UNITED STATES OF AMERICA
FOOD AND DRUG ADMINISTRATION
CENTER FOR DRUG EVALUATION AND RESEARCH
+ + + + +
ENDOCRINOLOGIC AND METABOLIC DRUGS
ADVISORY COMMITTEE
+ + + + +
MEETING
+ + + + +
WEDNESDAY
SEPTEMBER 25, 2002
+ + + + +
The Advisory Committee met at 8:00 a.m. in the Ballroom of the Hilton Silver Spring, 8727 Colesville Road, Silver Spring, Maryland, Dr. Glenn Braunstein, Chairman, presiding.
PRESENT:
GLENN BRAUNSTEIN, M.D. Chairman
ERIC ABADIE, M.D. Guest
THOMAS T. AOKI, M.D. Consultant
HENRY G. BONE, III, M.D. Consultant
STEVEN R. CUMMINGS, M.D. Guest
KENNETH G. FAULKNER, Ph.D. Guest
MARC C. HOCHBERG, M.D., Ph.D. Guest
MARIE C. GELATO, M.D., Ph.D. Member
DEBORAH GRADY, M.D., M.P.H. Member
SUNDEEP KHOSLA, M.D. Guest
LYNNE L. LEVITSKY, M.D. Member
BARBARA P. LUKERT, M.D. Consultant
ROBERT MARCUS, M.D. Guest
MICHAEL R. McCLUNG, M.D. Guest
PRESENT: (CONT.)
RENE RIZZOLI, M.D. Guest
GIDEON A. RODAN, M.D., Ph.D. Guest
ALLAN R. SAMPSON, Ph.D. Consultant
JANET H. SILVERSTEIN, M.D. Consultant
WILLIAM V. TAMBORLANE, M.D. Member
CHARLES H. TURNER, Ph.D. Guest
NELSON WATTS, M.D. Guest
ERIC COLMAN, M.D. FDA
David Orloff, M.D. FDA
BOB MEYER, M.D. FDA
BOB TEMPLE, M.D. FDA
NANCY WORCESTER, Ph.D. Member, Consumer Rep.
ROBERT ZERBE, M.D. Industry Rep.
KATHLEEN R. REEDY FDA Executive Sec.
C O N T E N T S
Introduction of Advisory Committee 4
Announcements by Ms. Reedy 7
Opening Comments by Dr. David Orloff 9
Presentation by Dr. Colman 15
Presentation by Dr. Abadie 29
Presentation by Dr. Bone 43
Presentation by Dr. Rodan 64
Presentation by Dr. Rizzoli 77
Presentation by Dr. Turner 84
Questions & Discussion 90
Presentation by Dr. Faulkner 106
Presentation by Dr. Hochberg 133
Questions & Discussion 147
Open Public Hearing
Presentation by Dr. Dere 192
Presentation by Dr. Marriott 201
Presentation by Ms. Alina 206
Presentation by Dr. Cummings 216
Questions and Answers and Discussion 281
Final Comments 429
P R O C E E D I N G S
8:07 a.m.
CHAIRMAN BRAUNSTEIN: Welcome to the September 25th, 2002, meeting of the Endocrinologic and Metabolic Drugs Advisory Committee. I'm Glenn Braunstein, Chair. We'll start by going around the room and asking for everybody to make introductions. We'll start with Dr. Marcus.
DR. MARCUS: Good morning. My name is Robert Marcus. I'm Emeritus Professor, at Stanford University. I'm a former member of this Panel, and I'm medical advisor at Eli Lilly and Company.
DR. FAULKNER: Ken Faulkner. I'm currently working at G.E. Medical Systems; also an Associate Adjunct Professor at the University of Wisconsin at Madison.
DR. CUMMINGS: Steve Cummings. I'm a Professor of Medicine Epidemiology and Biostatistics at the University of California at San Francisco.
DR. HOCHBERG: Marc Hochberg. I'm a Professor of Medicine and Epidemiology and Preventive Medicine at the University of Maryland in Baltimore.
DR. TURNER: I'm Charles Turner. I'm a Professor of Orthopedic Surgery and Bioengineering at Indiana University.
DR. RIZZOLI: I'm Rene Rizzoli, Professor of Medicine at the Geneva University of Medicine Hospital in Geneva, Switzerland.
DR. RODAN: I'm Gideon Rodan. I'm the head of Bone Research and Osteoporosis at Merck, and another Adjunct Professor of Pathology at the University of Pennsylvania.
DR. SILVERSTEIN: I'm Janet Silverstein. I'm a Professor in Pediatric Endocrinology at the University of Florida in Gainesville.
DR. ABADIE: I'm Eric Abadie. I'm Director of Therapeutic Evaluation with the French Agency, and I'm also Vice-Chair of the CPMP. The CPMP, for those who don't know, is the licensing body in Europe. And I'm probably here because I was rapporteur for the Osteoporosis Guideline in Europe.
DR. GRADY: Deborah Grady. I'm a Professor of Epidemiology and Biostatistics and of Medicine at the University of California, San Francisco.
DR. TAMBORLANE: I'm Bill Tamborlane, Professor of Pediatrics at the Yale University School of Medicine.
DR. GELATO: I'm Marie Gelato. I'm Professor of Medicine at the State University of New York at Stonybrook.
CHAIRMAN BRAUNSTEIN: Glenn Braunstein, again. I'm Chairman of the Department of Medicine at the Cedars-Sinai Medical Center and Professor of Medicine at UCLA School of Medicine.
MS. REEDY: Kathleen Reedy, Administrator of this Advisory Committee at the Food and Drug Administration.
DR. AOKI: Tom Aoki, Professor of Medicine, Division of Endocrinology, University of California, Davis.
DR. LUKERT: Barbara Lukert, Professor of Medicine, Division of Endocrinology, University of Kansas.
DR. SAMPSON: Allan Sampson, Professor of Statistics, University of Pittsburgh.
DR. LEVITSKY: Lynne Levitsky, Chief of Pediatric Endocrinology at Mass General Hospital, Associate Professor at Harvard Medical School.
DR. ZERBE: I'm Bob Zerbe, Quatrix Pharmaceuticals, and I'm the Industry Representative.
DR. WORCESTER: Nancy Worcester, an Associate Professor of Women's Studies and Continuing Studies, University of Wisconsin, Madison, and I'm the Consumer Representative on this Panel.
DR. BONE: Henry -- oh, I just needed to hit it a couple times. I'm Henry Bone. I'm Director of the Michigan Bone and Mineral Clinic, and Past Member of this Committee.
DR. WATTS: Nelson Watts, Professor of Medicine at the University of Cincinnati.
DR. McCLUNG: Michael McClung, Associate Professor at the Oregon Health Sciences University in Portland.
DR. KHOSLA: I'm Sundeep Khosla, Professor of Medicine at Mayo Medical School and Mayo Clinic, in Rochester.
DR. COLMAN: I'm Eric Colman, a Medical Officer in the Division of Metabolic and Endocrine Drugs at the FDA.
DR. ORLOFF: David Orloff, Director of the Division of Metabolic and Endocrine Drug Products, FDA.
DR. MEYER: Bob Meyer, the Director of Office of Drug Evaluation II at FDA.
DR. TEMPLE: Bob Temple, Director of the Office of Medical Policy at FDA.
CHAIRMAN BRAUNSTEIN: Thank you. We'll now turn the podium over to Kathleen Reedy.
MS. REEDY: This is the acknowledgment related to general matters waivers for the Endocrinologic and Metabolic Drugs Advisory Committee, September 25th, 2002. The following announcement addresses the issue of conflict of interest with respect to this meeting, and is made part of the record to preclude even the appearance of such at this meeting.
The Food and Drug Administration has approved general matters waivers for the following special government employees, which permits them to participate in today's discussions: Dr. Thomas Aoki, Henry Bone, Glenn Braunstein, Deborah Grady, Lynne Levitsky, Barbara Lukert, Allan Sampson, Janet Silverstein and William Tamborlane.
A copy of the waiver statements may be obtained by submitting a written request to the Agency's Freedom of Information Office, Room 12-A-30 of the Parklawn Building. In addition, Drs. Marie Gelato and Nancy Worcester do not have any current financial interests in pharmaceutical companies.
Therefore, they do not require a waiver to participate in today's discussions. The topics of today's meetings are issues of broad applicability. Unlike issues before a committee in which a particular product is discussed, issues of broader applicability involve many industrial sponsors and academic institutions.
The Committee Members and invited guests have been screened for their financial interests, as they may apply to the general topics at hand. Because general topics impact so many institutions, it is not prudent to recite all potential conflicts of interest as they apply to each participant.
FDA acknowledges that there may be potential conflicts of interest, but because of the general nature of the discussion before the Committee, these potential conflicts are mitigated.
We would also like to note that Drs. Robert Zerbe, Gideon Rodan, Kenneth Faulkner and Dr. Robert Marcus are participating in today's meeting as nonvoting industry representatives. As such, they have not been screened for conflicts of interest.
In the event that the discussions involve any other products or firms not already on the Agenda for which FDA participants have a financial interest, the participants' involvement and their exclusion will be noted for the record.
With respect to all other participants, we ask in the interest of fairness that they address any current or previous financial involvement with any firm whose product they may wish to comment upon.
CHAIRMAN BRAUNSTEIN: Thank you. Now, Dr. David Orloff, Director of the Division of Metabolic and Endocrine Drug Products, will give some welcoming opening comments.
DR. ORLO3FF: Good morning. I want to thank everybody in advance for coming. I understand, and I guess everyone knows around the table, that this was intended to be a three-day meeting for the Advisory Committee; has been now foreshortened to one.
And again, we thank you for your understanding. Certainly, in light of busy schedules that's a -- both a sort of unfortunate error and a boon, I suppose. I also want to thank in advance Henry Bone for his invaluable assistance in putting together the Agenda for this meeting.
And I want to recognize and welcome Dr. Braunstein back to the position of Chair of this Committee, which was a position he held back from `91 to `95. So as he said to me earlier, we've taken him out of mothballs, I guess.
But he was -- he served us very well in the past, and we anticipate that it will be repeated here again. Today's meeting is not a typical, I guess, Advisory Committee Meeting, based upon most people's experience. There's no specific drug application that is the subject of the discussion.
Nevertheless, this is an important issue that we'll be discussing with the potential for great impact on the way we and interested sponsors and investigators proceed in the area of research and development of new therapeutics for osteoporosis.
The title, if you will, of today's meeting -- I don't know that it's specifically written anywhere -- is, "Standards of Evidence for Approval of Drugs for Prevention and Treatment of Osteoporosis." And the purpose, generally, is to frame the issues around the possibility or the consideration of alterations in our -- that is, FDA's -- Guidance to industry on the development of osteoporosis drugs.
By our way of thinking, any consideration of change in that regard must be in light of the accumulated epidemiologic data, of the large and increasing body of clinical trial results, of experience with animal models, of advances in our knowledge of bone molecular physiology and of the pathophysiology of osteoporosis, our understanding of the mechanisms of actions of several existing and emerging drug classes, and the validity of intermediate measures of efficacy as predictors of true clinical benefit; thus, the panel of experts in the field that we've convened today.
A few days ago I sent a brief memo to the members of the Committee and the invited guests and consultants, and I'd like to share some of the salient points that I raised in that memo with you now.
As you'll hear from our first speaker, Dr. Eric Colman, in a few moments, current FDA Guidance regarding development of drugs for osteoporosis has proposed a weight-of-evidence approach to the demonstration of acceptable safety and efficacy of new treatments.
The approach begins with preclinical or animal data showing bone efficacy and safety, continues through assessment of bone mineral accrual in humans and of bone microscopic morphology in patients, and culminates -- for all drugs but estrogens -- in randomized placebo-controlled trials assessing efficacy in reducing incident fractures.
This last piece in the puzzle has been required in the final analysis because of concern whether bone mineral density, in conjunction with other studies of markers, can serve as a reliable predictor of bone quality and strength.
As many of you know, in June of this year the National Institutes of Health and the American Society for Bone and Mineral Research convened a meeting to address growing concerns over the appropriateness of placebo-controlled fracture trials, based on ethical constraints against the use of placebo when effective therapies for the target disease already exist.
Participants at that conference did put forward proposals for ways to address the ethical issues around the use of placebos in fracture trials raised by increasing numbers of IRBs and others involved in this field.
Indeed, numerous patients -- papers; excuse me -- have been published since the revisions to the Declaration of Helsinki in 2000, proposing constructs by which to decide on the appropriateness of placebo versus active controls in given instances, with several references included, I believe, in the Committee's package.
Without passing judgment on the approaches or proposals at the NIH meeting, it is I think fair to say that the future does not appear bright for the routine use of placebos in trials examining the effects of new therapies on fracture incidence in the development of osteoporosis drugs.
It is important to note that the NIHA SBMR meeting did include discussions of non-fracture endpoints as valid surrogates for change in fracture risk, not a new topic. Today -- one that will certainly be central to today's discussion.
A central theme of the NIH meeting, though, appeared to be that demonstration of anti-fracture efficacy would be critical to the wide acceptance of a new therapy. Whether and in what circumstances anti-fracture efficacy should be required for approval is a central question we wish to address in today's meeting.
What may be the preferred, practical or scientifically valid or ethically acceptable route to a promotable fracture effect is a subject reserved for another discussion. So in sum, the purpose of today's meeting is not to revisit the ethics of placebo controlled fracture trials.
Rather, we feel that the time is right to turn the discussion to the question of whether we -- meaning the regulatory authorities, pharmaceutical sponsors, investigators, doctors and patients -- should reassess what types of clinical and preclinical information should lead us to accept as safe and effective, new drugs for osteoporosis.
What guidance can we offer drug companies -- that is, FDA -- as to requirements for convincing evidence that fracture benefits will accrue, even if in some circumstances we are unable to put a number on that effect size.
As you can see, today's Agenda will begin with a session as necessary background on the U.S. and European guidance documents currently in use. The second session includes discussions of the validity of animal models as predictors of clinical efficacy and safety of the bone.
The third session will address the validity of BMD, bone mineral density, as a marker of fracture risk based on information from observational studies and the results of therapeutic trials.
We will conclude the morning with the open public hearing, and after lunch, we'll turn to trial design issues with the stage set by a presentation of the implications for osteoporosis trial design of the choice of placebo versus active controls.
Finally, the discussion of issues following the formal presentations will, at least at the start, be directed by consideration of four hypothetical new drugs in development. These are a new bisphosphonate, a new estrogen or estrogen agonist on bone, a new mechanistic class antiresorptive agent, and a new bone anabolic agent.
And I'll have more to say before we start that discussion. So for now, I'll say that's all and turn it back over to Dr. Braunstein. Thank you.
CHAIRMAN BRAUNSTEIN: Thank you, Dr. Orloff. We'll start with the history of the U.S. Guidance by Dr. Colman, and after the first three talks we'll take some time for questions and answers and clarification before going on to the next set of talks.
Dr. Colman.
DR. COLMAN: Good morning. What I plan to do over -- well, what I plan to do is go back one slide now. Okay. What I plan to do over the next 15 minutes or so is discuss three general topics, starting with the regulatory history of estrogens as they relate to osteoporosis, then talk about the non-estrogen compounds, and finish up with some of the highlights of the development of the Agency's Osteoporosis Guidance, which was first issued in 1979.
Before I do that, for those of you who are not familiar with some of the osteoporosis lingo, two abbreviations that you'll frequently see, postmenopausal osteoporosis is usually PMO, and bone mineral density is usually designated BMD. So you'll see that throughout my presentation.
The regulatory history of the estrogens dates quite a bit back in time. In 1942 the FDA approved conjugated estrogens for menopausal symptoms. It was then some three decades later that the DESI procedure, that stands for drug efficacy study implementation, and involved the National Academy of Sciences.
A group of individuals sat around and discussed what the available data on estrogen and osteoporosis were, and they made the rather undefinitive comment that estrogen was probably effective for selective cases of osteoporosis.
From about `90 onward, the labeling for the estrogen products as they relate to osteoporosis, there's been some changes in the language in the labeling. If you were to look at the PDR for some of the estrogens in 1990, you would see this wording: "The mainstays of prevention and management of osteoporosis are estrogen and calcium."
Management, we really weren't clear what that meant. So in 2000 it has been further simplified to read: "It's indicated for the prevention of osteoporosis or prevention of postmenopausal osteoporosis." We removed "management."
Some people took that to mean that treatment was rescinded, when in fact, estrogens have never been indicated or approved for the treatment of osteoporosis. It has always been prevention, and in the past, "management" was also in there.
This slide, this gives you a sense of where I think we stand from a regulatory standpoint with estrogen and osteoporosis. There's no question that estrogens increase bone mineral density. They do that in a dose-dependent manner.
The recent publication of the Women's Health Initiative data clearly suggests that estrogen plus progestin -- in this case it was Premarin -- significantly reduced the risk for osteoporotic fracture, including hip fracture, certainly the largest database confirming a favorable effect of estrogen progestin on osteoporotic risk fracture.
However, there's nothing that is free in this world, and the risk versus the benefit of this compound is under close scrutiny and debate right now, and I'm sure it will continue for some time to come. We don't -- I don't personally -- know where estrogen's role in the treatment or prevention of osteoporosis will fall out in time.
Next month, the NIH will have a meeting to discuss the WHI data. Again, it's -- this is a summary. Estrogen is currently approved for the prevention of PMO, but not the treatment of PMO. I think that'll become a little clearer as we get into this, as to why we chose those words.
Moving on to the non-estrogens, I want to talk a bit about calcitonin, fluoride, the bisphosphonates and the one selective estrogen receptor modulator, SERM, that's approved for osteoporosis; that is, raloxifene or Evista.
Injectable Calcitonin was the first non-estrogen approved for osteoporosis in this country. That was back in 1984. Total body calcium was one of the primary endpoints. The total body -- the total number of subjects taking part in these trials was relatively small, less than 200.
They did have two-year studies. This went to an advisory committee. It was a fairly small committee. There were a couple of members who voiced concern about approving this drug without fracture data. There were three members, however, who felt that the data that were presented were sufficient for approval.
The drug was approved based on total body calcium data, in part. But the message was clear to the company that you really should do a fracture trial. So the company did embark on a Phase IV Fracture Trial shortly after approval. Due to a number of problems, such as a very high dropout rate, recruitment was very slow. The study, in essence, never materialized and never produced usable, viable, accurate data.
Starting with fluoride, this really marks the 1980s and the early `90s as critical periods for the development of osteoporosis drugs, and in turn, for the regulation of those drugs.
Some studies early in the `80s were published that suggested favorable effects of fluoride on bone density and fracture risk in postmenopausal women. Larger studies were done after the initial studies, and one study in particular certainly demonstrated that the fluoride increases bone mineral density of the spine by a dramatic amount.
But in that study the rates of vertebral fracture in the fluoride group versus the placebo group was not significantly different. Furthermore, there was some suggestion that fracture rate in some skeletal sites may actually be higher on fluoride than placebo.
So this really wrangled some thinking. It caused people to question the validity of using an increasing BMD as an automatic indicator of reduction in fracture risk. People have pointed out that fluoride is known to cause mineralization defects in animals.
So this is not a complete surprise, to see this association. Nonetheless, it did set the stage for the next drug that was brought to the FDA, and it had a similar unfortunate correlation. Etidronate was the first intravenous bisphosphonate brought to the FDA seeking indication for osteoporosis in 1991.
This did go to an advisory committee, as well. It was known up front that etidronate could cause osteomalacia in animal models. That was a concern that was certainly sitting in the back of people's minds.
That information, coupled with a rather strange finding in the third year of two American trials, where the BMD and the fracture rates were what you would expect up to the two-year time point, but during the third year while the BMD was maintained, the fracture rate was not maintained and actually was going in the other direction.
There was quite a bit of controversy over how to interpret these data. I don't want to get into that. The point here is to highlight that fluoride and Etidronate caused some people to step back and say, well, how valid is an increase in BMD in terms of predicting a reduction in fracture risk?
If you add animal data to that equation, I think you have a better sense that we certainly knew that fluoride and Etidronate could cause mineralization defects in animals, and that added to the complexity of interpreting these two development programs. But they did have a big effect on drug development and regulation.
Calcitonin nasal spray was approved in 1995, again, based primarily on bone mineral density. There was still no definitive fracture data. There was -- the proof trial was underway, a five-year, 1,000-woman trial.
