and get people to talk about the use of placebo controls in depression trials? Who would like to start? Dr. Wolfe would like to start. DR. WOLFE: We were able to get a copy of the -- I guess it hasn't been fully published yet -- AHCPR funded a massive study on looking at the effective effectiveness of antidepressants, both old versus new, and new versus placebo, and so forth. And we just want to show just a couple of overheads, just taken directly from that. The main findings were that, very consistently, and we're talking about major depression, as I mentioned earlier in response to the question for someone -- The less severe and the shorter duration the depression is, the much more likelihood that it will have a placebo response. DR. WOODCOCK: Dr. Wolfe, can you talk in the microphone, please? DR. WOLFE: Yes. Right. These were the meta-analyses in this study funded by AHCPR, Agency for Health Care Policy and Research, and I think if you bring it down a little, I think those are the placebo versus -- what is it on the top -- it's the SSRIs. And what you can see is pretty consistently, the point estimates are all to the right; in other words, improved outcome, and maybe half of them have confidence intervals that do not go below 1. The next one is placebos versus tricyclics, I think, is that what it is? Newer antidepressants. Sorry. Versus placebo. And again, one sees that these are all for serious depression, effective. And finally, the last one was the one that concluded that there really wasn't any difference between the old and new drugs, and you can see with remarkable adherence to the one, and you see the studies are just -- and it lists study-by-study, that they reviewed them. Anyway, the point is, from all of this and from what we knew before is that, for major depression at least, antidepressants are very effective. For minor depression, particularly in the situational variety, there is a huge placebo response rate. One of the other things I learned was that, in a meta-analysis of 101 studies, they failed to find out, to show, that a wash-in phase really made any difference. They said that the placebo run-in does not, quote, "One, lower the placebo response rate; two, increase the drug-placebo difference; or three, effect the drug response rate, post-randomization, either inpatients or outpatients, for any antidepressant drug group." As I mentioned earlier, the main point of confusion, I think, in looking at the effect of drugs on depression, is the failure to stratify by severity. One study involved 146 depressed outpatients, who met the criteria for inclusion. Eighty received placebo; 27 Imipramine, and 37 Alprazolam. And what they found was that, the hypothesis was that the shorter the duration, the more of a placebo response rate there would be. And what they found was that, people who were depressed for one year or longer, the placebo effect was 22.5%, whereas people who had not been depressed as long, it was 44.9%. And there is study after study in the literature, that compares either in terms of duration of the depression, or the severity initially, as rated by the Hamilton Scale, or the presence or absence of an acute precipitating event, that there is a huge difference. Their description, which is just three sentences long in this AHCPR Study of depression is, "Detrimental effects on personal productivity" -- and we're talking again about serious, endogenous, DSM-, I guess, now 4, Depression, Major Depression -- "Detrimental effects on personal productivity, interpersonal relationships, and the ability to perform usual daily activities are pervasive. "Studies examining the effects of depression on health-related quality of life, demonstrate decrements that equal or exceed those of patients with chronic medical illnesses, such as diabetes mellitus, or ischemic heart disease." Now, I'll just summarize by saying that, given that we know that for major depression, that we have a large number of antidepressants that work; that there really does not seem to be any important difference between them; that to relegate people, particularly to a wash-out, which despite this meta- analysis seems to still continue, and to another period of time without the drug, is really cruel. And I would strongly disagree with Bob's initial slide, saying that for psychiatric illness, be it psychosis, or depression, that as long as patients were informed, which itself is a problem because of insight, and it's been taken up by the National Advisory Commission on Bioethics, in terms of the vulnerability of such a population to informed consent. But, aside from that issue, this is inflicting a lot of pain for a benefit that is really very questionable. We have not really gained, other than the difference in adverse effects, which obviously can be measured in a comparative study, we haven't gained, really, a lot. You are trading one set of anticholinergic effects with the tricyclics and tetracyclics for a very high incidence of sexual dysfunction and so forth with the SSRIs. So, I think that, whether