The Committee Members felt that it was reasonable to approve this formulation without definitive fracture data, as long as this Phase IV study is ongoing. Then along came alendronate in `95, and the development program for Alendronate really took into account what happened with Etidronate and fluoride.
The company conducted three-year fracture trials. It showed in these trials that the lumbar spine BMD increased by a significant amount, and that that increase was associated with a significant reduction in vertebral fracture risk.
So there was no discrepancy here between increases in BMD and reduction in fracture risk. That's what you would expect. Animal data also did not suggest there were any problems with mineralization. So the preclinical data looked good.
Having the fracture data with the BMD that confirmed this association, the company then was able to get a prevention-of-PMO indication by studying early postmenopausal women who were osteopenic for a two-year period, comparing it to placebo, and that was based solely on BMD, although we did have the knowledge that in a older group of women the increases in BMD were associated with a reduction in fracture risk.
So the -- what we learned here was applied to this population, although BMD was the primary endpoint for the prevention indication. And again, the correlation between increases of BMD and decreases in vertebral fracture risk, this again was resurrected as a reasonable surrogate.
Let me quickly just run over -- the only SERM that's approved in this country is Raloxifene, approved in 1997. It was deemed an estrogen from a regulatory standpoint back then. So the company was able to secure a prevention-of-PMO indication based solely on BMD, because it was viewed as a estrogen.
Shortly thereafter this approval they had a long-term fracture study ongoing done primarily to satisfy a European regulatory request. So we shortly learned that, in fact, that the modest increases in BMD did translate into a modest reduction in vertebral fracture risk.
But this was a development plan that differed from the non-estrogens and the estrogens, to some extent. Risedronate was the second bisphosphonate; again, followed the very similar program to that for Alendronate.
This is just to quickly highlight some of the points of the last 60 years. Because of their age, they've been around in regulatory parlance for 60 years. Estrogens and non-estrogens have been treated, regulated, developed along slightly different pathways.
The clinical trials have become enormous. If you'll recall, the Calcitonin trials in the early `80s had about 200 patients. The risedronate database in the mid-'90s had nearly 15,000 patients. BMD was -- took a bit of a beating with fluoride and Etidronate.
I think BMD has again risen up to its proper place as a reasonable surrogate for fracture, as long as animal data support that. You'll note that all of the primary, pivotal, long-term big trials for osteoporosis have been placebo-controlled, although the option to do active control has always been there for them.
And what we have, after years of this investigation, is -- we have a number of drugs that we know reduce the risk for vertebral fracture, and in some cases, nonvertebral fracture over a three-year period.
Let me finish up quickly with some highlights of the development of the FDA's Guidance document, first issued in `79, updated in `84 and `94. You can see from the very beginning in the late `70s, people thought that it was going to take a fair amount of time to study these drugs before they were approved.
This had to do with the relative weakness of the drugs at hand. The technology to measure bone mass was fairly rude and crude and insensitive. So from the beginning these were envisioned to be long-term trials.
This was some of the technique that was available. Single photon absorptiometry allowed you to measure mass in the forearm but not the spine; again, crude techniques. Evaluating fractures. From the beginning fractures have always been important.
The Guidance said that they're highly desirable to get fracture rate. In the same breath it also said, well, it may also be quite difficult to do that because of the large sample size. And again, back here a large sample size was 300 patients.
So they did -- the Guidance did articulate a middle ground, which in essence said, if you can demonstrate that the bone you form with your drug is normal, then a measure of bone mass would be an adequate surrogate for approval.
If, however, there's any evidence that the normal -- the bone is not normal that you form, you're going to have to do a fracture trial to prove that the drug reduces the risk for fracture. In `84, there were a few changes made to the Guidance, nothing too significant.
Prevention studies were now discussed. You could take early postmenopausal women, expose them two years to drug versus placebo, and use BMD as a primary endpoint. Dual photon absorptiometry now allowed measurement of -- major screw-up.
Anyway, the techniques became better at that time to measure lumbar spine BMD, and calcium and Vitamin D supplementation was recommended for all trial participants. And then we move on to the updates in the 1994 Guidance.
And the `94 Guidance, you will see, has incorporated many of the lessons learned from the Etidronate and fluoride experiences. Okay. Here we go. In the `94 issue, there was a clear delineation between what estrogens needed to do versus what non-estrogens needed to do to garner osteoporosis indications.
Preclinical data became very important, again, because of what happened with Etidronate and fluoride, in that you could pick up abnormal bone histology with those studies. DEXA was available. You could do skeletal assessment at various sites. It was a fairly accurate technique, lower radiation.
Fracture assessment at this time was also becoming more refined. The techniques were improving to allow assessment of vertebral fracture, which was the primary endpoint for most trials in the `90s.
This is taken verbatim from the `94 Guidance, and it pretty much -- very clearly lays out what a company needs to do to get a drug approved for the treatment of PMO, approval of treatment of PMO based on three-year clinical data.
The third year, I believe, is a direct result of what happened with Etidronate between the second and the third years. The third year puzzle, I think, led to -- let's go to at least three years to look at this.
You could follow this path if you could demonstrate in animal models that the bone quality is normal, histology's normal, bone strength is normal. Again, that was an issue that was raised with Etidronate and fluoride. If you studied those, you would see problems.
We also had to have at least a positive trend in three-year fracture data. You had to have a subset of patients in the trials that had a bone biopsy, and those biopsies also had to show normal bone.
The BMD had to increase by a statistically and clinically significant amount, and the fracture study must continue to five years or until a definitive benefit is shown. In practice, most of these studies were designed to go to three years, not five years.
This gives you a quick glimpse of what the current regulatory policy has been in the past few years. It differs from what's written in the `94 Guidance. Clearly, estrogens have all along been getting prevention-of-PMO indications from two-year BMD studies.
Treatment of PMO, unlike what is voiced in the `94 Guidance, we would like to see large prospective databases to -- before we would approve an estrogen for a treatment of PMO indication. The SERMs have actually drifted towards the non-estrogens with respect to what is required before approval.
Prevention of PMO can be based primarily on BMD if we have some fracture data that indicates the drug is effective at reducing fracture risk. And as always, a treatment of PMO requires a fracture.
Now, hopefully, at the end of the day, I won't look like this when people continue to ask me what's going to happen with the future Guidance, because I think most of us in the Division don't know exactly which direction things are going to take.
CHAIRMAN BRAUNSTEIN: Thank you, Dr. Colman.
Move on to a discussion of the evolution of the European Guidance. Dr. Abadie.
DR. ABADIE: Okay. Thank you very much and good morning, everybody. It's a great pleasure and a great honor for me to be here. I would like just to say before starting that I'm not a specialist, you know.
I'm only rapporteur of the EU Guideline and the translation, the EU, I would say, translation of the rapporteur is that we work as a coordinator, but we are surrounded by many experts.
And I think that what I'm going to present to you today is more or less the reflection of the EU experts in this field, and I think that, broadly speaking, most of the experts, EU experts, agree with that.
Every good development starts with labeling. There are two types of indications that could be sought by an applicant. The first one is the treatment of osteoporosis, and the second one is the prevention of osteoporosis.
Let's start with the treatment, and that is something which clearly differentiate today the EU and the FDA Guidance. We need -- we need fractures at the initial stage of registration. We need fractures and we need new fractures.
And if we think about the primary endpoints, we would like to see the percentage of patients with new incidents of fractures. That is, the patient sees the sample unique. We are not interested as primary endpoint with the worsening of fracture.
Therefore, the percentage of patients with new vertebral fracture will be the primary endpoint. Another very important issue which didn't appear in the -- I would say the first briefing book that you have, because you have the former EU Guidance in your first briefing book and in the new one, you have the new EU Guidance.
And there was a major difference between those former and new EU Guidances, that we now require both spinal and femoral fracture as co-primary endpoints in a Phase III trial. This Phase III trial could be the only one, and therefore, there will be some stratification between spinal and femoral.
Or you may have two Phase III pivotal trial, the first one being aimed at the spinal and the second one being aimed at the femoral. And I think it makes sense because the population, broadly speaking, between spinal and femoral is totally different.
The indication will be granted only if the antifracture efficacy has been demonstrated at one site, and there is no deleterious effect at the other sites. And at the end of the day the results will be specified in the labeling, for instance, treatment of postmenopausal osteoporosis to reduce spinal vertebral or vertebral fracture. The effect has been also shown in hip fracture, or the effect has not been shown in hip fracture.
So the results of the pivotal trial will be mentioned in the labeling. The prevention of osteoporosis is something which is more straightforward. It's clear that the aim of prevention is to maintain or increase bone mass and strength in order to avoid the occurrence of fracture.
But it's clear that the prevention will come afterwards and after the treatment. We, again, I repeat that we need fracture as the initial stage of registration. That's quite a preclinical package for us. It's probably, probably less sensible than in -- for our U.S. colleague.
Obviously, a robust preclinical package would show, first, no adverse effect on bone quality through histomophometry or histology data and the increase in bone mass and strength. There are probably specialists in the room. I will not dwell on that.
However -- and I think it's important -- the impact of the preclinical package will not be very important on the burden of proof required in fracture study. That is, again, we need fracture in order to get the first approval for an antiosteoporotic drug.
The clinical trials indication treatment, it's relatively straightforward. The minus 2.5 standard deviation on the spine and/or with or without fracture, and obviously there will be stratification. It's the indication treatment, and in the indication prevention.
It's also interesting to see that we have in the last version defined two populations, the first one within the five years after menopause and the other one more than five years after menopause, because more or less I think most of us believe in the transience.
And therefore, we think that there are two populations, within and after five years. And therefore, the risk factors for these two populations are either the BMD plus other risk factors for the population within five years, while we will concentrate on the BMD that is the definition of osteopenia according to WHO; that is, between minus 1 and minus 2.5 in a patient after five years of menopause.
But in Phase II, due to the complexity of the study, we will not require any BMD data before recruiting those patients in Phase II for the prevention trials. The endpoints, very rapidly as I said before, the fracture is the first endpoint in Phase III in the indication treatment.
It would be the incidence of patients with new fracture with the serial x-ray once a year at the minimum, and a BMD would be the first endpoint in Phase II and in Phase III for the indication prevention.
But we think today that BMD is overall -- and I repeat overall -- not an appropriate surrogate for fractures. And if you see these charts, it shows that in fact the relationship between the percentage of reduction in vertebral fracture risk on the Y axis versus percentage increasing BMD on the X axis is not marvelous, and you have all the data on the right-hand side of this slide -- Calcitonin, Raloxifene, Etidronate and Alendronate.
Now, I don't mean that there are not differences between bisphosphonate and the others. It's clear that -- and I think we will have this point to be discussed in the future in -- later on -- but it's clear that the relationship is more convincing for bisphosphonate overall than for the others.
Nevertheless, nevertheless, overall -- and we wrote down the Guideline for the whole class of pharmacological -- pharmacological class of antiosteoporotic guidelines, overall the relationship is not marvelous.
Regarding the endpoint, the biochemical markers, again, it was interesting probably to have those markers, indicators of bone resorption as co-primary endpoint in Phase II with the BMD, while for the stimulator of bone formation, the situation is for us a little bit unclear today.
Criteria of safety, it's important especially for those intermittent -- intermittent -- treatment. With the new bisphosphonate it will be important to have the serum level of PTH and 25 OH determined, because if there is a large increase in PTH with a corresponding hypoglycemia, it will be very important for this type of drug to know that and to monitor that from the beginning.
And finally, connective bone histomorphometry, which will obviously depend on the particular preclinical testing, it's not, I would say, very pleasant for the patient to have a bone biopsy. So it's clear that we will be more demanding on the data about bone biopsy in the patient, depending on the results of the preclinical testing.
So there is here now probably the most important and the most, I would say, innovative in the near future, which is the problems of the placebo and the comparator.
It's clear that today we would like to see a placebo controlled trial, and/or a non inferiority trial versus a comparator of three years duration, with calcium and vitamin D supplementation, which is clearly recommended that I would like to point out here, that we are more or less in the kind of add-on trial since all those patients are already treated by calcium and vitamin D.
And finally, the problem, the major problem, the question about the placebo, is it desirable and is it feasible? About the placebo controlled trials -- and I will not dwell on that because, again, there are people in the room which are much more competent than me -- but let's say that about the placebo, it's the most efficient tool to assess efficacy and safety of a test, few number of patients and easier to interpret when you have a statistically significant difference.
However, the ethical concern is absolutely obvious. And if had a mother or aunt with osteoporosis with a fracture, a prevalent fracture, I wouldn't like her to be treated by the placebo, that's for sure.
Should ethical concern be the same in all population? That is, I think, a very important and interesting issue that we will discuss just later on. About the active control trial, you should know that it's a common requirement.
I would say overall in the EU, we are fond of active controlled trials, which means that we would like to see first the placebo, and I'm not talking here specifically on osteoporosis, but we do like to see a placebo and also an active controlled trial, and the advantage of this active controlled trial is of use.
You can make a relative benefit/risk comparison to other therapeutic strategy, which will give you the possibility to place the compound in the therapeutic armamentarium. However, it's absolutely obvious that there are some drawbacks of noninferiority trial, and I would like to refer you back to a very good document, which is called the ICH Guideline E10, "Choice of Control Group," which clearly explains the situation of noninferiority trial.
It applies a number of assumptions which are difficult to verify, and finally, the choice of the delta is critical, especially with the osteoporosis topic. For the prevention, no major difficulties.
If you use with the prevention the same formulation and the same dose as the treatment, you will have the indication or the company will have the indication, with a placebo control trial with a BMD, as first endpoint in a two-year study.
And if you have a new dose, a new route or a new formulation, that will be a three-arm study with a placebo. Those are formulation which have been shown effective in reducing incidence of fracture, plus the new dose, new route or new formulation, so a three-arm trial.
The first conclusion is that I think the main difference between the FDA and the CPMP rests on the role of the BMD and the preclinical safety, which is probably more important in the U.S. than in the EU, where we will concentrate more on the fracture rate for initial drug registration.
However, it's of use that we will have to cope in the very near future with a difficult problem, which is the design of confirmatory trials in the treatment indication, and there are, at least for me, maybe not two but three alternatives which may not be mutually exclusive, to use placebo in a certain category of patients, to shorten the duration of confirmatory trials and/or to modify the endpoint, and finally also, probably to use the add-on design.
So I will take, if you'll allow me, Mr. Chairman, the last five minutes to try to elaborate on the future. And I will immediately put a disclaimer saying that we have discussed that, but informally, in Europe.
And again, what I will -- going to present you first -- will probably overlap with the future description. That's why I will be extremely brief, but there is something which has not been confirmed at the EU level.
So coming back to the placebo, I think that the placebo could be feasible in certain circumstances. First, the placebo in osteoporosis without fracture; secondly, the placebo in trials of shorter duration; and third, the placebo in add-on trials with patients already treated by an agent of a different class than the test.
I will go directly on the third indent, first to tell you that that is something which may be interesting. We had this example recently in the scientific advice at the EU level. So we won't talk about that especially, for reasons of confidentiality.
But it is an interesting issue that could be debated. The major problem that I see as far as the EU is concerned is the labeling -- is the labeling. And I think in the labeling -- and also for the prescriber in its clinical practice -- we will have to recognize that the patients before taking the test have been treated by another drug, and that may not be totally -- I would say -- the will of the applicant when you submit the dose here.
But I think that as far as we are concerned in the EU, the results of the clinical trials should mimic the clinical practice. The results and the design and the population of patient should mimic the clinical practice.
We have here a table which is interesting and which tried to -- maybe to elaborate a little bit more on the problem of low risk and potential extrapolation from low-risk to high-risk patients.
And here it's interesting because you see that -- in this table which has been borrowed from Pierre Delmas in the recent paper that he published in The Lancet -- it's clear that you see that for Alendronate, for instance, in the low- and high-risk profile, on the right-hand side the relative risk is relatively similar in both populations.
The same holds true for Raloxifene, on the one hand, with the low population and the high population -- high-risk population where you see that the results are relatively consistent. And if we take the vertebral U.S. and the vertebral multinational regarding Risedronate, we have also -- broadly speaking -- some consistent results.
So the question is, could we use placebo in osteoporosis without fracture, knowing that broadly speaking the relative risk may not be totally different from the high risk, which raises a difficulty if you use the placebo.
Before -- I think that it's a matter for debate. I think the use of placebo in low-risk patients is legitimate for these reasons that I've just stated, and also for the fact that the risk, as you saw, is low, that the patients, at least in Europe, are not systematically treated when they have only an osteoporosis based on BMD, and finally, that they have to sign an informed consent under the GCP recommendation, which are more today recommendation, which has an obligation.
And so for these three reasons, I think that at the very least we could discuss that, so there will be pros the solution, which would tell you -- or the extrapolation -- which would tell you that the relative risk is broadly similar between high and low risk, so we could extrapolate.
But there are cons who could say, the relative risk is the same or nearly the same, but the absolute magnitude of treatment effect -- that is, the number needed to treat -- is totally different between both situations -- that's for sure -- since the baseline is different.
And so the question that we should raise is, is regulatory extrapolation of low risk to high risk possible? Now, I will skip very rapidly the concept of sustained versus unsustained efficacy. For the Risedronate, you know that between the first year and between the years thereafter there are no major difference, again.
In the vertebral multinational, in the vertebral U.S., for the PTH we have exactly the same profile, not very different from the beginning until the end of the trial. For the Raloxifene we have also exactly the same picture.
For the strontium ranelate, which is a compound which has not been submit today, either in the EU or in the U.S., but which will be submitted in the future, the problem is similar. The issue is similar. The first versus the third, you have approximately the same results.
So my first conclusion is, is it possible again to shorten the duration of confirmatory trial, but the ethical concern may persist with respect to placebo? And secondly, it is possible or it could be possible to extrapolate from low to high risk.
If we think that the regulatory extrapolation is not possible, therefore, we should go along with non-inferiority trial in high-risk patients, which would remain the only option. And therefore, we have tried to work on this alternative scenario, but it's clear that we came out with a choice of data of 20 percent, with a realistic sample size.
If you want to shorten or to narrow the data to ten percent, we would have some sample size which would be probably unrealistic, but colleagues dealing with the statistics in the metallurgy would probably confirm that.
And so -- it will be my last slide -- if the extrapolation is permitted. It could be -- if it's the placebo control trials in low-risk patients with the endpoint vertebral and hip. And if no other, it's the noninferiority trial with the vertebral or the hip -- but let's say the vertebral -- as the first endpoint. Or an in-between solution, which could be the placebo control trial in low-risk patients, and an active control trial in high-risk patient with an endpoint, which could potentially be BMD, which would probably ease the realization and the achievement of those trials.
So that is something that I think will be the topic for the future discussion. Thank you very much.
CHAIRMAN BRAUNSTEIN: Thank you. Our next speaker is Dr. Henry Bone, who will talk about the rationale and durability of the U.S. Guidance.
DR. BONE: Thank you, Dr. Braunstein. I guess one of the reasons I was asked to talk about this is because I can actually remember when we went through the previous drafts of these things, as it -- I guess that's an advantage of advancing age. Hope it's correlated with some wisdom or insight.
I'm going to talk a little bit about what the reasoning is. I think Dr. Colman has given an excellent resume of the history, and a lot of the rationale was incorporated into that history. I'm going to try to emphasize a few points that are especially pertinent to today's discussion.
The first thing I would like to do is remind everyone that there are a wide spectrum of disorders that could be considered osteoporosis. On this slide we see the ones that have been recognized, by and large, as indications of proof by major regulatory agencies.
And by far, as you see, most of the emphasis has been on postmenopausal osteoporosis. This is where the Guidance documents have been most extensively developed.
We also have had some registrations for glucocorticosteroid exposure, and there's a disorder that's been described or an indication of male osteoporosis, which obviously is just osteoporosis in men and doesn't really specify any particular pathophysiology.
Numerically, as well, the postmenopausal osteoporosis is of course by far the most prevalent. But one of the important points to remember here is that there is a very wide spectrum of severity within the scope of the term "postmenopausal osteoporosis."
This ranges from women with low bone density to women who are in very impaired health as a result of osteoporosis. And this -- while this spectrum has common features of pathophysiology, and as Dr. Abadie has said, a fairly consistent response to certain interventions, the spectrum of the disease clinically is extremely wide.
There are some other examples -- which in this list is not comprehensive -- of osteoporosis, mainly ones we might think of as accelerated osteoporoses, which are not being currently addressed by sponsors or regulatory authorities to any great degree, but which have some certainly common features with regard to histology, fragility and perhaps even interventions.
There are three main Guidance documents, and the Committee Members I think have received copies of all these. The earliest to be developed, as Dr. Colman has described, was the FDA Guidance, which has gone through three iterations now.
There was also a WHO Working Group Guidance which has, of course, no regulatory authority, but which has been published as what I hope would be something more than an academic exercise, at least from the standpoint of providing a rationale for an approach to this.
And of course, there's the EU CPMP Guidance that Dr. Abadie has just described. There are a number of similarities between these documents. The main differences involve the role of bone density as opposed to direct assessment of the effect on fracture for the initial registration.
Now, Dr. Colman has told you about some of the experiences leading to the U.S. Guidelines revision in 1993 and 1994. Prior to that revision the Guidance had said that if bone quality remains normal, then increasing bone density is adequate indication of a favorable therapeutic effect, and this is simply applying the laws of physics.
It's clear: the increase in the mass of a structure, as long as its architecture and material properties are not altered, will increase its strength. This is not an opinion. It's an inevitable fact of engineering and physics.
And I don't think anyone thought that the laws of physics had been revoked when the Guidance was revised. But there was some recent experience in -- as has been mentioned with drugs that appeared to increase the bone mass, but didn't necessarily appear to decrease fracture rates or to produce a sustained improvement.
These two drugs, the two drugs that were a particular concern, were fluoride and Etidronate. Both of these drugs were well known for a long time to cause histologic and biomechanical abnormalities in preclinical studies.
Neither of these qualified under the Guidance, the previous Guidance and certainly not under the present Guidance, for registration under -- on a BMD endpoint full stop.
What I think the major effect, however, of this recent experience was on the revision of the Guidance that we carried out in 1993 and 1994 was that the Guidance became much more specific about the requirements for preclinical testing. so that this wasn't a fuzzy gray area vaguely referring to normal bone quality, but the requirements were enumerated for evaluating this.
We also had the failed trial with subcutaneous salmon Calcitonin. So the issues that were being addressed at that point were the meaning of bone quality in -- as it affected the relationship between mass and strength.
What emerged was a document that says that -- it doesn't actually underline or use the word "robust," but that's the point -- robust preclinical testing can identify drugs that cause a disparity between mass and strength.
However, at the time of the writing of that Guidance it was -- we had relatively little actual experience with drugs that had been successful. A lot of our experience was with identifying the problems.
So it was considered that a "belted braces" approach was probably a prudent idea. And the Guidance said that drugs that did not harm bone quality with a thorough preclinical testing could be approved, based on a primary BMD endpoint in the Phase III trials provided that an ongoing fracture endpoint trial showed a confirmatory trend, as Dr. Colman has told you, and these trials had to be completed.
It put the FDA in a different position from that which they had encountered with the salmon Calcitonin subcutaneous situation. In that Phase IV trial, there was really no way of being sure that the trial would be successful as a trial, irrespective of the outcome of the evaluation.
Whereas, in this situation if it -- obviously, if a trial is fully enrolled and it's far along, enough along for an interim analysis -- the agency could be quite confident that it -- that this sort of Phase 3.5 trial would be successful in at least adequately evaluating the effects.
It was clear that drugs that caused abnormalities in preclinical testing could not be evaluated on this basis, but if the sponsor did carry through a trial -- for example, as we've seen with anabolic agents now -- that a fracture rate was the only acceptable endpoint.
One of the issues that was a little muddy, I think, in some people's minds that -- when you go back and look at the transcripts and remember the discussion in the previous Guidance Advisory Hearing -- was that model systems actually have several purposes, and we're going to hear about model systems from experts on model systems and that.
But -- and it's important not to confound the points that are made by different kinds of investigations using model systems. You can model a disease and its response to therapy. You can use models to detect specific adverse effects.
You might very well look at model systems to detect a mineralization defect, for example, in a system that might not be particularly informative about efficacy of a drug or pathophysiology of a drug. So it's a question of which model system answers which question.
And I think it's extremely important to understand the ability of our model systems to detect effects of drugs that might alter the relationship between mass and strength. You certainly can use model systems to evaluate specific pharmacokinetic or pharmacodynamic phenomena, but of course, we're all familiar with species variations and so on in this area. So this is a general issue in drug development at an early stage.
Again, the point that emerged was that the preclinical testing was generally probably reliable, but it was felt that -- at the time of the development of the Guidance -- that these results required clinical confirmation.
And I think that's -- at some level of evidence that's always the case. We would never approve a drug for human use without, you know, a certain amount of clinical information.
With respect to the specific issue that pertains today, the role of preclinical testing is complimentary to toxicology, and in fact, it could be considered a highly specialized form of toxicology.
The ability of the preclinical testing to identify situations in which a problematic divergence between bone mass and bone strength might occur is crucial. The Guidance is quite specific about -- and thorough about -- the studies that are required to evaluate the architecture, mass and strength of the bone, and this will be discussed.
More limited requirements were indicated for estrogen because of the physiologic nature of the agents and because the experience prior to the Guidance had been supportive of the safety from a skeletal standpoint, and the efficacy of long-term treatment with estrogen.
We do not have specifications in the Guidance -- and this might be a point for discussion -- about when is a SERM an estrogen in the bone. For example, what is the skeletal definition of an estrogen? I think this is -- will be an important point because as Dr. Colman mentioned earlier, Raloxifene was evaluated as an estrogen. We've been in a little bit different mode subsequently.
There's some back and forth on this point, and if we had a very clear understanding of when the appropriate evaluation would -- in the clinic would resemble that for an estrogen or when it might not -- that would be extremely helpful in terms of future clarification.
But the primary objective, clearly, is to make sure that we don't have some undermining of the effect. There was an extension of what had been a two-year observation period to three years. And there were really two reasons for this, two specific reasons.
The one -- the first reason had to do with the bone mass effect of a drug. The subcutaneous Calcitonin trial was two years long, and the numerical, total body calcium estimate at 24 months was lower than it was at 18 months.
And while this was not a statistically significant decline, it was pretty obvious on all the graphs. The total effect was actually small in the first place. The increase in total body calcium by the neutron activation analysis was only about two percent.
There was concern that the biomechanical regulation of bone mass might actually result in a re-equilibration at the original bone mass. This is a plausible concept, particularly at the time, and that what we -- there was concern that what was being seen was just a transient antiresorptive effect that then would be re-equilibrated back to baseline by the -- what's now known as -- by some of us as the mechanistat, following Harold Frost's terminology.
And the other reason, as mentioned, for the three-year observation period on fractures was a concern that if you had a drug that had an antiresorptive effect that might transiently increase bone strength and bone mass by its transient effects on bone remodeling, but nevertheless have some adverse effect, such as a mineralization defect, that this might not be detected in the early phases of the trial, so that the on-treatment observation period was felt to be -- needed to be at least three years.
At that time there was no discussion about whether the placebo phase of that study needed to be quite that long, however. I think the emerging concept was reminiscent of a previous approach to arms control, which was "trust but verify."
It was considered important to confirm the qualitative effects in humans by evaluating the fracture rate. This allowed just -- us to determine whether both vertebral and nonvertebral fracture rates were improved, and it helped to support specific claims.
Part of this has to do with initial registration, if the -- when we look at the minimum requirements for initial registration -- and part of this has to do with: what do doctors actually have to know as a practical matter about their drug?
If we have drugs that are registered with less information than practicing physicians actually need, then we're going to have to get that information with additional studies, and those claims that result from those studies will still have to be registered in order for the information to be disseminated in the marketplace.
One of the questions that has come up repeatedly is whether fracture studies need to be repeated for every single additional indication. If our major concern is the possibility that a drug might induce a discrepancy between bone quality -- or between bone strength and bone mass -- is it reasonable or plausible to suppose that this is going to be different in every sub-indication?
In other words, if the drug doesn't cause a qualitative abnormality in women, is it likely to do so in men? Is there likely to be a qualitative abnormality caused by the drug in steroid-induced osteoporosis, if it's not in postmenopausal osteoporosis? And this has been, let's say, a thorny point in the evaluation of drugs until now.
One of the points that I would like to personally suggest we take into account is how we evaluate multiple endpoints or anatomical sites. Is it really necessary to do multiplicity adjustments, for example, when we are showing that the second or third or fourth anatomical site behaves the same way as the first one that we tested? Or does it make more sense to take into account the prior supportive information of antifracture efficacy, allowing us to reduce the sample size required to achieve that information?
When we contrast the WHO, FDA and CPMP Guidances, we see that there are similar preclinical testing recommendations.
They're really quite consistent, and the Phase II requirements are quite consistent, as well, for the indication of osteoporosis treatment, with the use of biochemical markers being taken into account for dose finding and mechanistic studies, but a consistent requirement for one-year bone density studies for the primary endpoint for Phase II-B.
In other words, we're going to Phase III, based on drugs that have had a full year of bone density evaluation. Efforts to short circuit this have come to grief in most cases in drug development.
The main differences are that the WHO Guidance would recommend registration with no fracture trial at all, if the preclinical testing is robust and satisfactory. The CPMP, as you have heard, requires definitive anti-fracture efficacy for initial registration. And as I mentioned and as Dr. Colman has mentioned, the FDA has required confirmatory fracture information, but not necessarily definitive efficacy at the time of initial registration -- taking into account that during the review period most of those trials that have had the interim analysis of the biopsy will be completed.
This slide is another way of summarizing the same point. Mind you, preclinical testing is done, as I said, in much the same manner as toxicology. In fact, it's really properly regarded as a part of the toxicology program, in my opinion.
And this means that excess dosing is required. Five times the expected clinical dose or its equivalent is specified in the U.S. Guidance, and the European Guidance is a bit less specific, but generally consistent with this.
If there are no problems, as I mentioned, the approaches in the clinical trials on the left are permitted in the U.S. -- well, are possible. In yellow, I've highlighted the current U.S. approach, amongst the other three.
Now, what happens if you have a drug that causes a histological abnormality? We know, for example, that parathyroid hormone changes the histological appearance of bone if you give five times more than the therapeutic dose, and so will many other drugs that might be considered as we go forward.
This certainly would, I think, diminish confidence in the bone density endpoint to the -- and under the current Guidance that is not an option. If so, if a company has a drug or a sponsor has a drug that does produce abnormalities under this testing scheme at the high dose, they can do a fracture trial, or they can quit, under the current situation. We haven't really thought through or developed the experience with some of these drugs yet, to see if there's another way.
Now, when the Guidance -- the current Guidance was developed -- there were fewer therapeutic options, and they certainly weren't as thoroughly tested as the ones that we have available now.
We'd had the recent experiences that have been mentioned, and we've had relatively little experience with everything working very well. We've had more recent drugs tested where the premise of the Guidance has been validated repeatedly in an affirmative way, rather than just showing that preclinical testing could detect problems.
There have been some changes in the last ten years, and I'd like to just review how a few of those might interact -- might verge on today's discussion. We certainly have a lot more experience with drugs in classes that were not ones we'd used extensively at the time of those previous deliberations.
We have had advances in the technology of evaluating bone density and in the ability to evaluate very subtle fractures. And as you will learn, some of what we call fractures are deformities that don't have a clinical correlate, but give very similar information to those -- to clinically apparent fractures.
We've had additional experience in relating the preclinical and clinical measurements, and we have much more information now about what the risk estimates for events are in the clinical trials. There's been an interaction between the Guidance documents, as Eric Colman mentioned, and I think this has had important consequences.
Alendronate and Raloxifene were both registered under the Guidance, pretty much as it was written. But the subsequent registration packages have been designed to simultaneously file in both the U.S. and Europe, or more or less simultaneously file in both the U.S. and Europe.
So in the instances where the U.S. might not have rigidly required fracture data and the CPMP has required those data, the -- obviously, the sponsors have applied under the -- with the strictest interpretation in order to have the project done as expeditiously as possible.
So a lot of the change in what we have been seeing in development plans has been actually this idea of making essentially the same submission to both regulatory agencies.
Clinically, we've had some evolution, as well. We now have several drugs approved that are about 30 to 50 percent effective in reducing the risk of -- the relative risk of fracture. Fracture rate reduction has become the clinical outcome that is widely accepted as an efficacy measure.
This is what the people who do efficacy reviews, this is what the people who do evidence-based medicine analyses use as the target or the indicator of how well drugs work. And this is considered to be the closest to the practical, clinical implications of drugs for effectiveness measurements.
The prevailing standard of care is no care at all. The vast majority of patients in the United States, Canada, U.K. and Europe don't -- even those who could be identified by bone densitometry if the patient's had bone densitometry -- do not take even calcium and Vitamin D.
And we have seen several studies indicating that even after clinically serious fractures, a tiny fraction of patients actually receive pharmacotherapy. Even in those situations the majority of patients probably don't even receive calcium or Vitamin D supplements.
So the standard of care has changed in a way that I think is minimally -- and it's been a source of disappointment, I would say, but it's a fact -- the standard of care for the great majority of patients that's prevailing is no care at all.
As we go forward in the development of new drugs for osteoporosis, we want to look at novel mechanisms -- particularly those that might be capable of increasing bone mass substantially. We are looking at alternative regimens which will improve convenience and adherence, and make treatment more attractive for physicians and patients.
And we'll undoubtedly be wanting to look at combination therapy with complementary mechanisms. Now, a point -- as it was raised earlier -- was the question of the add-on trial. We're really doing add-on trials now.
And as I'll mention in a moment, it's a little bit of a misnomer when we refer to our present approach as a true placebo. But one of the problems that we have if we do add-on trials is, that if we are using an antiresorptive agent and superimpose it on another antiresorptive agent that is potent, we may very well undermine the efficacy.
There is a theoretical concern about the possible adverse affects of such a combination. It might be worse than either one alone, and this is a problem. So the -- we should be -- we have to be a little careful about talking about add-on therapy here, because of the characteristics of our benchmark therapies that are in general use.
We want to limit the risk to participants for participating in studies, obviously. There are several kinds of risks to participants. And we want to keep the development costs and time in a range that does not prohibit development.
Now, an issue that's been mentioned already a couple of times -- and will be undoubtedly a part of the background, even if it's not the foreground, if I can put it that way -- for today's discussion, is the question of the so-called placebo controlled trials with fracture endpoints.
Please keep in mind that in essentially all trials that anybody's going to be discussing, background calcium and Vitamin D are included for all subjects, regardless of their treatment assignment.
When we refer to placebo in our parlance, we're talking about a placebo tablet being used or a placebo infusion or a placebo injection or a placebo nasal spray being used to mask the treatment assignment that that -- for the active patients.
There's a current view that such trials are now acceptable in patients with relatively low risk, but not in patients at high risk, that is to say, patients with prior fracture experience, and you'll hear more about this. This has a lot of implications for trial design.
As we then want to look at the reevaluation of endpoints, we want to ask the questions, what do we need to know and when do we need to know it? What do regulators need to register a drug as safe and effective, and what do physicians need to know in order to make good clinical decisions, bearing in mind that the FDA is responsible for both the initial registration and subsequent claims? -- this was a distinction that was made earlier in the charge to the Committee -- and recalling that if less information is required early in the course of development, we may need to do more later.
We're specifically going to want to look at the relationship between preclinical testing and registration requirements, clinical trial endpoints to look at both these kinds of problems, and the question of how we analyze the data, and particularly this question of multiple indications.
Dr. Abadie has talked about the spine and hip both being evaluated, and I think this is something that will come up in our discussions as we go along. With that, I'd like to conclude and I believe, Dr. Braunstein, you're going to have discussion now or --
CHAIRMAN BRAUNSTEIN: No. Actually, in order to try to stay to the schedule I think we'll just move ahead with some of the additional talks. And either later on this morning, if we have time, or early in the afternoon we'll have time to question the speakers.
Our next speaker is Dr. Rodan, who will be discussing the preclinical models of drug efficacy and skeletal toxicity.
DR. RODAN: Dr. Braunstein, esteemed guests, members of the Panel, thank you for the opportunity and I would like to thank my colleagues at Merck for their input and to Dr. Dave Thompson at Pfizer.
My task is to discuss the contribution of preclinical studies to the evaluation of osteoporosis therapy. The use of animal testing is a cornerstone of drug development, especially for establishing the safety of future therapy.
It's not perfect, but it's the only thing -- the best thing we have, and it works most of the time. This can be expanded to establishing the safety of therapeutic agents to bone, as already mentioned here. We have now excellent models for postmenopausal osteoporosis in various species, and can get an idea about future efficacy of agents in humans by evaluating their effect in those models.
And last but not least, preclinical studies. And that's probably the major, potentially the only way, to study the mechanism of agents, not only for their pharmacodynamic effect, but also for their potential adverse effects, because this can provide shortcuts to studying the side effects and their potential impact on safety in general and in bone, as well.
As already mentioned, this has been very well covered over the last few talks. The change in the Guidelines in 1994 was driven primarily by the experience with Etidronate, which in the third year did not decrease fractures.
And this was a primary motive at that time -- I was at that meeting -- to extend the prior two-year requirement to a third year, and the experience with fluoride, which although it increased BMD, it did not decrease fracture incidents and may have increased it.
And I'll illustrate how preclinical studies actually, with relatively little costs compared to clinical trials, could have preempted the experience which was obtained in the clinical studies. So let's start with Etidronate.
With Etidronate, actually, mechanistically it was known that there is a risk of osteomalacia. And it is preclinically a very good way to study osteomalacia.
The risk was known from preclinical studies which were conducted, actually, by the sponsors, in dogs, and which showed that Etidronate treatment, albeit at very high doses, produced fractures in dogs.
And other studies, also at a multiple of the therapeutic dose, impaired fracture healing in dogs, and this was the background when we studied Alendronate and we knew that osteomalacia is the problem.
And osteomalacia can be very rapidly studied in animals by looking at the accumulation of nonmineralized osteoid in the growth of a long bone. This was described by a Swiss gentleman maybe 30 years ago.
And this is a bone. In blue here you see bone and in red you see nonmineralized bone, and today there are very good ways to measure the quantity of bone in blue, and of nonmineralized bone in red. This can be automated, actually.
And when you look at the effect of Alendronate you can quantify how much bone is retained. This is due to inhibition of bone destruction or resorption, and how much osteoid is present here in red due to a matrix which has to be mineralized. If you have osteomalacia, you have more of the red stuff.
And you can see this with Etidronate and you can quantify that. So this is shown here. You can do a dose response, you increase the doses of the drug to real multiples of the therapeutic dose, and you measure the blue and you measure the red.
The blue is a measure of efficacy because you accumulate bone, and the red is a measure of toxicity because you accumulate nonmineralized bone. And all the bisphosphonates actually inhibit mineralization and keep the accumulation of nonmineralized bone approximately at the same dose, which is about six milligrams phosphorus per kilogram, and so does Etidronate.
The problem with Etidronate is that this happens at a dose which is very close to the effective dose in this model because of the low efficacy of this drug. The new amino bisphosphonates have a much larger window between efficacy and safety risk. This is shown here.
So with Etidronate you get exactly the same place, accumulation of osteoid, but you get also approximately at the same place, accumulation of bone, so that this exaggerates sort of the problem by showing that the window is very narrow.
So this type of study, when you know the mechanism of action of a side effect, this study takes only about ten days and it's recommended for any agent that has the potential of causing mineralization defects. So understanding of the mechanism of the side effect can be approached head on in preclinical studies, as an example.
Now, you can detect this in a less specific manner if you look at the relationship between the amount of bone, which is measured here by bone mineral density, the way it's measured in the clinic, and the strength of that bone.
These are very simple mechanical principles that are going to be expanded on by Dr. Turner, but it's almost intuitively obvious that if the bone is normal, when you have more bone, you should have a stronger amount of bone.
But this is for the same -- this is an engineered, fixed volume which contains more bone, so as the amount is larger, the strength is larger. And this follows a certain curve which was scientifically shown to be the third power of the amount of material, based on engineering principles.
And this study can be conducted, actually, with any kind of future therapy to see if the accumulation of bone, which is the purpose of many therapies, is actually going hand in hand with the increased strength of that bone. Because if the accumulated bone is not normal, then the strength should not increase in proportion.