it is withdrawing someone from an antipsychotic drug in a wash-out period and allowing them to become psychotic, or putting them in a placebo group, and having them become psychotic, or relegating people with serious depression, which as was said, is equal or in excess of those people with diabetes or ischemic heart disease. I don't think that a placebo control design for this kind of serious depression, is indicated. For more mild depression, for a mild depression, the placebo response rate is so high that, from a practical perspective, one would wonder, assuming that these people are not suicidal, because you can be suicidal, even without the other criteria for DSM-4 Depression, I would wonder why people wouldn't be tried on non-pharmacologic therapy at first, rather than doing more clinical trials with mild depression, which according to the review of the AHCPR, result in very confusing kinds of results. DR. WOODCOCK: Dr. Temple? DR. TEMPLE: I think -- I'm really stunned. I think you're saying, it wasn't worth developing any of the newer antidepressants, because they're not more effective than tricyclics. I'm positive there's someone in the room and they can speak to that, but no psychiatrist believes that is remotely true. Attempts to -- you -- it's perfectly true that you get one kind of side effect on one, and one kind on the other -- DR. WOLFE: In terms of effectiveness in treating major depression? DR. TEMPLE: No. No. In terms of effectiveness, they're all the same. That's absolutely right. That's -- DR. WOLFE: Well, that's what I said. DR. TEMPLE: That's very important. DR. WOLFE: And there are different kinds of adverse effects, I'm sure. DR. TEMPLE: That's right. But, you needed to know that the SSRIs were effective. You had to have a study design that would show they're effective. And an equivalence design could not do it, for reasons that I showed you, nobody's contradicting that. DR. WOLFE: Let's assume we're talking about now. We're not talking about five years ago. We now have two, three, four classes that have been developed and at this point, the question is -- which is the case that's presented today -- do we need to keep doing placebo control trials for major depression? That's the one I'm answering no to. DR. TEMPLE: Only -- only if you want more antidepressants. If you don't believe you need any more drugs, then you don't have to do it. DR. WOLFE: If you want more that are no more effective, as is likely to happen, from the ones we have now -- DR. TEMPLE: Right. It's extremely clear that nobody is going to be able to distinguish one antidepressant from another. They've been trying for decades. They all have roughly the same effect, within the limits of these studies to detect them. But they differ enormously in their side effects. And if anybody thinks that SSRIs are the sort of ultimate and optimal therapy, they just haven't talked to -- DR. WOLFE: Well, using your own logic -- DR. TEMPLE: They just haven't talked to the women who -- DR. WOLFE: -- the ability to distinguish between the classes, in terms of adverse effects, would be very clear. I mean, certainly, the nature of, and the degree of adverse effects with SSRIs versus tricyclics or tetracyclics is pretty clear, and you don't need a placebo control trial, if that's what you're postulating, for the next class that comes along. DR. TEMPLE: No. You need to know it's effective. You have to find out it's effective, and you cannot do that without a placebo control trial. DR. WOLFE: I disagree. DR. TEMPLE: So, if you -- you disagree? How do you disagree? DR. WOODCOCK: Wait, wait, wait -- Let's get to the technical issue. Dr. Wolfe thinks, I gather, that it's unethical to withdraw people and wash them out, and expose them to placebo, who have major depression, okay. And Bob thinks -- I'm summarizing everybody -- that it's impossible to determine whether a new agent is effective, unless you do a placebo- controlled trial of it and depression. Is that true? Okay. But they disagree with that -- there's a disagreement about that conclusion here. Dr. Wolfe, you believe it is possible to determine the effectiveness of an antidepressant -- DR. WOLFE: With severe depression. I mean, again, we're not talking about where there's a high placebo response with mild or kind of depression, severe depression -- DR. WOODCOCK: With a head-to-head trial, is that what you're -- wait, let's see what alternative trial do not -- a head-to-head trial versus an active control approved antidepressant. DR. WOLFE: Right. Right. DR. TEMPLE: There's just no track record in those. Those studies are actually severe depression, you know, with suicidal and everything, scares everybody off. No one wants to study those, and we have relatively little experience. However, I can tell you, when inpatient studies are done, you have huge placebo responses, just the same as here. Tom Laughren is here and he can attest to that. All of the people in these trials meet the diagnosis for major depression. They're not absolutely the most extreme, but they're