Then on the background of etidronate and fluoride -- when alendronate was developed, we did these studies -- I think this was the first time it was done for drug development -- and did it in many, many species for many periods of time and found that, indeed, when you retain bone as a result of bisphosphonate treatment, the retained bone generates increased strength, commensurate with the amount of bone retained.
Now, to show that this actually works, I'll show you that with fluoride this is not the case. It was actually reported way before, in 1987 by Mosekilde, that the strength did not increase with bone mass. And much later it was shown that this is due to a defect in mineralization.
But this can actually be detected with the type of study I have illustrated, and this is shown here. We did a head to head comparison between the effect of fluoride and the effect of Alendronate on mini pig bone, and with Alendronate, as previously shown, you see the expected increase.
However, with fluoride you get an increase in the pigs in the amount of bone, measured here histologically as bone volume, but you do not get the expected increase in bone strength. This should have gone in a line which is parallel to this one.
But, actually, the largest increases in bone mass were those which had the largest proportional decreases in bone strength. And when we measured the fluoride content of these bones, the strength decreased with fluoride content, which actually illustrates in an animal study that increases in bone which are filled with fluoride may not lead to increased strength, and therefore, would project a lack of proportionality between BMD and fracture risk in these patients.
So the preclinical models for safety -- and I agree with Dr. Bone that this could be looked at in an extension of the safety study. By doing bone measurements; histology, which gives you quite a bit of information on the structure of the bone; and measurement of strength, as illustrated, which Dr. Turner will expand on, in animal models with multiples of the dose, like five times because this is a toxicological study, one can detect -- one actually did detect the deleterious effect of Etidronate and fluoride, and this could be a sensitive method to evaluate the safety of future therapy.
Now, the studies I illustrated were of relatively long duration. The pigs were studied for a year. The monkeys were studied for two years and rats were studied for a year or so. In order to realize some savings one could add these measurement of bone properties to regular toxicology studies.
There are toxicology studies, for example, for carcinogenicity, which have to be conducted for two years, and these could include the toxicology for bone. And there are other studies which may take six months or so. And so this could be incorporated as a toxicology study for agents which are aimed at the treatment of bone.
Now, moving to efficacy studies, we have really very good models of estrogen deficiency bone loss, and this loss occurs more rapidly in cancellous bone, which is the interior of the bone, than in the envelope of the bone, which is called cortical, and occurs actually in most mammals.
It's very pronounced in humans after menopause. It's seen in rodents. It's seen in primates. It's also seen in other species, but because it's driven by a reproductive hormone, usually estrogen, it's dependent on the reproductive cycle.
And some species where the cycle is seasonal show this phenomenon in a seasonal manner, and one can pay attention to that. For example, sheep, dogs and more recently, rabbits, also have estrogen deficiency bone loss in a seasonal pattern and could be used to evaluate estrogen deficiency bone loss and its prevention or correction by agents which are developed for osteoporosis.
Now, agents that increase bone mineral density and bone strength, as illustrated in those models, and have been tested clinically have been shown to reduce fractures. And now we have quite a database: it's bisphosphonates; estrogen, just published a few weeks ago for the first time in a placebo controlled prospective study; the SERMs do it; PTH is not yet approved, but they did it as well.
However, the quantitative relationship between the effects in animal species and in humans can probably not be easily projected. So for the quantitative relationship the human data are most likely necessary.
Now, I'll just summarize. With certain principles which -- the `94 Guidelines are extremely specific in outlining what studies to do for what duration and so on, and maybe a committee can work on that.
But I'll just list the principles that can be followed for providing the maximum information from preclinical studies for the development of antiosteoporotic therapy. The first one is to use adult animals to eliminate the confounding effects of growth.
I think that any species that loses bone as a result of estrogen deficiency can be used, because the science that has accrued over the last ten years or so shows that there's really mechanistically no significant difference in the way this happens.
One should deal with several parameters, histology, densitometry, now micro CTs are being developed, mechanical testing, biochemical markers. And there should be internal consistency between the outcomes of these different measures in order to provide further confidence in their validity.
And also mentioned by Dr. Abadie, several doses should be used. Excuse me. For prevention, one should have prevention studies. For treatment, one should document in their own species, and this can be done, that you can actually reverse or restore bone that had been lost, and follow what happens in both cases after cessation of therapy, because this may be predictive of what might be happening in humans.
For mechanism study, I mentioned already that they provide important insights for both understanding the safety of agents when the side effects actually can be identified. And in many cases this is today possible.
One can identify from histology the type of bone. Dr. Bone mentioned PTH effects on histology. We looked for lamellar bone, which can be easily identified. You can look for ectopic ossification for bone formation agents.
And efficacy. There's actually, at the tissue level, no difference in the way the inhibitors of resorption work. From estrogens to amino bisphosphonate and anything else, at the tissue level, they work through the same mechanism, suppression of bone turnover and they reverse the negative bone balance.
So suppression of bone turnover leads to a positive bone balance and accumulation of bone. So from the tissue-level mechanism point of view, we do not detect differences between the various resorption inhibitors.
And we know, from a very nice article by Profitt published two years ago, the mechanism for destruction of the cancellous bone or destruction of the cortical bone are the same. So inhibition of resorption of cancellous bone and the inhibition of cortical bone are following the same path.
The Guidelines stipulate that different species should be used for those two; this may not be essential. So inhibitors of bone resorption retain existing bone, which is normal because these are patients that lose normal bone, and produce a positive balance.
And as already mentioned, unless there is an effect on the bone mineral or some other problem, the bone that is retained by inhibitors of resorption should be normal bone. Formation stimulators, on the other hands, generate new bone and this bone should be examined for its so-called quality, histologically and by any other means available now, micro radiography, micro CT and so on.
So, in summary, preclinical studies can evaluate bone safety -- the way they evaluate safety in general -- of osteoporosis therapeutic agents and examine if increases in bone mass mineral content, which is clinically measured either as BMD or BMC, are associated with increases in bone strength.
And if they are, then they strongly suggest that the bone accumulated is normal. Preclinical studies can test the efficacy of prospective therapeutic agents in animal models of estrogen deficiency bone loss and potentially other types of bone loss mentioned by Dr. Bone and provide some indication for the future efficacy in patients these models are actually trying to mimic. They can provide mechanistic insights into the mode of action, also with respect to side effects, and with respect to the pharmacodynamic action of these new agents.
And if something in the preclinical studies should be considered and reviewed, as already mentioned, the need for multiple species should be evaluated and the duration of efficacy studies. Efficacy can sometimes be determined in relatively short-term studies, but safety requires several turnovers of the skeleton, which in different species takes different durations. And this could be an add-on to the toxicology studies conducted for longer times, anyhow.
And maybe one should define different criteria for preclinical studies aimed at approving resorption inhibitors, which act through the same mechanism at the tissue level.
So I don't know if one should actually differentiate between them. This could be a topic of discussion later on. So thank you for your attention.
CHAIRMAN BRAUNSTEIN: Thank you. Our next speaker is Dr. Rizzoli.
DR. RIZZOLI: Mr. Chairman, ladies and gentlemen. I'm a little bit confused. I'm a MacIntosh user, and this one is not my most favorite instrument.
(Pause)
DR. RIZZOLI: Okay. Mr. Chairman, ladies and gentlemen, I feel very honored to be here, being a non-European and non-U.S. representative, and despite the fact that my group is a WHO collaborating center, I will not speak on behalf of this organization, nor of the group who set the guidelines in `98.
So my thrust today will be to discuss a little bit what we could do with preclinical studies in terms of the developmental strategy of new drugs for osteoporosis.
Now, if we look at the three main guidelines you have in your handout, in the objectives, something is very common. Everybody is interested to look at the strengths of bone under the different conditions. So it's very similar from one guideline to the other. And in terms of safety, as well, everybody's interested to look at if a new drug could be potentially harmful to bone.
But what differs a little bit is one of the objectives proposed by the WHO Working Group to look at mechanisms beside bone strength or safety. And, finally -- and probably this is related to safety -- there is a need to look at the effect of an intervention during fracture repair, since we will be probably at risk of giving this new drug to a patient already with a fracture who will experience a new fracture during therapy.
So now, if we look at the different details, and we will not go into, as pointed out before by Gideon, by Henry Bone, the different working groups have set very strict, precise conditions to investigate drugs at the preclinical level in terms of species, in terms of design, in terms of duration, in terms of scheduled dose and the different variables to look at.
And if we consider the different drugs which have been now registered and how the clinical outcome has been predicted by preclinical data, we will notice that it's relatively or extremely good, since many of the outcomes we have collected in large, randomized, placebo-controlled trials could be foreseen by looking at the preclinical file.
And just to give you a few examples, you know the data with Alendronate. The left-hand side has been presented already by Gideon. An increase in BMD is associated with an increase in bone stress and this, in clinical trial, is ending in a reduction in fracture rate.
The same is true for Risedronate. Before it was in baboons, here in mini pigs: better strengths, lower incidence of fracture rate. And now, if we look at another category of compound, the SERM, once again, in ovariectomized rats, Raloxifene is associated with an increase in bone strength in the vertebral body, and this has been done by the next speaker.
And if you look at the clinical trial, everything is going in the same direction. And even for a bone forming agent like PTH, you see that you can detect -- in the lab with animal data, you can detect the increase in bone strengths and then you can detect the same effect, favorable effect on bone by looking at the incidence of fracture.
So if we summarize that all together, there's a good relationship. Now, the issue of fluoride, and the issue of fluoride is perfectly supporting this aspect, since if we are to take into consideration the preclinical data, maybe we would have saved a lot of time and money and, ethically speaking, some problems with the patients since none of the preclinical trials were in favor of a positive effect on bone strength, and most of them were even in favor of a deleterious effect.
I'm sorry. I forgot one of the initial references. So if we summarize all together, we have a good relationship between the effect on one side, BMD, bone strengths in preclinical trial, and then a good relationship with what happened in humans in terms of BMD and also in terms of fracture reduction.
And as mentioned previously by Dr. Abadie, the recent data with strontium ranelate goes in exactly the same direction. It has been recently reported that this compound is reducing the incidence of vertebral fracture and probably also of hip fracture, and this was associated in animals with an increase of bone strengths.
So having said that, now the question is how the three guidelines are taking into consideration these values. And there there is some discrepancy because for the FDA document, for instance, the complete file should be submitted at the end of the Phase III.
On the other hand, the CPMP suggests that this could provide some information to the development of the Phase II trial, and most strongly, the WHO document relies on that to design a good Phase III trial, and then it will integrate this data in the overall analysis of the outcome.
Now, if we put that also together, you see here the details. There will be some information about quality, bone abnormality, and this will be integrated in terms of the final convincing procedure to see whether or not we will be convinced by the quality of the file.
However, there are some issues which are not exactly solved at the present time, and we should probably take into consideration for further assessment of drugs in preclinical trials.
First of all, is the relationship between bone mass -- and when I write bone mass, it's bone density, either volumetric or area bone density, so the amount of bone, irrespective of the type of expression -- and bone strength.
And the idea would be what could we foresee to detect the exception to the nice relationship. And despite Gideon before emphasizing the point that for a category of drug at the tissue level the expression should be the same, if you would look at this recently published data in which, in a face-to-face trial in ovariectomized rats, we gave either SERM or bisphosphonate at doses leading to exactly the same increase of bone strengths, you see that there was some small discrepancy in terms of area bone density on the left side, and volumetric or trabecular bone content at the middle.
And then in terms of bone turnover, for the same increase in bone strength, there were some small differences in the same category, but not in the same class of compound. So this should be considered in all the determinants of bone strength.
Then another possibility is the heterogeneity of action within a direct category. We know, for instance, for the class of SERMs, they are triggering so many different genes.
And if we look now at one recent poster presented two days ago in San Antonio, you see here with one SERM in which the vertebral strength was measured at the maximal effective doses, and compare -- this new SERM is in yellow -- compared to Raloxifene, there was a small difference in terms of bone strength, and this was associated maybe with another mechanism of action, since measuring IGF-1 in these rats, it turned out that the level was increased.
So it has been already pointed out that the safety window should be considered. For instance, with anabolic agents or even with a new antiresorbents with a new mechanism of action, maybe this dose, proposed at five times the effective dose, maybe we could end up in some nonphysiological or even the toxicological side of bone.
And, finally -- and this has been drastically proposed by the WHO document -- the integration of preclinical studies in the drug development strategy, taking into account that both approaches are complementary, could help in the definition of study design, including the endpoints, the type of population. And you have in the document of the WHO a very boring table, but this table tries to synthesize, to summarize this complementary approach, this synthesis between the preclinical part, which would help you to define the endpoint and the population to study, and the different outcome to organize in a Phase II and a Phase III trial.
So with this table that I invite you to carefully analyze, I'm finished, Mr. Chairman. I thank you.
CHAIRMAN BRAUNSTEIN: Thank you very much. Our last speaker on the topic will be Dr. Charles Turner.
DR. TURNER: Thank you, Mr. Chairman, and members of the Committee. My topic today is this term "bone quality" that's come up several times previously. And this has been used as an explanation when bone mineral density doesn't necessarily explain changes in fracture risk.
And what I'd like to do or try to do is show some different aspects of bone quality and what we currently know about the different effects of drugs on this. Now, I'm picking several examples of therapies. We've already heard about fluoride, and you probably understand why I picked it. And I'd also like to address antiresorptives, with bisphosphonates as the paradigm there, and anabolics with the parathyroid hormone fragment as a paradigm.
And the reason that we're very interested in fluoride is that this has been the most famous failure in osteoporosis therapy. You see, this is from the Mayo Clinic trial where nonvertebral fractures were actually increased threefold after the therapy, even though there were tremendous increases in bone mineral density in the vertebral bodies.
So this clearly is a serious problem that this therapy caused, and what was subsequently shown is that it was a problem in the mineralization of the bone.
Here we see what should be healthy bone in the light color and this undermineralized bone throughout this biopsy sample. Now, if I can get this to forward. I've lost control.
(Pause)
AUDIO-VISUAL ASSISTANT: Why don't you go ahead and keep talking. I'll fix it. It doesn't like your presentation.
DR. TURNER: Apparently not.
(Pause)
CHAIRMAN BRAUNSTEIN: Well, why don't we take our ten-minute break now.
DR. TURNER: Yes.
(Laughter)
CHAIRMAN BRAUNSTEIN: And when we come back the technical difficulties will hopefully be resolved.
(Whereupon, the foregoing Meeting went off the record at 10:13 a.m. and went back on the record at 10:27 a.m.)
CHAIRMAN BRAUNSTEIN: I'd like to ask everybody to take their seats, please. Okay. We were unable to fix the glitch in Dr. Turner's PowerPoint presentation, and so he's going to give a five-minute overview of what he was going to show, and then we'll take about 15 minutes or so for questions from the panel and the guests to the initial six speakers.
DR. TURNER: Okay. Very well. Welcome back. It doesn't look like we're going to get to -- be able to get to the histological slides because that's where it hung up, and really the focus of the presentation was on some of the microstructural details that occur with different treatments.
And I think you'll have to take my word for this, but I will explain what we've learned from the histology and the different types of bone quality changes that have been seen in preclinical studies.
Just to summarize, there have been two incidences or two observations where a drug treatment has actually caused more fractures than it prevented, and one was with fluoride treatment and this was done in clinics. And the example, the main example was the Riggs study, published over ten years ago.
The other example was in dogs with high-dose etidronate that Dr. Rodan referred to, and this study also caused multiple spontaneous fractures in these dogs. And there were a number of theories on how etidronate caused fractures in these dogs, because the histology showed that the fractures were associated with very low bone turnover.
And there was one theory proposed by Michael Profitt and some others that this was due to the lack of repair, of turnover of the tissue so it was not able to repair accumulation of micro damage, and therefore, the micro damage caused the fractures.
We've subsequently shown in studies with David Burr in our laboratory that that's not indeed the case. The major problem with Etidronate therapy was exactly what Dr. Rodan illustrated. It's a problem with mineralization in that although there is an accumulation of micro damage when bone remodeling is reduced -- and this would apply to all bisphosphonate therapies -- this is a minor change in bone quality and not the major effect.
And we've seen with anabolic therapies, or at least with the currently investigated anabolic therapies; that is, the full-length peptide PTH and the PTH fragment 1 to 34, that the gain in bone mass occurs alongside a rapid increase in bone turnover rate and a rapid increase in bone porosity.
So the question there was whether or not this could be potentially dangerous if you had too much bone porosity, and the answer in preclinicals and now in clinical studies is that no, it isn't, that bone porosity, while it's a great concern, is not a dangerous side effect.
So in summary, we've pretty much identified several different bone quality effects in a tissue. One is impairment of mineralization or osteomalacia. This is definitely a problem and this can be shown in preclinical studies, and it has been shown clinically to actually increase fracture rates.
The other histological findings that are associated with bone quality are increase in micro damage, increase in mineralization, which also occurs when you have decreased bone turnover, both of which do occur and they have been demonstrated but have not been shown to cause great detriments in the efficacy of the drugs, and then the increase in bone porosity which occurs with increased bone turnover.
So with many effective therapies we have combinations of positive effects and negative effects that go together, but other than the impairment of mineralization, these effects tend only to blunt the efficacy of the drugs, not to cause actual detrimental side effects or increased fracture rates. So I'll stop there and we'll move on.
CHAIRMAN BRAUNSTEIN: Thank you for responding under very difficult circumstances. We'll take a few minutes to, again, have the Panel and guests free to ask questions for clarification from any of the initial six speakers. Yes, sir, Dr. Cummings.
DR. CUMMINGS: Charles, how long would it take in bone to see -- in the case of an antiresorptive in humans, how long would it take to see detrimental effects on bone strength if such were to occur as the result of an inhibition of resorption?
Would you see that in three or four years? Would it take longer, five, six, seven, eight? Do you know what the differences affect? Because most of these preclinical and other studies that you're talking about have been done over the course of a very short period of time.
DR. TURNER: That's an important question and a difficult question to answer in the clinical setting because typically an antiresorptive therapy will cause a combination of an increase in the amount of bone tissue that's available for structural support.
So that's a positive effect, and maybe that bone tissue might have a little bit more mineralization, which tends to make it a little bit more brittle, or it may have some micro damage that accumulates because it doesn't repair as well.
Now the best we can say is that these two effects must balance each other to some extent because the outcomes at six, seven, eight, nine, ten years tend to still show fracture efficacy with antiresorptive treatment.
CHAIRMAN BRAUNSTEIN: Yes, Dr. Gelato.
DR. GELATO: I was going to ask Dr. Rodan, given the presentation that you made, or Dr. Rizzoli rather, or both, since you both talked about preclinical trials, whether the preclinical trials could be used as a screening mechanism.
In other words, if you find that there is an agent that shows evidence of osteomalacia, which you so elegantly showed in both of your presentations for Etidronate and fluoride, and maybe that would say this is a drug that should not go on to development or, you know, if it is already in Phase I trials or whatever, that maybe it should be a drug that should be considered not to go further.
I mean, is that something that -- how these preclinical trials -- I guess it goes to the question that the other gentleman asked about, you know, how long does it take to see these effects in an animal because clearly they mirrored exactly what was seen in the clinical trials.
DR. RODAN: If the defect is in mineralization it can be readily detected relatively rapidly in animal studies based on the experience of decades now. And I think it would be wise not to proceed with such a drug into the clinic.
CHAIRMAN BRAUNSTEIN: Yes.
DR. ABADIE: It was exactly the sense of the role of the importance of the preclinical studies in the CPMP Guideline. I mean, that as far as the registration is concerned, I would respond to that that we are not that, I would say, happy with considering the preclinical studies.
But to go into Phase II, exactly as you pointed out, we think that's important because if there is a defect in bone quality, it's clear that we will not encourage the company to go into Phase II.
CHAIRMAN BRAUNSTEIN: Okay. Yes.
DR. TAMBORLANE: I think I'm directing this at the Agency. But if you're looking at -- say you have an approved drug and you're looking at a new indication, say the use of a bisphosphonate for glucocorticoid-induced osteoporosis, do you ever go back and ask for preclinical studies if the mechanism might be different in a different indication?
CHAIRMAN BRAUNSTEIN: Dr. Orloff, do you want to --
DR. ORLOFF: Yes. We're consulting with our pharmacology colleagues.