all major depression. A typical response is that you get a change of 10 or 12, I mean, I just showed you six of them -- 10 or 12 in the placebo group, and if you're lucky, on a good day, the drug adds three points more. That's a set-up for not being able to tell anything in an active control trial, and you couldn't. If we were to reach the conclusion, and one could reach that conclusion, that you cannot -- DR. WOODCOCK: Talk into the microphone, Bob. DR. TEMPLE: If you were to reach the conclusion that it is not appropriate to take people off their current therapy, and that the only people you could study, for example, are people who don't respond to other therapy, you could probably work up a drug that way. Whether it would be successful or not is not clear, but I would say, if you were to decide that it's unethical to ask people to endure a period of depression, even though they say it's okay with them, you are going to have trouble developing any new antidepressants, and I would allege that the current crop is not so good, that that's a very, a price we should accept very willingly. DR. WOODCOCK: Dr. Stolley, yes. DR. STOLLEY: I think one of the problems here has to do with the severity of the disease. The thing that nobody wants to study is the thing that requires the most treatment. DR. WOODCOCK: Please speak into the microphone -- DR. STOLLEY: And these are the people who have a major depression; who lose 15 to 20% of their body weight in four to six weeks; who either become highly agitated, or immobile; who have a 15% suicide rate over a six-month period; and if you want some idea of how much they suffer, you can read William Styron's account of his depression, which became a bestseller. These are the people who have to be studied and have to be helped. On the other hand, having them use a placebo would probably be cruel. DR. WOODCOCK: Bob DeLap. DR. DELAP: I think this is a very difficult situation for me to evaluate, too. I'm not a psychiatrist and haven't been involved in using these products. I recognize that they are very valuable products, and they are very helpful to people that have this particular kind of life illness. And I do think there are some issues about informed consent in this situation, that have to be thought through. Part of it is, expending effort to develop kind of more of the same. Of course, one outcome that we would be delighted to see would be a head-to-head trial, we'd be delighted to see a head-to-head trial that would show the new drug was actually superior to the existing drug, but you know, again, I think we are not really expecting to see that, based on historical precedent. We're thinking that the new drugs that come along will turn out to be, at best, kind of as good as the existing drugs, with maybe some advantage in side effects. So, we are left with this efficacy question. But if we did have, again, some trial that showed a new drug was clearly better than a comparator, I think we would accept that. We'd be delighted to see that, in fact. DR. WOODCOCK: All right. Well, this sounds like it is a dilemma. Let me take some questions from the audience, and then we're going to have to close up. Tom Laughren. DR. TOM LAUGHREN: Can I just make a couple of comments? DR. WOODCOCK: Yes, you may, if you speak into the microphone. (Laughter.) DR. LAUGHREN: I think one of the problems pervading the entire discussion is that, it seems that folks from FDA see a different set of trials than folks who are not from the FDA. And I think that's most apparent with the depression example. DR. WOLFE: The ones that don't get published. DR. LAUGHREN: Right. Dr. Wolfe focused on the AHCPR Report, which is based on published studies. At FDA, we see all the studies that fail. And we see all the negative studies. And I think for that reason, we have a much better sense of how much extraordinary variability there is, in placebo response rate, across different trials. And these are all patients who meet diagnostic criteria for major depressive disorder. Now, these are not the kind of, these are not the endogenous patients who have to be hospitalized. And so, it's true, you know, we're not studying probably the patients who most need treatment. But there is, you know, for this population of major depressive disorder patients, there's enormous variability in placebo response. And that's why it's so difficult to interpret a trial which shows no difference. There's another point I just want to follow-up and make and that is, you know, it seems to me that one of the issues here in terms of ethicality is how much clinical trials differ from what's done in clinical practice. And I can tell you that it's not uncommon in clinical practice for patients who have mild to moderate depression, who meet diagnostic criteria for Major Depressive Disorder, to be observed for some period of time, before treatment is initiated. So, what we do in trials, in terms of giving patients placebo, is not so different than what is done in clinical practice. DR. WOLFE: The question I have for