DR. TAMBORLANE: I can make -- rephrase the question. Would there be a usefulness in doing such?
DR. ORLOFF: The answer with specific respect to the glucocorticoid-induced osteoporosis indication is no, we did not ask for specific preclinical studies in an animal model of glucocorticoid-induced osteoporosis, in that instance because the sense was that there was not a good animal model available.
Do you want to elaborate on the question?
DR. TAMBORLANE: Well, it just seems that, with the data they have presented, that these were good predictors. And Dr. Bone raised the issue that there may be different mechanisms with, you know, the effectiveness of the drug in glucocorticoid-induced osteoporosis might not -- might be different.
I'm just using that as an example, but you know, that's -- so it seemed to me that -- in the discussion that that might be -- useful information might be derived from that kind of approach.
DR. BONE: Could I respond to that? I actually didn't mean to imply what you inferred. What I was getting at is that -- I was suggesting that we probably only needed to look at fracture data if we need fracture data to confirm that there is a consistent relationship between mass and strength.
We probably only need to do it in one indication, unless there is a particular reason to think it might be otherwise in a different indication. In other words, if preclinical testing or some good theoretical reason related to the drug's mechanism of action as it would relate to one of these other kinds of osteoporosis raised a serious question about whether the results might be different, then you would have more of a reason to look at fracture data.
But if there were no such reason, then what we're really asking is do we think there's a toxicity or not. Do we think that the drug in some way undermines its own benefit. And so that's what I was trying to get at.
I was sort of saying, if such a circumstance existed, then you might have to go and look at that as a completely separate entity, but that wouldn't necessarily be the case. I'm not suggesting that that should be the rule.
CHAIRMAN BRAUNSTEIN: Yes, Dr. Levitsky.
DR. LEVITSKY: Bill's question is actually a rather important one. It was my understanding that a number of the other disorders which have become peripheral beneficiaries of these drugs do have alterations in bone matrix, which is different from the osteoporosis associated with the loss of estrogen, and I wonder whether there shouldn't be a closer look before we sort of generalize.
CHAIRMAN BRAUNSTEIN: Which disorder is that?
DR. LEVITSKY: Well, I'm thinking, for instance, of glucocorticoid-induced. Aren't there changes -- don't the glucocorticoids change because they change protein turnover? Don't they have effects on bone matrix? Or for instance, in pediatrics, osteogenesis imperfecta, where there are some very excellent trials showing an effect of some of these drugs, which is wonderful. Obviously, the animal model is a little difficult, but nonetheless, it is a concern, perhaps.
DR. BONE: Gideon or Rene, in glucocorticoid steroid osteoporosis, do we know of any reason to think that we would have a toxic effect with a drug in that model that wouldn't be apparent otherwise?
DR. RODAN: At the resolution at which we can evaluate it, there's no detected difference in bone as a material, in the composition or otherwise, in those disorders like glucocorticoid-induced osteoporosis. It is different in OI, osteogenesis imperfecta, where the collagen has a different structure.
However, it seems, especially based on the response to clinical intervention, that a major component of the fragility is increased bone turnover. That has been reported by Prokoff in OI 40 years ago. And this really is probably why there is response to antiresorptive therapy.
CHAIRMAN BRAUNSTEIN: Dr. Sampson.
DR. SAMPSON: It's a question, I guess, to Drs. Colman, Abadie and Rizzoli. It appears that in the FDA Guidance and the CPMP Guidance, in terms of fracture assessment, its incidence of fracture is the primary efficacy variable.
And I was wondering what would have led WHO to suggest that "time to fracture," to use their language, might possibly be a primary endpoint, for example, in studies of hip. The difference in choice of primary endpoints in the three documents, or maybe I'm just reading it -- I'd like to have further information on that, please.
CHAIRMAN BRAUNSTEIN: Dr. Abadie, do you want to start?
DR. ABADIE: The first endpoint is, for us, the patient as a sample unit. But the time, the time to event, which is the time to first fracture for one patient is extremely important, and obviously, we will ask for that for every submission.
DR. SAMPSON: But would that be considered secondary or is that considered a primary response variable in actually evaluating the primary efficacy?
DR. ABADIE: I think it will be a very important secondary.
CHAIRMAN BRAUNSTEIN: The major endpoint being fracture and secondary endpoint being how long it takes to develop a fracture?
DR. SAMPSON: Yes.
CHAIRMAN BRAUNSTEIN: Dr. Temple.
DR. TEMPLE: As a practical matter, if there are a lot of dropouts it's sometimes easier to do a hazard ratio based on time to first whatever it is. That's true whether it's cardiovascular endpoints or these.
I suspect that has something to do with the reason, too, although I must say it's much easier for people to understand fractures at six months, fractures at a year. That's more tangible than hazard ratios, I think, but I -- that may be the reason. It's an easier thing to calculate if not everybody stays in the study.
CHAIRMAN BRAUNSTEIN: Dr. Rizzoli, do you want to comment at all? No. Okay. Yes, Dr. Silverstein.
DR. SILVERSTEIN: Yes, thanks. Doctor -- I guess this is to you, Dr. Abadie. When you gave your presentation, you said that bone mineral density was not generally a good predictor of fracture risk, although it was good for bisphosphonates or better for bisphosphonates.
And from a lot of the other presentations, it appears that if the preclinical studies show that the bone structure and strength in the preclinical trials are good, then bone mineral density appears to be a better predictor of how they're going to do in the clinical trials as far as fracture risk.
So if you are using drugs with good bone histology, good bone strength in the preclinical trials, can bone mineral density be used as a surrogate marker for fractures, do you think?
DR. ABADIE: Well, the point is well taken, but I'm afraid that in Europe we will not consider, as I told you before, the importance of the preclinical studies, and we will mainly focus on the fracture.
And the reason for that is, I think, sensible insofar as the BMD and the fracture may be qualitatively related in some sense, but probably not quantitatively related at least for most of the pharmaco class today.
I mean, if you take into account the preclinical and the BMD, and if you say the preclinical is extremely important and the BMD is also important, therefore, you could go as the drug guidance of the Food and Drug Administration.
The problem is that we discard, more or less, in this reasoning the preclinical. And therefore, we are left with the BMD versus clinical fractures. As I told you, we are not absolutely sure that the relationship, and especially the quantitative relationship between BMD and fracture is sufficient to approve a drug based on BMD alone.
CHAIRMAN BRAUNSTEIN: Okay. Dr. Grady.
DR. GRADY: Yes. You know, I think we've had a couple of presentations that suggest that preclinical studies are pretty good at picking out agents that, even though they increase bone density, may not decrease fracture risk, and at least in retrospect have been pretty good at figuring out why that might be, at least for Etidronate and fluoride.
But I guess I'm -- that seems to me to be old history. That's not exactly what I'm worried about right now. I'm actually more worried about estrogen, for example, and harmful effects that actually have nothing to do with bone. I mean, I guess you can't answer this, but that's my problem.
CHAIRMAN BRAUNSTEIN: Dr. Marcus.
DR. MARCUS: I'd like to address some -- now that this topic has been introduced of BMD, there have been some things that have --
CHAIRMAN BRAUNSTEIN: Well, we're going to have a discussion, a presentation on BMD and then there'll be ample time for elaboration on that afterwards. So why won't we wait until -- if it's a BMD question, why don't we wait until after we have the BMD discussion?
DR. MARCUS: That's fine.
CHAIRMAN BRAUNSTEIN: Okay. Yes, Dr. Bone.
DR. BONE: All right. Just a further comment about the question that was raised about time to first fracture. It's important to realize that the most of the fracture events that are counted in clinical trials are not clinically symptomatic.
These are -- what you'll hear about from Dr. Faulkner are deformities of the vertebrae, and we're not doing an x-ray every day. We're only doing those at the specified event, or specified times, like annually or something like that.
So that kind of fracture, which constitutes a large percentage of the events that are counted in the trial, would be only detectable on sort of a per year basis, or something like that. You couldn't get the time-to-event comment.
One other thing I was just going to ask Dr. Rizzoli to comment about, because he has extensive experience with the category, the broad category, with many sort of the pleiotropic category, if we can call it that, of selective estrogen receptor modulators, is that, in the one marketed drug in this category -- and I'm sure that a lot of drugs have been washed out because of the testing that's been required -- but in the marketed drug we do have this discrepancy between the significant reduction in vertebral fracture and a relative risk of about one for the nonvertebral fracture, the hip fractures.
Can you give us a little discussion about if we're seeing a disconnect there between the bone density effect and the fracture rate?
DR. RIZZOLI: I wish I could do it because this is an issue which has puzzled many, many people. I cannot give an answer why for the same decrease in fracture at the vertebral level and given an increase in bone mineral density, the peripheral fracture are not influenced in the same way.
So I cannot answer you. But you raised two other points. The first point is, within a category of compounds with probably exactly the same mechanism of action maybe the BMD is a relevant issue.
For instance, a new amino bisphosphonate having been shown in preclinical data that the relationship between BMD strength is the same, there is no mineralization impairment, the pharmacokinetics is the same, probably the effect on the fracture rate is likely to be similar.
On the contrary, with the SERMs, for which as you know the mechanism of action is probably very, very different from one compound to the other, it would be very difficult to draw a conclusion from one compound to the other.
CHAIRMAN BRAUNSTEIN: Right. Yes, Dr. Watts.
DR. WATTS: In thinking about the preclinical studies I think it's important to separate where you're looking for safety or toxicity problems and where you're looking for efficacy problems.
In particular, the doses of Etidronate that were shown to impair mineralization in animal studies were much higher and the exposure much longer than the doses of Etidronate that were used in clinical trials.
And while I don't want to get into details on the Etidronate study, it was not powered to show an effect on fracture, but it did have extensive bone histomorphometry data available for at least seven years of treatment, and there were no problems with mineralization identified there.
So the dose that you study for toxicity certainly raises the possibility that there might be a problem with the lower dose used for efficacy, but it doesn't mean that a lower dose would be ineffective.
Dr. Rodan, on the efficacy side, pointed out the strong relationship between increases in bone volume and increases in bone strength, and I asked him at the break and would appreciate a clarification: Since none of these agents have been shown to increase bone volume in iliac crest biopsies in humans, why should we extend the observation of this relationship in animals to the antifracture effect in humans.
DR. RODAN: You have the question. So I mentioned that in our three-year baboon study we did not see increases in volumetric bone amount in the ilium. We were very surprised about it, but we did see it in the spine, and this is what we published.
So the ilium has problems of sampling and it's a nonloading bone and so on. And the data are usually collected in the spine, so there are site differences which limit extrapolation from the ilium to the spine. So this is what I told Dr. Watts.
Now, I meant to answer some of the questions here. The very strong correlation one sees in amino bisphosphonates between bone density and a reduction in fractures may have a quantitative component to it because amino bisphosphonates are the most efficacious inhibitors of resorption now used.
So the amount of change in bone whether due to mineralization or to increase bone, volumetric bone mass, is larger than for the other agents. And so we may not have the power to detect -- there may be a quantitative aspect to it.
We may not have the power to detect the fracture efficacy with agents that are not as robust in their antiresorptive effect, and there may not be a mechanistic difference between the action, all of them inhibiting resorption, but some less and some more.
So that's why the bisphosphonates came out to have such a strong correlation.
CHAIRMAN BRAUNSTEIN: All right. Thank you. I think we'll go ahead and move on to the next set of talks on measures that -- pardon?
DR. ORLOFF: I just want to make one comment, if I might, before you go on.
CHAIRMAN BRAUNSTEIN: Yes.
DR. ORLOFF: I wanted to make sure that Dr. Grady's question didn't get completely dropped, if indeed it was a question. And I think what it gets to, at least from my interpretation, is a concern that if certain drugs were approved based upon trials that didn't go as far as to assess fractures --so foreshortened, if you will, because of a requirement only to examine BMD -- there is no necessary reason why those trials have to be shorter or smaller, and that the safety concerns or the need for safety information will always drive the size and durations of trials, in this instance and for other, you know, parallel conditions, in a chronic asymptomatic disease in the vast majority of patients who are, you know, affected at any given time.
So there's always -- there will be opportunity to get safety information, both for the skeleton and at nonskeletal organ systems.
DR. GRADY: Well, that's true if estrogens are allowed to be approved based on only BMD studies. I mean, the average sample size there is a few hundred, compared to, you know, a few thousand in the fracture studies.
DR. ORLOFF: That's a minimum sample size based upon considerations for efficacy. But I'm just saying that there's no reason why the trials have to be limited, and it's up to us to ask for more patients and longer duration to make sure that we're not overlooking some sinister effect that might accrue over the longer term.
CHAIRMAN BRAUNSTEIN: Thank you. We'll move onto the measures of clinical efficacy. The first speaker will be Dr. Faulkner, speaking about measurement of bone mineral density in vertebral fractures.
DR. FAULKNER: Thank you very much. I'm very pleased to be here today. I wish to acknowledge, by way of disclosure, that I am an employee of G.E. Medical Systems. We do manufacture densitometry and x-ray equipment. However, that's not the subject which I address today, differences in the equipment.
I am here to address the techniques in general. So I hope you'll find that acceptable; and acknowledge, also, the significant contributions of Professor Harry Genant, actually one of my mentors in my early career. He is joining us by video conference and will be available if we have a question regarding the radiology, specifically in vertebral fracture assessment.
So I'd like to just start with the basics. My training is in biomechanical engineering, and one of the things which I have learned and has been confirmed to me repeatedly is that there is an exponential relationship between the density of bone tissue and the strength of that bone tissue. And this is done predominantly in excised specimens, but it has been well shown by decades of research.
So that we know that this exponential relationship is such that small declines in bone density correspond to large differences in strength, and in particular, fracture risk. I quote here a mets-analysis of Debbie Marshall which combined a lot of studies that had been done over the past several years showing this to be true.
But this also means that small increases in bone density, if we drive the curve the other way a few percent, can reduce fractures by 30 to 90 percent. So, again, the nonexponential feature of the curve is that which I wish to stress.
It's not that if you wish to reduce fractures by 50 percent that you have to increase bone density by 50 percent. It's not even close to the case. It's very, very small changes leading to very dramatic changes in strength.
In fact, the conclusion of the Marshall review was that the predictability of bone mass was better than that of serum cholesterol for cardiovascular disease. You can see here that for bone density and fracture, as bone density increases the relative incidence of fracture decreases greatly.
There's about a tenfold gradient between those who have low BMD, lowest quartile, to those who have high levels of bone density. Wherefore, cholesterol in comparison, you can see that there is a similar type of relationship, but not nearly to the degree, about a fourfold change going from the lowest quartile -- or actually, the highest quartile of increasing cholesterol down to about, as you can see here, much less of a steep gradient for cholesterol and heart disease.
There are a lot of different ways to measure bone density. I'm going to do a quick review of those here. They've been mentioned previously. It was possible, and still is indeed possible today to measure the peripheral skeleton using conventional x-ray techniques.
This is really unsuitable for a few reasons, though, for our discussion today. One, we'll talk a little bit about the utility of peripheral bone density measurements -- that is; nonspine, nonhip measurements, as I define them -- for monitoring changes in efficacy of drugs.
Also, conventional x-ray systems have limitations of using -- of requiring calibration phantoms, or they're not as well used, either clinically or in research today. There are other options, to use smaller peripheral-based x-ray units, as well. I show several here.
They are nice in that they're portable, but they have limited measurement sites, again. And as you have been hearing, most of the time the requirements have been to measure spine and hip and these devices are not equipped to do that.
Ultrasound has been seen as a little more recent advance in the field of densitometry, though it has been around for some time. Ultrasound is speculated to maybe measure properties of bone which are beyond just bone density and might be related to some of the infamous quality issues, which we're discussing here.
But this really has remained to be determined, though. I think that this point, in particular, I believe that ultrasound is primarily an alternative measure of bone density.
It probably has some component due to other factors, as well, but at least your colleagues on the radiologic devices panel chose to approve it as an estimator of bone density, and I think for our purposes that's true, as well.
It's -- but not using x-rays, so that that does have some regulatory and safety considerations, though x-ray dose with all these techniques is extremely low. Again, limitation here, even though some of these devices may have clearance from the radiologic devices panel of the FDA for monitoring, I think that we realize, as well, that it's by monitoring those sites in the peripheral skeleton, which don't change as rapidly, that you are -- they are not as efficacious for monitoring for the purpose of our discussion today.
Most of the Guidance, I think all of the Guidance documents that we've reviewed up to this point have concentrated on the use of central bone density measurements, that is, spine and hip measurements predominantly, using a technique, DXA.
It has the advantage of measuring not only the spine and hip, but can also measure forearm, total body measures. These can be of particular importance, for example, with some agents that may not have the same type of effects on the skeleton as antiabsorptive bisphosphonates.
For example, with a recent application with PTH, concerns over effects on cortical bone brought out the importance of doing total body and forearm measurements. They are somewhat larger, their office space a little more expensive, but really considered the clinical standard.
In fact, all of the registration studies which have been done to this point have been based upon bone density information acquired using DXA. There are techniques using CT scanners that look at a slightly different property of bone.
A quantitative CT measures the volumetric density. The previous techniques use an area or a projection density, or grams per square centimeter. This is actually a volumetric technique, grams per cubic centimeter, are predominantly being done at the spine, has been done in some subsets of study populations, but has not been considered as a primary endpoint for the registration studies, possibly because of its limitation to the trabecular bone in the spine.
I'm showing a CT scan here. It really doesn't give you the full spectrum of both cortical and trabecular bone. But it has proved important for looking at some agents in a research setting, and maybe in subsets of studies.
So it's often asked, I get the question quite frequently, well, you've got a lot of options for measuring bone density, and how well do the various bone density measurements correlate? And since I figured we'd have that question I would give you the numbers here.
It corresponds to -- I think my basic law of correlation is that anything in your body will correlate to anything else at about .6 to .7. And we see some variability around that number here.
But essentially, if you look at the different bones and the different skeletal sites using the various technologies, you do see modest correlations, not surprisingly, but not perfect correlations.
The bone density at say the spine will never really correlate -- will not correlate to a high degree of bone density at hip or other skeletal sites due to the fact that you've got completely different kinds of bone, cortical and trabecular ratios.
You've got a large variability in the blood supply, surface to volume ratio, weight-bearing. So it really isn't -- would not be expected that they should be. And I think it's appropriate to evaluate several skeletal sites when looking at efficacy of bone density for therapeutic agents for these reasons.
But really, correlation in itself is not of that great of interest clinically, even though you may have a disagreement between the raw bone density values at different skeletal sites. What really is important is how these different skeletal sites and measures predict ultimately fracture.
And it has been reported in several trials, and again, showing the meta analysis from the Marshall paper, that hip fractures can be predicted by all BMD measurements, but that hip BMD itself is the best predictor of hip fracture.
I'm showing the age adjusted relative risk for fracture here as a function of the various bone density tests that are performed. And again, not surprisingly, you would think that a direct measurement of the hip would have the strongest relationship to eventual risk for hip fracture.
But it is also true that other skeletal sites, measurements at the spine and the heel and the forearm, can indeed predict fracture, but the relationship is not quite as strong as direct femoral measurements.
When looking at overall risk for fracture the measurements turn out to be very similar. You don't see one -- a preference for one skeletal site over another. In fact, all BMD measurements in the Marshall meta analysis were just about equally predictive of fracture risk.
So I think this has led to the conventional wisdom in the field that if you wish to predict fracture at a skeletal site, at least clinically, that you should try and measure that skeletal site directly.
But for overall risk of fracture of any osteoporotic fracture, then you can really measure any skeletal site and get similar types of results. but in our context of our discussion today, I think that we have to realize that the diagnosis or assessment of fracture risk is important at one level, but also, we are interested in monitoring changes over time.
And these are very different challenges for bone densitometry. When we want to diagnose someone or assess fracture risk, we have to have an accurate bone density measurement. We have to make sure that the number is a true reflection of that patient's density.
We need to have valid reference ranges. We have to know what is normal in order to classify someone as outside of the normal range. And we've also, properly to assess risk, should -- need to include additional risk factors: age, prevalent fractures, family history and many other features must be incorporated, as well, in order to get an overall picture of fracture risk, because it is not just a feature of bone density alone, which you'll be hearing a lot from other speakers.
If we're looking at changes over time, though, if we're now looking at the ability to monitor, precision and instrument stability is really the most important feature here. We've got to have precision or reproduce-ability so that we know that changes that we see over time are true changes in the patient, and not due to alterations in our technique, either alterations in our equipment or alterations in our measuring procedure.
And this needs to be carefully controlled in clinical trials. We've also got to measure response of skeletal site. If you choose a skeletal site that is maybe not as responsive, it may be difficult to see bone density changes, not due to any problems with the technology, but just due to the fact that you're measuring a site that is not changing very rapidly.
And also, appropriate follow-up time. If you wish to do a treatment study looking at change in BMD and confine it to a one-month duration, you will be disappointed because the changes, at least with the current therapies, don't occur nearly that quickly.
So you've got to have -- in most clinical situations it usually takes in individuals about two years to see clinically significant changes, maybe less for steroid-induced osteoporosis. I will point out, indeed, this is for individuals.
When you have group effects you can show changes much more quickly. So in monitoring responses, show data here from one of the postmenopausal registration studies, looking at this time at Alendronate and HRT compared to placebo.
In this case a slew of skeletal sites were measured. We had both the posterior and anterior spine, lateral spine measured, as well as the femoral neck, total hip, the forearm, both at the ultra distal region and the one-third region, and the total body.
And the -- percent BMD change at 24 months, shown here the largest changes indeed occurring at the spine in this early postmenopausal population with this metabolically active bone in the spine. And that was true both in the posterior, anterior and lateral view.
The lateral view looks at a little bit more trabecular bone than the PA view, but in this case they were fairly similar. And note that at least in the case of this particular study that you saw in some cases a loss of bone, or no change in bone density at some skeletal sites, when indeed, the metabolically active site of the spine was showing a significant response.
So TROI's subskeletal site is indeed important. As I mentioned, though, 60 to 80 percent of -- when we take bone specimens of the bone specimen strength, this related to its bone density and it is both cortical and trabecular bone that are important.
If you look at the vertebral body here, for example, in a slide from Dr. Genant's lab, you can see trabecular bone components shown in red and the cortical components shown in blue. And at least at the vertebral body, the predominant weight-bearing site is in the vertebral body here, the posterior elements being used predominantly for muscle attachments and torsional stability.
So this is important to maintain this weight-bearing bone, but I think I point out in the spine it has been suggested by some that trabecular bone is of prime importance.
But I think you can see here from this picture that both cortical and trabecular bone are indeed present at the spine, and even to a larger degree at such sites as the femoral neck, where cortical bone may represent as much as half of the bone density there.
Clinical studies have indeed confirmed that fracture risk is reduced by treatments that preserve bone quality, increase bone density and decrease bone resorption.
So at least as I've reviewed the information that we have, that those studies that have shown a positive effect on quality through various animal studies and bone density through both animal and clinical studies and decreased bone resorption, have by and large gone on to show some degree of efficacy for reducing fractures.
Though I agree with Dr. Rodan that to quantitate the exact relationship does require you to look at the fractures in detail, but by and large, that if you see this type of positive results here that you'll see a positive result in fracture studies, as well.
This is a meta analysis which was done by Richard Wasnich, looking at some 13 clinical trials that were done and its potential -- the potential for a change in bone density to have an effect on vertebral fracture rates.
And it was noted here that those compounds which have a small change in bone density tend to have less of effect on reduction in vertebral fracture risk than those changes that have a large effect on bone density.
So this can be shown when you combine these multiple studies together into a meta analysis, though I will admit for individual studies it has been somewhat confusing as to why some compounds show an affect on bone density somewhat discrepant with their expected change on fracture risk.
I personally believe part of the difference is that we've got extremely disparate populations that we're studying. Some are early postmenopausal, some late postmenopausal, and it's not really fair for us to lump all of these together unless we do something like we've shown here, using a Poisson regression that accounts for differences in sample size.
But the general trend at least is that those compounds that show the greater increase in density, show the greatest reductions in risk. So we do have a lot of different methods for assessing bone density.
I breezed through a brief introduction of them all, but I think for our discussions here today we find that those that are monitoring response and predict fracture are the ones that we're trying to include in our investigations of these various therapeutic compounds.
DXA probably has the big advantage, which is why it's used, because of the fact it can measure the clinically relevant sites, both the spine and the hip. It has been well-documented. There are excellent procedures for performing quality control.
We understand the technology quite well and its ability to monitor has been shown, as well, in all these studies. So let me move on to a discussion of vertebral radiographs and the section that was predominantly prepared by Dr. Genant.
It's -- the idea with vertebral radiographs are we need them as a detector of vertebral fractures, because not all fractures which happen in the spine, in fact a majority of them, are detected clinically.
Many of them pass by without being symptomatic to our patients, at least to the degree that they would present for some type of evaluation. And they have played a key role in establishing the efficacy of drugs in osteoporosis treatment and prevention.
And they have to be interpreted, though, very carefully, as it's not -- I'll show you some examples here -- it's not easy to do these without expert knowledge of anatomy and pathology and some experience looking at these films.
The challenge is really to look at shape recognition. I would sometimes go into Dr. Genant's office when I worked in San Francisco and watch while he read films, and he gave the example of the fact that you can look at a vertebral body and with a trained eye very readily determine whether or not it's fractured.
Whereas, sometimes with an untrained eye it's not easy to do that, and in some cases by using measurements of height, which we'll discuss later, have been proposed as a surrogate. But consider if you would a car that's been involved in an accident.
It would -- it's usually quite easy for us to tell visually if that car is a total loss. Whereas, if you were asked to put six points on that automobile and then based upon the placement of those six points determine if it was a total loss, I think you'd be quite frustrated.
So it's important to have I think at some point a visual assessment. Let's talk about the various deformity indices that we have. There are, as I mentioned, clinical or symptomatic vertebral fractures. This is sometimes used as a secondary endpoint.
That is, those patients who present with back pain or some symptom that would cause them to indicate that something might be wrong, would present and then upon verification with a spine film, that would be called a clinical vertebral fracture.
But many vertebral fractures are not clinically captured, and we've also determined that these nonclinical or morphometric types of fractures can be associated with an increased risk for subsequent fractures, as well. So they are indeed important to capture.
There are several methods to define fractures based upon spine films. There's a simple visual assessment, either a yes or no, based on a radiologist's read. Semi-quantitative visual assessment. That is a technique which divides vertebral bodies into a zero grade, being normal, then from a one being mild, two moderate, three severe.
This I'll explain in a little more detail in a moment. And then we have morphometry, which is a simply measure of heights of the vertebral body at various locations and looking at the ratios of those heights and comparisons within vertebrae or between vertebrae to determine if a fracture exists.
For vertebral radiographs, quality is first and foremost, as the same with bone densitometry. You need to have good quality radiographs. As an excellent example, if you have garbage in, you will get garbage out.
So exposure is important to control. Patient positioning is extremely important because you can mimic the features of a fracture with poor positioning. A depiction of anatomy; ideally, you'd like to see T4 to L1 on the thoracic view and from T12 down to the sacrum on the lumbar view.
Having the overlap between the two views does allow us to accurately quantitate the vertebral levels. For visual assessment you of course need to have, as always, adequate film quality, but I think in this point the experience in trained observers is very important.
To distinguish fractures and other clinical conditions, technical and positional variations in the films requires a trained eye. I know that we've had the privilege of doing some studies comparing radiologists and it's surprising how frequently different radiologists will disagree upon whether or not a fracture exists in a film, just due to differences in their experience, but it has -- requiring the need for some kind of standardization in training.
These are probably going to be a little bit difficult to see with the lights up. I don't know if we can turn them down, but I wanted to provide for you just some of the challenges associated with vertebral fracture.
Is it possible maybe to dim the lights somewhat? Do we have someone that could do that? Thanks. You can see right here, one of the requirements is to have dim lighting while you want to read these, as you could see.
But here, we have an orthograde film, orthograde in that when we have the visualized vertebral bodies here that you do have a view where you're looking down the endplates in a way that allows you to accurately assess vertebral heights.
Whereas, you have here in this film if you can appreciate it, there are the endplates here seen slightly at an angle, which give them somewhat of an oval appearance; so it's very difficult to distinguish whether or not this is in fact maybe some kind of a biconcave fracture, or whether it just has to do with differences in positioning.
And differences in x-ray technology, you can see an under-penetrated or over-penetrated film can be -- make it quite a challenge to assess whether or not someone has a vertebral fracture, as well. In this case, we've got two examples, though, of true osteoporotic fractures, at least according to Dr. Genant's eye, but I believe him.
You could see here deformities of the endplates that are shown here. This is a close-up view where you can see there's quite a decrease in vertebral height in these particular endplates. So these are examples of osteoporotic vertebral fractures.
But there are various different ways you can be tricked, such as here in osteomalacia, you can see that these vertebral bodies here are showing this bow tie or fish vertebrae appearance, in this case not due to osteoporosis, but due to osteomalacia. So it's important to be able to distinguish the differences.
Also, you've got examples here of Cushing's disease, and also steroid-induced osteoporosis. I think it was questioned whether steroids caused differential effects on bone. At least a trained radiologist can appreciate some differences in the spine due to the presence of steroids.
Because of difficulties, though, and the qualitative nature involved with measuring the spine -- or visual assessment of vertebral deformities, there have been creation of semi-quantitative grading scores.
The most well known is that developed by Dr. Genant and his colleagues at the University of California, and have created this pictorial definition of the various grades of vertebral fracture for semi-quantitative grading.
And this is, as you can see, mild, which is approximately a 20 to 25 percent reduction in vertebral height; moderate, 25 to 40 percent, roughly, and severe, about 40 percent or greater reduction in vertebral heights, and this is for both wedge, biconcave and crush fractures, which examples are shown here.
When doing a semi-quantitative assessment you need, of course, to have adequate film quality, as always, experienced and trained operator, but a well-defined fracture criteria and standardization to an atlas.
So when doing studies you'll find that the majority of them have provided some type of training and some type of an atlas so that those reading the films can indeed be brought into synchronization as to what's termed a fracture.
And very often, you have centralized analysis of these things so that you can have a consistent reading across studies. So there is an example of a grade one fracture, which is shown here, and a grade two fracture and a grade three, finally, a severe fracture based upon -- these are representative examples of the criterion which have been evaluated and proposed and used in the majority of the studies today.
You also have the ability now to look at incident, severe and moderate fractures. You see here vertebral bodies. Here, this is a grade zero, which has become a grade two on subsequent follow-up, and also, a grade one fracture here, which has subsequently worsened to become a grade three fracture.
Actually, I think these are in -- may have gotten out of track here. But anyway, the progression of these is something that's difficult to appreciate sometimes in the films.
Vertebral morphometry, the final technique which I'll talk about, the measurement of vertebral heights themselves, requires highly standardized radiographic techniques and very careful patient positioning to evaluate the heights of these vertebral bodies.
It's important to have screening of experts, by experts for the appropriate vertebral levels and exclude vertebral bodies that aren't appropriate for measurement. You need to digitize the films, which has to be done in an appropriate way that allows them to be evaluated on computer analysis work stations using well-defined normative data and an algorithm for fracture which is in line with consensus readings.
This is normally done and can be done straight off of the films, but I think in more recent studies this has been done off of digitized radiographs, using an electronic cursor, if right off the films, or using software tools specifically designed for this purpose.
And you can see here, quantitative morphometry with six-point placement. It's currently -- all the studies that have been done have usually looked at six points, evaluating the different heights of the posterior and the anterior, and then the mid-vertebrae.
But it becomes quite challenging in some cases. You can see here in this particular vertebral body where you've got an endplate deformity which has occurred. In this case it becomes quite difficult to know where to place the mid-vertebral point, as in you have two margins.
You have the margin here and then an inter-margin here, and the standard technique requires you to split the difference and come halfway in between. Also, obliquity can cause some problems in these point placements.
So it is important, I think, to not rely exclusively on measures of vertebral height, but in the case of question to have a trained radiologist provide visual assessment, as well.
The morphometric deformities have been defined in most of the studies as having a three standard deviation or greater decrease in the AP or mid-vertebral height. This is one that's been used in -- for several of the studies.
For an incident deformity, that is, during the study, a 20 percent or greater decrease in either the anterior, posterior or mid-vertebral height has commonly been labeled as a morphometric incident, morphometric fracture.
Of the clinical trials, many of them which have come before this particular Committee, they have used a combination of the quantitative morphometry and the semi-quantitative visual reads of the films. That is, using the ability to measure the vertebral heights and having that confirmed by a visual assessment using a semi-quantitative read.
Various permutations have been used, but I think these two techniques together have been virtually well-accepted as a good endpoint for vertebral fracture assessment.
It is possible to use some of the bone density equipment, as well, to assess whether or not someone might have a vertebral compression. This is an example here of a scan that was done with the bone density system.
The nice feature here is you have the equipment in many of the sites that are measuring bone density and it can measure the entire spine in one sweep, and you see here as a fracture which was identified, and a possibly important clinical tool, but it's not clear that this is going to replace spine films at this point.
There's the ability to measure patients both in the supine lateral view by laying them on their side. Other instruments use a decubitus position, and use dual energy techniques to equalize soft tissue variations.
Here again, appreciate that there was a fracture at this point. But I present this as a potential for future technology and give the quote from Jackie Rea's article just a few years ago when she evaluated the ability of -- in this case she called it vertebral x-ray absorptiometry to assess vertebral fractures, and concluded it showed good sensitivity in identifying moderate and severe deformities and an excellent negative predictive value in distinguishing subjects without those -- without subjects from those with vertebral deformities on a per subject basis.
The part which I didn't show here, though, is I think for mild fractures, grade one type fractures, it does not perform nearly as well, missing potentially a third up to a half of these mild deformities.
And in addition, it becomes difficult to see vertebral fracture from about T6 and above using this particular technology. So at least at this point I don't think -- believe it as a replacement for vertebral film technology.
So let me conclude that at least from a BMD standpoint we know it's strongly related to bone strength and fracture risk. That is something we know and I think agree on. Virtually all clinical studies have used DXA measures of spine and hip for determining efficacy of compounds, but they've been supported by bone quality, turnover markers and eventually fracture studies.
Vertebral fracture determination requires high quality radiographs and highly trained readers to be done properly, and a combination of visual that is semi-quantitative, and morphometric reads represents the current best practice, which we have. And I thank you for your attention.
CHAIRMAN BRAUNSTEIN: Thank you, Dr. Faulkner.
Our last speaker in this session is Dr. Hochberg, who's going to speak about relationship of drug associated change and bone mineral density to fracture risk.
DR. HOCHBERG: Well, while we I guess get all set up I want to thank Dr. Braunstein and the Committee for the opportunity to be here today and to speak to you about a topic which has become one of my favorite areas of interest.
Now, I have to admit that I am not a card carrying endocrinologist or metabolism specialist. I am actually trained as a rheumatologist. So I come here from a different sub-specialty of internal medicine.
And specifically, my title is "The Relationship of Drug Associated Change in Bone Mineral Density to Fracture Risk." Now, a number of individuals this morning have commented on this issue of bone quality.
I'm not going to address bone quality, in particular, as it may be measured by bone turnover. But just to mention that the new definition of osteoporosis, which was proposed by a consensus conference from the National Institutes of Health and which was held about two years ago, suggested that measurement of bone mass with bone mineral density, as well as a measure of bone quality, possibly with bone turnover, were integral components of the assessment of osteoporosis.
Now, I think the question that you want me to address and that I will try and address is, are changes in bone mineral density which occur with antiresorptive therapy, and I will also address it in terms of anabolic therapy, important in explaining the antifracture efficacy of approved agents for the treatment of osteoporosis.
And I'll address this in the context of both vertebral fractures, as well as nonvertebral fractures. Now, several people have commented on the laws of physics as applied to bone, and I just wanted to summarize the earlier comments from this morning in the slide that I made during the break.
So this is an "if and then" relationship. If the material properties of the structure remain normal, then an increase in mass of the structure will lead to an increase in strength of the structure. And this has been reviewed by Drs. Bone, Rodan and Rizzoli this morning.
It's been shown to be applicable to antiresorptive agents of different classes demonstrated in preclinical studies. And the different classes are the nitrogen containing bisphosphonates and the selective estrogen receptor modulator, which may work by different mechanisms at the molecular level, although they all decrease bone resorption.
And it's also applicable to teriparatide, recombinant human PTH, which is not as yet approved. And I'll come back to this later on with some new analyses, but the concept would be that for antiresorptive agents, those that are currently approved for the treatment of osteoporosis in the United States, I noted that estrogen is not actually approved for treatment, although it is approved for prevention.
I'm not sure what the difference is between management and treatment, to be perfectly honest with you. I didn't look it up in Black's Law Dictionary. And then the anabolic agents, as well, neither of which are currently approved.
So let me start with vertebral fractures, and you got an excellent review just now by Dr. Faulkner of the ways in which vertebral fractures can be defined and the ways in which they have been defined in some of these clinical studies.
Now, the analysis that I'll show you is really as a result of three meta analyses. So I'm not going to review the data from individual trials. I will say that earlier this morning Dr. Colman sort of reviewed the evolution of the relationship between BMD changes and vertebral fracture risk reduction from individual trials, and sort of went over the data for Etidronate and then the more recent bisphosphonates, as well as fluoride.
And then Dr. Abadie in his presentation showed one graph with the point estimates in the 95 percent confidence intervals for vertebral fracture reduction, plotted against the changes in bone mineral density compared to placebo for those agents, and I think suggested that there was not a sufficient or strong relationship between these.
Now, in the meta analysis that was done by Richard Wasnich and Paul Miller, which was briefly referred to by Dr. Faulkner, they identified 13 placebo controlled trials of antiresorptive agents that reported both vertebral fracture incidents, as well as change in bone mineral density.
And they used the regression model to relate the change in bone mineral density to fracture risk reduction and they reported their best fit model. They did report sensitivity analyses where they eliminated individual trials, as well as all trials for individual agents and stated that this did not alter the results of the study.
But note that this analysis was performed and published prior to the publication of the data from the Risedronate vertebral fracture studies. And you've seen this graph just before, which is taken from their paper and shows the relationship between change in spine bone mineral density measured over the course of the study on the x-axis.
And this is the difference between the mean difference -- let me say the difference between the mean of the treatment group versus the mean of the placebo group, plotted against the relative risk reduction for vertebral fractures.
Now, we have to remember that in this pooling we're pooling heterogeneous populations, because these are not all women with osteoporosis in these studies, and we're also pooling across different definitions of the outcome.
Dr. Faulkner showed you that there are different ways of defining vertebral fractures, and not all the studies defined a new vertebral fracture as a greater than 20 percent decrease and greater than four millimeter reduction in vertebral height.
Nonetheless, they did report a statistically significant relationship between change in spine bone mineral density and reduction in the risk of vertebral fracture. But what was also importantly reported in their study was that even that the model predicted for a drug which did not increase bone mineral density compared to placebo, that there was still a statistically significant reduction in the risk of new vertebral fractures.
It's also worth noting that they included studies of agents which are not approved for the treatment of osteoporosis in the United States, such as tiludronate, and they also included some of the topical estrogen studies in terms of estrogen patch.
Now, Dr. Cummings, who's going to be speaking this afternoon in conjunction with Dr. Black and others, performed a separate meta analysis which was published earlier this year in the American Journal of Medicine.
They limited their analysis to randomized placebo controlled trails that lasted two or more years in duration, and had an ample number of fractures, five or more fractures per treatment group. They used a slightly different regression method, but again, examined the change in spine bone mineral density in relationship to the reduction in vertebral fracture risk.
And they reported a linear relationship, where the expected or estimated relative risk was equal to an aught .75 minus .03 times the increase in lumbar spine bone mineral density. So assuming a linear model, there was a significant reduction in relative risk of new vertebral fractures, independent of any change in spine bone mineral density, but nonetheless, a small additive effect with changes in spine bone mineral density versus placebo.
Okay. And this table summarizes the two different models for a drug which would have no change in bone mineral density and a drug which would have an eight percent increase in bone mineral density.
Now, note that both of these looked at the change in bone mineral density occurring over the entire course of the study, not just within the first year of treatment. Okay.
Dr. Cummings and colleagues also recognized that the observed changes in lumbar spine bone mineral density explained only a small proportion of the actual reduction in the risk of vertebral fractures, and I think furthered this by coming up with another model where the observed relative risk in the study couldn't then be estimated from the expected relative risk, given the change in bone mineral density, and in fact, that the expected relative risk from the first model underestimated the true relative risk which was observed in the study.