you, Tom is, let's assume that we had figured out what the most optimal dose is for a known placebo-proven, existing antidepressant. And we do a randomized control trial of that, with severe, endogenous depression, people above a certain level, with a new drug. I mean, why do you think -- I mean, I agree with Bob, even though I haven't seen the unpublished studies, that there isn't much experience with that, but I mean, one of the reasons for the seminar is to think about encouraging new kinds of experience. Why do you think that it isn't possible, from such a design, to get a useful answer? Assuming that the population is at least somewhat similar to the population that the control drug was originally studied against, in terms of -- DR. LAUGHREN: Even for severely depressed patients, as Bob pointed out, in hospitalized trials, in patients who were hospitalized, you still see a fair amount of variability in placebo response. And it's actually not surprising. DR. WOLFE: This is a randomized study, though, so you sort that out in a randomized study, yes. DR. LAUGHREN: Maybe I didn't understand your question again. DR. WOLFE: No, I'm just saying, why with a drug proven to be effective, you're using it at the right dose, with a population that is similar if not identical to the population in whom it was shown effective against the placebo, you now take that drug, at the right dose, with the same group, kind of people, generally, randomize them to that, versus a new drug. Why would you not, Bob seems emphatic that you would never get any kind of useful information from a study like that. DR. LAUGHREN: It's not predictable. That's the problem. Here's the problem, the same diagnostic criteria are always used. You always use DSM-4 Criteria for Major Depressive Disorder. You apply those criteria in a standard way, despite that fact, I mean, you would hope that you would get a predictable placebo response when you apply those criteria, but you don't. And you know, despite anything you said about being able to predict a placebo response from this based on severity or duration, all of that falls down. I mean, there was a conference on this some time ago, and it basically falls down -- DR. WOODCOCK: But, Tom, are you saying also that you can't reliably distinguish active, good treatment from placebo, is that what you are saying? In a trial. DR. LAUGHREN: You cannot reliably do it. Again, as Bob pointed out, about, you know, a third of trials, maybe 40% of trials, fail to distinguish drugs, standard drugs that we know work, from placebo. DR. WOODCOCK: Even in that setting. DR. LAUGHREN: Even in that setting. And again, it's an issue of those data not being available more widely. But we see them, because they come in as part of an NDA. You don't see them, because they're not published. DR. WOODCOCK: Well, I would like to find some way to make these data available. That's one of my missions. DR. WOLFE: The Conference Center would be glad to receive them, I'm sure. DR. WOODCOCK: I think Peter Lurie and then Jay Siegel. DR. LURIE: Okay, a couple of points. One is, I don't think that we, meaning Sid, I, and Paul have adequately emphasized how much we agree with Dr. Temple's earlier statement that three-armed studies, even if you have to have a placebo, can we at least agree that three-armed studies would be better than two-armed ones? DR. TEMPLE: In depression? DR. LURIE: No, no, in depression, I understand you're talking about in depression. DR. TEMPLE: In depression there, that's what's almost always done. DR. LURIE: Okay, what -- that -- DR. WOODCOCK: Wait, wait, wait, wait. Finish, Dr. Lurie, please. DR. LURIE: I'm just saying, I'm not saying that that is not the case in depression. All I'm saying is that that needs to become much more common in all areas of FDA regulation, particularly as there become vanishingly few conditions for which there is no available therapy, and most of the time, we're talking about me-too drugs that are very likely to be better than placebo, because they are in fact, you know, some minor chemical modification on known effective therapy. The second point is, there agreeably might be times where you have, say, a particular study which doesn't, you know -- it can't show differences, Bob, but when you go back and you look at only the publishes studies, even, on depression, what you see is there have been a large number of studies. And what you're really doing is you're looking at the family of studies, and interpreting them as a group. And that a single, or even two, quote, unquote, false negative studies, if you will, can be interpreted in the context of the totality of clinical experience. The pharmacological data, the toxicological data, animal data, you know, if relevant to the condition, and the large number of clinical trials that are often conducted. So, you have to put the whole packet together, and one trial here or there, you know, probably isn't going to sink or swim the drug. DR. WOODCOCK: All right. Dr. Siegel, then our questioners, and then we're going to have to