Now, in preparation for today's meeting it was suggested to me that we go back and look at these studies and try and limit -- repeat an analysis, which was limited to agents which are currently approved for use for the treatment of osteoporosis in the United States.
So we went back and repeated the Wasnich and Miller analysis, excluded trials of nonapproved medications, specifically Tiludronate, and added the Risedronate birth studies, and this produced the total of 13 trials.
Now, the results were largely unchanged. Change in lumbar spine bone mineral density remained significantly associated with reduction in the risk of new vertebral fractures. And from the Poisson regression for every one percent increase in lumbar spine bone mineral density the relative risk of new vertebral fractures was .9, significantly different from one.
And there remained an independent effective treatment, even without any increase in lumbar spine bone mineral density, and here the relative risk is .81, or about a 20 percent reduction in risk.
So there does appear to be a relationship between increase in lumbar spine bone mineral density and reduction in the risk of new vertebral fractures, although there is a residual effect which appears to be independent of the change in bone mineral density, and this is probably due to reductions in bone turnover, specifically bone resorption, which affect this indistinct and difficult to define concept of bone quality.
Now, Dr. Silverstein, in her question, highlighted this sort of conundrum which has been labeled the Raloxifene paradox by Dr. Riggs in an editorial earlier this year. And this is that some agents decrease vertebral fracture risk, but have not been shown to reduce the risk of nonvertebral fractures.
And these agents tend to have smaller increments in bone mineral density and bone turnover when compared to the amino bisphosphonates. So this actually prompted us to examine the relationship between change in bone mineral density and reduction in the risk of nonvertebral fractures.
And we published earlier this year in the Journal of Clinical Endocrinology and Metabolism, along with Drs. Wasnich, Miller, Greenspan and Ross, an analysis which pooled randomized, double-blind, placebo controlled trials, which reported changes in bone mineral density and/or changes in biochemical markers of bone turnover, as well as incidence of nonvertebral fractures.
Now, we limited these trials to trials which were conducted in women with postmenopausal osteoporosis, defined either by the presence of a prevalent vertebral fracture with bone mineral density, or a T-score less than or equal to minus 2.0 measured at the lumbar spine or femoral neck, to try and get some homogeneity of the patient population.
There are, however, some differences in the outcome because some trials report all nonvertebral fractures. Some trials reported only a few nonvertebral fractures. So we're still -- we still have the problem of some heterogeneity with regard to the outcome.
And this analysis focused on the change in bone mineral density which was seen within the first year of therapy, and then the overall reduction in the risk of nonspine fractures during the entire study.
So we identified 18 trials which had 30 active treatment groups, which had almost 70,000 women years of follow-up, 92 percent of which were present in the eight larger studies. And there were over 2,400 women who had an incident nonvertebral fracture.
Ninety percent of these fractures occurred in the eight largest studies. And this plot shows the relationship between the change in spine bone mineral density seen at one year in the treatment group as compared to the placebo group, and the relative risk for nonvertebral fractures.
This is a slightly curvilinear relationship where the intercept term, which is estimated from the model where there's no change in spine bone mineral density at one year actually goes through a relative risk of one or no risk reduction.
And the different trials are depicted by different size circles, given the number of person years. But you can see that there's a lot of variability in the estimates from the individual trials.
A similar relationship, although slightly steeper, was noted when one plotted hip bone mineral density, either femoral neck or total hip, depending upon which was reported in the study, as compared to the relative risk of nonspine fractures.
And again, this so-called intercept term where there's no change in hip bone mineral density versus placebo was not significantly different from one. Now, I'm not going to show you the data for change in biochemical markers of bone turnover, because that's not the point of the discussion.
But this summarizes the results such that for every one percent increase in lumbar spine bone mineral density versus placebo there was an estimated eight percent reduction in the risk of nonvertebral fractures, and for every one percent increase in hip bone mineral density, this is within the first year of therapy, there was an estimated 27 percent reduction in the risk of nonvertebral fractures.
Okay. The results were generally robust to removal of both individual trials as well as all trials of individual agents. Now, we also repeated this analysis, excluding trials of nonapproved medications, and this left us with a total of 15 trials.
Here again, the results were largely unchanged. Change in bone mineral density within one year remained significantly associated with reduction in the risk of nonvertebral fractures. The relationship was pretty much unchanged at the lumbar spine.
The amount of reduction was somewhat decreased with a change in hip bone mineral density. And again, there was no significant, apparent independent effective treatment without a change in bone mineral density.
So to summarize these results for antiresorptive agents for nonvertebral fractures, greatest and greater increases in bone mineral density within one year of therapy are associated with a greater reduction in the risk of nonvertebral fractures.
So my conclusions from data on antiresorptive agents are that increases in bone mineral density are important indicators of antifracture efficacy of antiresorptive drugs, both for vertebral, as well as nonvertebral fractures, and increases in bone mineral density appear to be necessary to decrease the risk of nonvertebral fractures.
Now, another issue that was raised was, is there a threshold effect for vertebral fractures, and this does not appear to be the case in terms of changes bone mineral density or reductions in biochemical markers.
And we had stated in our paper that the results could not be extrapolated to anabolic agents. But for today's presentation we actually repeated these analyses again, and included the data from the pivotal Phase III trial of teriparatide, published in the New England Journal of Medicine, and found that the results were largely unchanged in the analyses, both for vertebral fracture, as well as nonvertebral fracture.
Change in bone mineral density remained significantly associated with reduction in the risk of vertebral fracture, and change in bone mineral density at one year remained significantly associated with the reduction in the risk of nonvertebral fracture.
So when one incorporates the results from teriparatide, increases in bone mineral density remain an important indicator of antifracture efficacy for both antiresorptive and anabolic drugs. And I think the caveat here is based on the preclinical data, showing that one is making normal bone and that this is true for both vertebral as well as nonvertebral fractures.
So I want to thank you very much for your time and attention.
CHAIRMAN BRAUNSTEIN: Thank you, Dr. Hochberg.
We'll open both Dr. Faulkner's and Dr. Hochberg's presentations up for questions. I think Dr. Marcus will be first on the list.
DR. MARCUS: Yes. I have a question specifically related to Dr. Hochberg's presentation addressing the issue of heterogeneity among all the various trials that were put into your regressions.
It seems to me that one of the major sources of heterogeneity in those trials, various trials, was the initial bone mineral density of the patients on enrollment into the trial, and therefore, using as your outcome measure the percent change in BMD seems to me to be confounded by the fact that somebody who starts with a lower BMD might have, for the same increment in bone, a relatively higher percentage change, and I wonder if you've been able to look at those data, not looking at percent BMD changes but absolute BMD changes.
DR. HOCHBERG: We haven't looked at the data with regard to absolute BMD changes as opposed to percent. My -- I guess this a potential limitation in that you're right in that individuals who start out with a lower BMD will likely -- will have a greater percentage increase in bone mineral density with treatment.
One thing that we did was to restrict, at least, the BMD definition in the analysis of nonvertebral fractures to include studies just in women with postmenopausal osteoporosis. While there is a variability in BMD, it's not as great as in the studies which have looked at reductions in vertebral fractures. I think Dr. Cummings has a comment.
DR. CUMMINGS: We did it both ways and it didn't make a difference.
DR. MARCUS: Thanks a lot.
DR. HOCHBERG: Thank you, Steve.
DR. GELATO: This is for Dr. Hochberg. I guess the question I have is the drugs that don't show a change in BMD but do show a change in fracture risk, although it's only 20, 25 percent, if you use BMD as your primary outcome what would you do with those drugs? They would just be -- you know -- because clearly, as a clinician what I'm concerned about is, I mean, I see a number of patients who can't tolerate the bisphosphonates.
So you know, what do I do with them, you know? And so it becomes, if we're going to just look at BMD --
DR. HOCHBERG: Well, I share your concerns as a clinician in terms of treating patients with osteoporosis who don't tolerate oral bisphosphonates. I think the issues are several and I'm certainly not proposing to the Committee that they decide to recommend changes in guidance and ignore let's say non-BMD effects of therapies, because clearly, all of these analyses have demonstrated that there is a relationship with reduction in vertebral fractures for drugs which do not have a robust effect on changes in bone mineral density as measured in the clinical trials.
What I do in clinical practice is obviously probably different from what other people do in clinical practice, but I tell my patients about the caveats of the results of the trials and what they can expect from -- what I feel they can expect from the individual drugs, and I base my choice of therapy on that.
CHAIRMAN BRAUNSTEIN: Let me follow up with a question to Dr. -- actually -- Rodan and Dr. Turner about this. Are the antiresorptive agents also potentially decreasing the breakdown of the cross-struts in the vertebrae, or doing some other things that will maintain tensile strength, but you may not see a change in density because of the imprecision of the machines, or what?
DR. RODAN: Excellent question. Actually, they preserve the bone that is there, and on a very hypothetical basis it's possible that the bone that is added as part of the normal process of remodeling is added at places where it has the best mechanical function, because mechanical loads influence how bone is being built and remodeled.
So by giving the bone an opportunity to accumulate, the bone that is added may accumulate where it has the best mechanical function, and this is well established for 100 years now. So this may explain some of the discrepancy that you get increased fractures, resistance of fracture prevention, without actually seeing the cumulative bone. It's maybe where the bone has redistributed that is more favorable now.
CHAIRMAN BRAUNSTEIN: Dr. Turner, did you want to comment?
DR. TURNER: Yes. I'd like to respond to your question and also make a comment about BMD. First, it's important to realize that bone resorption is a focal process. And if you could imagine a beam that supports a building, if you had somebody with a jackhammer trying to cut a little piece out of the middle it would greatly weaken the beam, and much more than what would be measured if you simply measured the overall amount of material that was in the beam.
So if you can produce a drug that inhibits bone resorption you can take away all of these little focal stress raisers or jackhammers from the trabecular bone. And this may well explain -- this hasn't -- this is somewhat hypothetical, but it makes sense and it may explain why some antiresorptive agents, particularly the example of the Raloxifene, was brought up.
That worked better in the spine than they do in the hip, because the hip fractures are more of a cortical bone, biomechanical problem, and they don't require -- the trabecular strut aren't as important. And this, I think, is a very plausible hypothesis and it does link turnover with fracture reduction, and at a structural basis because, of course, turnover means nothing if it doesn't have a structural outcome.
And that's probably what's happening. Now, I do want to make one other comment and that has to do with, what is bone mineral density. This is a measurement that you get from a densitometer, but in the case of an antiresorptive agent you're actually decreasing bone turnover, which allows an extended period of secondary mineralization, and the amount of mineral in each strut of bone is actually higher.
So a bone mineral density that you measure with an antiresorptive agent will actually have more mineral for less volume. So it may mean something different than, say, an anabolic agent such as the parathyroid hormone fragment that's been reviewed by this body.
This increases bone turnover. So now, we actually have less time for mineralization. You have less mineral for each component and probably more bone volume. So you take an exact same bone marrow density with the antiresorptive, and with a -- this type of anabolic agent you'll have a different bone volume, different amounts of actual bone tissue and different degrees of mineralization within the bone tissue.
So just lumping them together may -- it's nice to -- for certain purposes, but it doesn't explain everything. And we have to realize that this is somewhat of a -- is an imprecise measure of what's going on in the structure.
CHAIRMAN BRAUNSTEIN: Dr. Marcus.
DR. MARCUS: Thank you. I'd like to reemphasize what Charles Turner just said, because I agree with him fully. And in fact, there's been some ambiguity in some of the presentations that have been made.
For example, Dr. Colman first stated that BMD has now "risen to its proper place." I'd respectfully like to disagree with that. I think, actually, there's been more questions raised about BMD within the last few years than maybe we had before.
When you look at what happens when you raise BMD there are a multiplicity of ways in which BMD can be raised. With antiresorptive drugs, certainly during the first period of several months when these little jackhammers, as Charles describes them, the resorption bays are being filled in, that does represent a true increase in the amount of bone tissue.
But subsequent to that, the secondary mineralization does mean that you gather more and more mineral over time just because the activation of new remodeling units to come and clean that up is much reduced.
Another way to increase BMD could be to increase the number of trabeculae without -- prior to anabolic therapy. That's something that has never been known to occur because the number of trabeculae are set in utero, actually before birth.
You could increase trabecular thickness with normally mineralized bone, and that would increase BMD, but that has not ever been observed to occur with antiresorptive therapy. Now, we have the advent of anabolic therapy.
And whereas, you might think it would be a slam dunk that under all circumstances anabolic therapy with definition pari passu increase BMD, we have now learned that that actually is not always the case.
For example, there was presented this week at the Bone and Mineral Society a very interesting study of growth hormone, which showed that early on in growth hormone therapy there's actually an apparent reduction in BMD.
Now, we always used to think that was due to increasing the remodeling space by opening up new resorption bays, but in fact, this was a study from Denmark which was a very careful histological study, histomorphometric study, that showed that that actually didn't happen.
What was happening was that growth hormone was laying down new bone, but early on in that bone's life it is relatively undermineralized. Remember, BMD is an artifice. It is a compound number which represents the bone mineral content divided by the area.
So by expanding the area by increasing new bone, but that bone not being as well mineralized as mature bone, there was actually the appearance of a reduction in BMD. And under certain circumstances it appears that a similar sort of thing happens early on in the treatment with Teriparatide.
Therefore, I think that depending on BMD is really fraught with a great deal of danger. The second point I want to make where there was ambiguity had to do with this term, "bone quality," which strikes me back to what I understood from my house officer days, is the meaning of the word idiopathic.
It's so vague as to be almost useless. And in fact, we had an awkward situation on this very panel when I was a member dealing with the presentation of the delayed release fluoride. Some of you were also on this panel at the same time, because measures of bone quality were used and introduced which really weren't highly validated and generally accepted as standards of measurement for the community.
Therefore, I think that what we have to do, and as a recommendation to this panel and to the Agency, is to come up with specific validated parameters of bone quality, rather than just talking in sort of vague terms, such things on biopsy as cortical thickness, connectivity index, an index reflecting the percentage of plates versus rods.
Those are all validated, statistically robust measures that should be -- you should ask for listing up front for biopsy data. And furthermore, now that we have MRI and synchrotron, other noninvasive sorts of approaches to looking at structural parameters in studies, we should also encourage you to have specified a certain number of those parameters that are highly validated also to be called on, rather than just using the general term, we're going to look at bone quality. Thank you for indulging my mind.
CHAIRMAN BRAUNSTEIN: Thank you. Well, we'll be discussing more of these things this afternoon. Dr. Cummings is next, then Dr. Khosla and Dr. Watts.
DR. CUMMINGS: I have to agree with Bob that the changes in bone density and subsequent changes in fracture risk have actually gotten to be much more interesting as one of the meta analyzers that Mark referred to.
We have looked closely at this data and one of the things that sort of magnifies this paradox is the fact that there have now been reported 60 to 65 percent reductions in risk in the first year, when the bone density changes are even less.
And the analyses that Mark and our group reported were for the aggregate of three years. If you try to do this for one year you would find that the discrepancy is much, much greater, but still, a gross underestimate.
Bone density increase is still -- it doesn't correlate very well, but it's just grossly underestimating the risk. And it's not clear how well it's predicting risk of fractures beyond that, you know, if you just look at second, third and fourth year because there, the reductions in risk of vertebral fractures are less.
Bone density is continuing to accrue. And so I think if we look at it more closely it is complex and it's not straightforward that there is not the clear-cut statistical relationship that we reported between change in bone density and change in fracture risk when you take time into account.
And then things really seem to change -- this relationship seems to change a lot over time, and that was it.
CHAIRMAN BRAUNSTEIN: Thank you. And Dr. Khosla.
DR. KHOSLA: Well, I guess I agree with all of the comments that have been noted about the caveats with bone density, that you know, there are individual clinical situations where a patient, maybe on therapy, may not have a change in bone density but has still benefitted from the drug.
But I guess the -- you know -- just stepping back, it's pretty clear that if you use BMD as some sort of a surrogate, that it's actually a very conservative bias, because you're actually vastly underestimating the potential benefit from the antiresorptive drugs.
So it's not like you're going to go wrong and overestimate the potential benefit. It's actually going to be a fairly significant underestimate of the benefit that you're going to -- you may get from that particular antiresorptive drug.
And actually, if I could just make -- because I wanted to ask Steve another question to follow up is that, have you or Mark actually combined, you know, looking at BMD changes and bone marker changes into a more global model to see if in a combination they may actually come closer to predicting the reduction in fracture risk with these drugs?
DR. HOCHBERG: Well, we tried to do that in the models to estimate reduction in nonvertebral fractures, and we couldn't get the regression models actually to work. And I think that was -- that's because of the of cross-studies there is a very, very high correlation between the reduction in bone turnover seen with the antiresorptive agent in that study compared to placebo and the increase in bone mineral density, which is seen in that study compared to placebo.
We actually found our square values of between .8 and .85, which are higher than the -- you know -- the sort of no expected correlations that Dr. Faulkner mentions of, you know, .6 to .7, for the reduction in bone turnover and the increase in bone mineral density.
So because of the very high correlation between the two, we couldn't force both into a single model.
CHAIRMAN BRAUNSTEIN: Dr. Watts.
DR. WATTS: I had notes on all of those points that I'd like to elaborate on just slightly. I think that it is more complex than bone density alone. The addition of anabolic agents makes it even murkier.
With teriparatide the increases in bone density were 50 percent or 100 percent larger than what was seen with antiresorptive drugs. Yet, the reduction in vertebral fractures was in the same order of magnitude.
We know that antiresorptive drugs have roughly a correlation between suppression of bone turnover and increase in BMD. The suppression of bone turnover occurs early. The rise in BMD continues to accumulate over years.
And as Dr. Cummings has pointed out, the reduction in vertebral fracture, at least numerically, is greatest early rather than late. The turnover and density changes are so linked that it's probably impossible to separate those out.
Now, Dr. Faulkner showed us the measure of density with DXA, which is the standard for these trials, measures both cortical and trabecular bone, and therefore, might underestimate changes in the critical component of the skeleton, trabecular bone being more metabolically active and preservation or destruction there being more important for maintenance of bone strength.
So it may be that there are better ways that we could look at bone density, independent of turnover. But I don't think for a minute that bone density, at least for me, serves as an adequate surrogate, even for the antiresorptive drugs, much less for drugs that might have a way of laying down new bone or changing the geometry of bone that would also have important structural implications.
CHAIRMAN BRAUNSTEIN: All right. Thank you. One last question, Dr. Grady.
DR. GRADY: Yes. I think these studies of the association of change in bone density and change in fracture risk are important for -- really, probably key for our consideration. So I hate to be dense, but I just want to ask -- I just want to understand how this was done.
So we have sample sizes of somewhere on the order of 13 to 18 or 19, right, and you looked at a predictor of univariate regressions, predictor variable of continuous outcome. And just in eyeballing those, they look pretty heterogeneous.
I wonder, number one, did you do -- you know -- did you do a formal test for heterogeneity, and -- they're homogeneous?
DR. CUMMINGS: Sufficiently that they could be pooled, the 13 of them.
DR. GRADY: Okay. So the sample sizes are rather small?
DR. CUMMINGS: They're -- yes, correct.
DR. GRADY: Okay. And secondly, I wonder, so those were --
DR. HOCHBERG: Can I -- you mean, the sample size in terms of the number of studies that are included?
DR. GRADY: Yes. That's the sample size when you're doing your regression analysis.
DR. HOCHBERG: Okay.
DR. GRADY: And secondly, did you look at any other variables in those models like, for example, age, you know, age since menopause, baseline BMD, lots of interesting sorts of additional --
DR. CUMMINGS: No. That information is often missing from the reports of trials. So it was just a heroic effort to be able to do -- just get the bone density that was sometimes variously reported. And the confidence, one thing you didn't mention is that there's a lot of variability around that.
The confidence limits around the relationships, mathematical relationships that Mark showed exclude no relationship, but you know, they're pretty wide because of that heterogeneity. That's why I don't think you could use the equations to predict the reduction risk of vertebral fractures from these meta analyses without doing a trial.
CHAIRMAN BRAUNSTEIN: Dr. Marcus had one addition on that.
DR. MARCUS: I just had one tiny point. This shows you that you can pool some of the people some of the time.