wind this up. DR. SIEGEL: I just wanted to comment on the issue, the critical importance of an issue Dr. Wolfe mentioned, the similarity of the population, the similarity in the active control trial, to the population in which the active control was initially studied. In a case such as Dr. Laughren mentioned, where active control works sometimes and doesn't work other times, there may be differences in the population that are not related to placebo response rate, a variability we've talked about, but that are related to responsiveness to the drug, that we are unable to identify. In many areas, we don't know the co- variates that account for the fact that people respond or don't. And one of the few areas where we've successfully used active control trials, the area of thrombolytics, we have the advantage of a database generated from tens of thousands of patients in placebo-controlled trials. In fact, when meta-analysis came around, it was discovered and noted probably correctly by many that we were doing placebo-controlled trials of thrombolytics long after we had convincing P-values with several zeroes that they saved lives. Well, obviously, assuming that's true, that shouldn't have been done, but having been done, we now have a massive database that we can look at, and we know things, such as, if you do your study in a country or a community that doesn't have emergency medical services, or that doesn't have patient education such that patients arrive within the first two hours of their heart attack, the drug is not going to make a difference. If you do it in a population enriched, which can happen, for patients with inferior MIs, or MIs with little ST-elevation, the thrombolytic is not going to make a difference. And what that means is that if you were to do an active control trial of a new agent to a known thrombolytic, even if that new agent were completely inactive, and you were to do it in those settings, either unintentionally or potentially intentionally, you could pretty well ensure that you would, by any statistical standard, be able to prove that there aren't important, whatever you set up as your margins, differences between the two therapies, because in fact, the thrombolytic doesn't work. So, this is the harking back to the caution Dr. Temple made, but I think that where we have a massive database, we know that there are inferential problems based on certain co-variates that react or respond. Where we don't have that database, but we see variabilities in response, it doesn't -- I think the important thing to keep in mind is, we don't know why some people don't respond, but we do know in fact that some people don't respond, and some trials don't show responses, and that raises significant concern about the ability to conduct an active control trial. DR. WOODCOCK: But that's a very good point, that as our science advances, that we will have better ability in cases, and we will be driven, and we will try to get that baseline data and those predictors. DR. SIEGEL: It's all genomic. DR. WOODCOCK: Genomics. Oh, jeez. Don't hold your breath. Our last two questioners. DR. SCOTT WOODS: Just to follow up on that point of needing to do better. We may -- my name is Scott Woods. I'm from the Yale Psychiatry Department. We may not be at the point that Dr. Wolfe thinks we may be at depression, while we could say with perhaps 95% confidence that an active control investigation, had it included a placebo, would have been informative. We may not be there yet, but we should try to get there, and think about ways where we might be able to, and the sloppiness that people have talked about that possibly would be encouraged in an active control investigation, there ought to be ways to control that. I mean, we could, for example, just instead, for patient selection. You say, patients might not really have the disease. Well, there would be a way of perhaps controlling for that. A videotape of the patient interview could be submitted, instead of the rating scale, and the rating could be done by someone who wasn't associated with the study at all. So, we ought to try to get there, even if we aren't there now, where we would be able to stop using placebo controls in depression studies. DR. WOODCOCK: Thank you. The last question? DR. JOHNSON: Yes, we've been discussing efficacy, and I'm curious to hear opinions on whether active controls are comparable to placebo-controlled studies, to assess adverse event profiles. DR. WOODCOCK: We need monosyllabic answers, please. DR. WOLFE: A definite maybe. DR. LURIE: A definite maybe. DR. TEMPLE: No, they're not. You don't know whether a given rate is larger than the placebo would have been. DR. WOODCOCK: This is Bob's equivalent of a monosyllable. (Laughter.) DR. TEMPLE: No, no, no. DR. WOODCOCK: Anyone else? Substantial disagreement on this point. All right. Well, thank you very much. I really appreciate the audience's tolerance for this long, and thank you to this Panel for this good discussion. (Whereupon, at 3:45 p.m., the Panel meeting was concluded.) 102