(Laughter)
DR. MARCUS: That was not the point. The point was --
CHAIRMAN BRAUNSTEIN: Can I --
DR. MARCUS: -- addressing the issue -- just one tiny BMD issue. All of this presumes that the agent you are using is not itself changing the mineral structure of the bone. And I must point out that fluoride, which creates a larger molecule, introduces not just BMD as an artifice, but an artifact, as well as I saw one agent that has not yet been -- shown its head in this country that has apparently -- is on the books in Europe, and that is strontium, because that will do the same thing.
You will see an artificially high BMD, which just represents a fact of the incorporation of a heavy metal into the bone, just like if it were lead.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Hochberg.
DR. HOCHBERG: Can I make a brief comment? Okay. First thing, in response to Dr. Grady, we did not adjust these models, as Steve said in his analyses, for age or baseline bone mineral density. We -- in our paper we actually reported the characteristics of the populations in the trials, but in the absence of having patient-based data from all the companies or the authors which sponsored the trials, we didn't do that, and that's a limitation, at least, of our analysis.
But recognizing that, you know, some agents in fact do artifactually change bone mineral density, none of these analyses include the more recent data for strontium ranelate, which was presented at the World Congress, I guess in May, or the older data from fluoride.
CHAIRMAN BRAUNSTEIN: Okay. Great. We'll move onto the public -- open public hearing now, and we invite the individuals who are going to speak to please come up to the microphone in the center there. There's two written submissions that are available outside on the desk from GlaxoSmithKline and from Roche Pharmaceuticals.
Our first public speak is going to be Dr. Ginger Constantine, Vice-President Women's Health Research, Wyeth, and we ask all the speakers to please not only identify themselves, but identify if they have any conflicts of interest or potential conflicts of interest.
And if the speakers could speak from the middle, we'll show the slides up here. Thank you.
DR. CONSTANTINE: You're pointing in different directions, which is part of the conversation here, I guess. Hi. I'm Ginger Constantine.
I am a representative of Wyeth Pharmaceuticals and I would really like to thank the Advisory Committee for allowing us to discuss some of the challenges from a pharmaceutical perspective company in developing these products, and also offer some suggestions for future development.
Now, I don't know how to flip the slide from back here. Thank you. On this slide is a list of eight different things which are predominantly the stumbling blocks to development for pharmaceutical companies, and these have predominantly been discussed this morning. So I won't bore you with all of them.
Obviously, the IRB and Ethics Committee -- obtaining approvals from IRBs and Ethics Committees have been a predominant issue. Country variability is -- has also been quite difficult in light of the global nature of trying to perform studies.
Trial size and cost are huge stumbling blocks. The complexity of the protocol for testing procedures, oftentimes depending on the compound that's being developed, may span several divisions within the FDA and concurrent sometimes on individual factors may be difficult.
Enrollment and retention of subjects, especially with longer-term trials, has become quite difficult and it's obviously very difficult for the physician, as well as the patient. On the next slide you'll see the Wyeth position with regarding to these issues.
We do feel that placebo controlled trials are most efficient and reliable. We do feel that active reference drugs provide important therapeutic context and that protocol requirements and guidances need to be the same worldwide, if possible.
Patient testing requirements need to be practical and trials need to be short enough to allow for high patient retention. And this is a huge issue, especially when we go on to do analyses. Next slide, please.
So in light of this we have some suggestions for existing compounds -- of compounds which would include SERMs, estrogen and bisphosphonates. With a primary endpoint we would suggest statistically and clinically significant change in BMD for vertebral and nonvertebral fractures, and perhaps enroll subjects with a BMD of minus 2 to allow for our placebo controlled trial.
A secondary endpoint would be reduction in the incidence of vertebral or nonvertebral fracture, with bone histomorphometry in a subset of patients demonstrating good bone quality. I'd like to just add the caveat that the first point certainly would be dependent on an adequate preclinical package showing good preclinical models for bone development and bone strength.
We would like to suggest a two-year trial for the durability of effect and to look at adverse events. And obviously, we would have to design an adequate safety database so that all of the safety issues that would come up could be adequately addressed.
I would like to thank the Advisory Committee for allowing us to present this position, as well as thank specifically the M&E division for their efforts in this and realize how difficult it is to address these challenges.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Orloff.
DR. ORLOFF: We need a -- before you go, we need a clarification on your primary endpoint. We don't understand -- you mean, to support an indication for the reduction in risk for vertebral and nonvertebral fractures, BMD alone?
DR. CONSTANTINE: Yes, with these other things.
DR. ORLOFF: With the secondary.
DR. CONSTANTINE: With all of these things.
DR. ORLOFF: Okay.
CHAIRMAN BRAUNSTEIN: Thank you. Our next speaker is Dr. Dere, Vice-President, Endocrinology, Lilly.
DR. DERE: Chairman Braunstein, Dr. Orloff and members of the Advisory Committee, Lilly commends the efforts of the Agency to provide a forum for discussion of this critical clinical topic. During the past years the FDA has approved a number of new agents for the prevention and treatment of osteoporosis.
The drugs were approved with heavy emphasis on the existing 1994 draft FDA Guidelines for the development of osteoporosis therapies. These Guidelines were developed when there were few options available to the medical community to treat this potential debilitating disease.
It is now time to develop new guidelines which must take into account -- consideration advances in medicine and science and the current climate of drug development. These guidelines must take into account workable strategies for testing and registering osteoporosis therapies for women and men with osteoporosis of various etiologies.
We offer the following points for consideration as you continue your deliberations today. Number one, there is a need to define a common standard for demonstration of efficacy that can be applied to drugs of different classes.
Lilly believes that while BMD is a useful diagnostic to identify those at risk for osteoporosis, we maintain that the change in BMD and in biochemical markers of turnover are not suitable to replace fracture as an endpoint for evaluation of efficacy of a new chemical entity.
The relationship between the change in BMD to that of a reduction in fracture risk is not the same for different classes of therapy and accounts for only a small part of the observed fracture risk reduction.
Lilly agrees with the current recommendation that a reduction in vertebral fracture risk is necessary to prove efficacy for osteoporosis compounds in order to obtain a treatment indication.
Using surrogates for vertebral fracture endpoints would make it difficult to establish the true antifracture efficacy of new drugs and would result in less informative and less competitive labeling for sponsors with new drug development programs.
However, we agree that treatment induced change in BMD remains an acceptable endpoint for new formulations and indications such as glucocorticoid induced osteoporosis and male osteoporosis for compounds whose fracture efficacy has previously been established.
Number two, while we recognize that a number of osteoporosis therapies are now available, Lilly maintains that a randomized controlled trial using calcium and Vitamin D for all patients should remain the standard for establishing efficacy and safety.
In the current environment there is a dilemma regarding the acceptability of these so-called placebo controlled studies for evaluation of compounds for treatment of a disease for which alternate treatments exist.
However, a relatively small placebo controlled study that clearly demonstrates superiority of a new drug over placebo may be more broadly useful and more ethical with respect to the number of patients exposed, than a larger study against an active comparator.
The European CPMP Guidance on osteoporosis drug development that was issued in 2001 states that: "Although active control trials are preferred, placebo controlled trials are still acceptable. Placebo controlled trials provide greater flexibility in study designs, for example, the use of escape clauses and stopping rules to maximize patient safety and use of add-on therapies, and should be considered for drugs in development."
Number three, there are considerable challenges in conducting active comparator trials rather than placebo controlled studies. For example, these include: a lack of access to data other than that present in the public domain for the active comparator may hamper elucidation of statistical and sample size estimations for hypothesis testing.
Also, noninferiority trials would require exposing a larger number of patients in potentially longer clinical studies. Next, trials designed to establish either noninferiority or superiority of a drug compared to an established therapy might be compromised due to the difficulty in replicating the effectiveness of the comparator active therapy, depending on the population studied and the conditions of the trial design.
Without a placebo controlled group one could not know whether the active compound had worked -- comparator had worked or not. Next, if an active comparator were required, how would a sponsor determine which therapy is best for comparison, given that different classes of osteoporosis therapies work by different mechanisms, have different pharmacokinetic profiles and even different target populations.
And finally, there may be a lack of understanding of the safety profile because the true adverse event rate for a new drug is best derived from placebo controlled studies. Number four, as I stated, Lilly maintains that the most appropriate study endpoint is the reduction in the incidence of osteoporotic vertebral fractures.
While demonstration of reduction of fractures at the hip is not required by current guidelines, guidance is needed for the purpose of label language on ways to be able to demonstrate efficacy at the hip.
It is not practical to limit studies specifically to hip fractures. For example, to demonstrate a 40 percent reduction in the incidence of hip fracture assuming a three percent event rate, the number of patients required for a placebo controlled trial is 5,000.
And for an active controlled noninferiority study with a 20 percent margin of noninferiority, the number of patients required is 33,000. And for an active controlled superiority study the number of patients required would be 40,000.
Therefore, we propose that a reduction in combined nonvertebral osteoporotic fractures, an increase in hip BMD and improvements in bone structural measurements such as those describe by Thomas Beck and colleagues from DXA scans should be considered adequate to demonstrate substantial evidence for a hip fracture reduction claim.
Number five, guidelines should provide for the acceptability of shorter duration clinical trials, such as 12 months with a vertebral fracture endpoint for an antiresorptive, and possibly shorter for anabolic agents, provided preclinical studies clearly show no detrimental effect on bone quality and sufficient safety data will be accrued during follow-ups such as in post-marketing surveillance programs.
While further guidance is needed on the number of years of follow-up required to assess clinical safety and durability of effect, we believe that a total exposure of three to four years should be considered appropriate for safety evaluation.
Number six, current guidelines do not consider histomorphometric parameters of bone biopsy as efficacy endpoints. Given the lack of treatment effect, i.e., fracture reduction predicted by changes in BMD alone, the Agency should consider accepting the use of advanced imaging and computer-based analytical techniques for demonstrating changes in bone micro architecture and quality.
For example, -3D analysis of bone structure using micro CT might provide efficacy measures of bone quality and structure, and could be used to define and distinguish true anatomical differences of different classes of osteoporosis therapies.
For the purpose of human studies, bone quality may be assessed by appropriate combinations of bone mineral densitometry, specialized radiographic techniques in vivo and in vitro, such as micro CT, spiral CT and MRI, and histologic assessments of trabecular and cortical bone mass, cortical thickness, trabecular connectivity and bone remodeling.
Sponsors should be encouraged to consider new assessments for bone strength that could include bone quality and architecture during clinical development.
Number seven, with the availability of a variety of therapeutic options, drugs are likely to be used for the treatment of osteoporosis in a number of ways, alone or in combinations. Guidance is needed to support claims for sequential or combined use of osteoporosis agents with the same or different mechanisms of action.
And finally, there will be a critical need for harmonization of guidelines between the various regulatory agencies to provide for similar registration requirements across countries. Divergent guidelines will make registration of new osteoporosis therapies needlessly expensive and difficult.
We believe that the points raised are critical. If they are not dealt with, patients needing advance therapies may suffer as important advances are delayed or prevented from ever reaching them.
We anticipate that the Division and the Committee will continue to encourage constructive dialogue with industry today and in the future. Additionally, it will be important to keep communications open with the CPMP and with public health agencies such as the NIH to address these critical questions and to provide recommendations for workable new guidelines for developing osteoporosis therapies. Thank you very much.
CHAIRMAN BRAUNSTEIN: Thank you, Dr. Dere.
The next speaker is Dr. Thomas Marriott, Vice-President, Development Research, NPS Pharmaceuticals.
DR. MARRIOTT: Good morning, Mr. Chairman. Thank you very much for giving us the opportunity to make some comments. We certainly appreciate the work the Committee is doing in tackling this difficult area.
This morning's speakers have done a good job of outlining many of the points and much of the data that need to be considered in the design of clinical trials and the clinical evidence necessary for the approval of new osteoporotic agents.
The summaries of the Guidance documents also make it clear that there's an urgent need for an ICH-like harmonization of the meaning of the terms "prevention" and "treatment" as they relate to osteoporosis and of the regulatory requirements for the approval of new agents.
At this point NPS still believes that the randomized double-blind calcium and vitamin D controlled trials are the best way to evaluate the safety and efficacy of new osteoporotic agents. However, our recent experience suggests that it is becoming increasingly difficult to conduct calcium and vitamin D controlled trials.
We are currently in the middle of a 2600 patient randomized calcium and vitamin D controlled trial in nine countries. In 2000, when we were initiating the study in the U.S. and Canada, several IRBs refused to approve the study because we included women with severe osteoporosis; that is, women with a BMD of less than minus 2.5 and a prevalent fracture.
In 2001, as we expanded the study worldwide, two of the multiple research ethics committees, the MREX in the UK and the Central Ethics Committee in Denmark would not approve the study because they considered it placebo controlled and requested that we add an approved agent to the calcium and vitamin D control group.
Thus, if we're to continue to employ calcium and vitamin D controlled studies, the scientific and regulatory communities must clearly describe why this study design is appropriate and better than alternative study designs.
We must also demonstrate that we have reduced the risk to our patients as much as possible. We suggest that there are at least three ways to minimize the risk to our patients. First, it's possible, we believe, to reduce the number of clinical studies.
Harmonization of the definitions of treatment and prevention may allow both indications to be investigated in a single trial. An obvious example would be the study of a true anabolic agent where a reduction in fracture incidents is demonstrated and virtually all patients show an increase in BMD, beginning in the osteoporotic range and increasing through the osteopenic range.
A second study to specifically investigate prevention divined by an increase in BMD should not be necessary in this case. Secondly, we would believe that it should be possible to reduce the recommendation -- the recommended duration of clinical trials.
It is clear that it is possible to demonstrate statistically significant increases in BMD in short periods of time with many agents, and statistically significant decreases in the vertebral fracture, incidence, for example, in less than three years.
The recommended duration of efficacy studies required for approval, we believe, should therefore be considered and in fact should be less in fact than two years. We believe it's also possible to reduce the number of patients in clinical studies.
There are two ways, at least, to reduce the number of patients participating in clinical trials from the osteoporotic agents. The first is to use the one-sided test to determine efficacy when the control group is calcium and vitamin D.
For example, it is obvious that a treatment would need to demonstrate better efficacy than that of calcium and D. Therefore, the null hypothesis is whether the incidence of fractures in patients receiving the experimental treatment is lower than the incidence of fractures for patients receiving calcium and vitamin D, not whether there is a difference in the incidence.
This question can be answered using a one-sided T-test or one-sided test, and the use of the one-sided test should in fact reduce the number of patients in the trial by 15 to 20 percent.
A second way to reduce patient numbers is to accept a lower level of confidence, for example, 80 percent, for the reduction in fracture incidence at a second fracture site once reduction in the fracture incidence at the first site has been demonstrated.
Since the significance level is the risk of concluding a difference exists when in fact there is no difference, the level of significance is chosen based on the consequences of this decision. Therefore, if a treatment has been demonstrated to reduce the incidence of vertebral fractures, for example, the question of whether it also reduces the incidence of fractures at another site, for example, the hip, should addressed using a lower level of confidence.
This does not substantially increase the risk of concluding that an agent with a deleterious effect at the second site is better than the control, but will require fewer patients and fewer fractures at the second site to reach the appropriate conclusion. Thank you for your time, and again, I thank you for your efforts in taking on this task.
CHAIRMAN BRAUNSTEIN: Thank you, Dr. Marriott.
Our last speaker is Ms. Amy Alina, from the National Women's health Network.
MS. ALINA: Hi. I'm speaking here on behalf of the National Women's Health Network, which is a nonprofit organization that advocates for national policies that protect and promote all women's health.
We also provide evidence based, independent information to empower women to make fully informed healthcare decisions, and the network does not accept financial support from pharmaceutical or medical device companies.
We're supported by a national membership of about 8,000 individuals and 300 organizations. We're here today representing the concerns shared not just by our members, but also by millions of women who, particularly in the wake of the news this summer about the Women's Health Initiative results, are really struggling with questions about the safety, effectiveness and the need for drugs prescribed at menopause.
And while the topic of this meeting is clinical trials for new osteoporosis treatments, it touches on issues that go far beyond clinical trials and affect the way that women are educated about bone health, screened for bone density loss, counseled on prevention strategies and finally treated for osteoporosis.
We recognize that this Committee and the FDA do not control all those aspects of women's healthcare, but we address them in our comments because the way that clinical trials for osteoporosis drugs, particularly prevention trials, are designed will have consequences for women's health education and care.
In the 1980s and earlier we were among the women's health advocates who agreed that the problem of bone fractures and their effect on elderly women's quality of life was being overlooked by the medical community and needed to be addressed.
And today, we think the pendulum has swung to another extreme for those women who do have access to healthcare and insurance coverage. Now, we believe it's the case that women who are in the healthcare system are commonly over-treated -- over-screened, over-diagnosed and over-treated for problems relating to their bones.
At the same time it's still true that there are women who would benefit from screening and treatment who don't get the care they need as a result of economic and other barriers to health services. But bone density screening has become a rite of passage for women approaching and entering menopause, and this means that women are being screened in their `40s and `50s, which we believe is far too early to use a test that hasn't been shown to be a reliable predictor of fractures that typically occur 20 to 30 years later.
And this is a problem, though many people might ask what the harm is in taking a measure of bone density. The assumption that osteoporosis screening must be a good thing fails to recognize its limitations or how it plays into the medicalization of menopause.
So the problem is that over-screening leads to over-treatment, and many of you who see patients must hear, just as we do in our office, from women who tell story after story about how they've been told that they have the disease of osteopenia, that they need a prescription for their borderline osteopenia.
And once women are diagnosed in this way they may be much less likely to do many of the things that could help them maintain their "borderline bone health," like staying physically active. So we're very concerned about that.
And we're also concerned about the fact that many of these women are given prescriptions and told that they have to take drugs to prevent their osteopenia from developing into osteoporosis, and then leading to bone fractures, the slippery slope.
Some of them may need help from a drug to prevent serious bone loss and debilitating fractures, but some of them don't, and the bone density test is not a sufficiently reliable predictor of fractures to support that use of it.
The experience of hormone replacement therapy should serve as a warning, we think, as an example of a drug that was prescribed to millions of women based on false assumptions about unproven benefits and inadequately tested safety.
So how does this relate to the discussion issues that you have to address? Clinical trial design, as I said, doesn't control clinical practice, but it does have an affect on it.
And as you think about the answers to the questions that the FDA staff has posed to you, we urge you to put them in the context of how the clinical trials of new osteoporosis treatments will affect the way that the drugs tested will be put into use in clinical practice and the way they will therefore, affect women's lives.
I'm going to respond to a couple of the specific questions, first on efficacy. When is bone mineral density an adequate primary endpoint? Well, my guess -- our answer to this question is never, and we recognize that this puts us somewhat outside the mainstream of discussion.
But we're not alone in questioning the value of bone mineral density measures, and it's already been pointed out this morning that the NIH Consensus Conference two years ago, the report itself raised questions about the accuracy of bone mineral density testing and recommended that more comprehensive ways of assessing risk for fracture should be studied, and so we're echoing that.
On the question of duration of study, what duration of study is appropriate for assessment of effectiveness, we link this very much with the age of women included in the trials.
If study durations are going to continue to fall in the two- to three-year range, or certainly, if they were to be shortened we think it's important that the prevention trials not be conducted on women who are younger than 65, unless those women are at particularly high risk for bone fractures, because of, you know, early removal of ovaries or long-term steroid use.
It isn't possible to determine in the short term of a two- to three-year trial whether a drug has effectively prevented bone fracture in a younger woman who hasn't yet reached the age at which she's likely to experience bone fractures.
On the question of using a placebo or an active control, the new understanding of the risks posed by hormone therapy significantly changes the terms for discussing this issue.
While the Women's Health Initiative showed us that estrogen plus progestin is highly effective for osteoporosis and will likely show the same for estrogen alone, it also demonstrated that the combined hormone regimen poses serious health risks which outweigh its benefits for healthy women.
So we would say that we can't hold out HRT to be the standard comparison for a trial of the new osteoporosis. And in prevention trials we still believe it's appropriate to use a placebo control by which we mean vitamin D and calcium.
In treatment trials, however, where participants have experienced a fracture prior to beginning in the trial, we think an active control is both ethical and appropriate, and we would also say it's desirable, because it will provide more valuable and useful results showing whether a new drug offers a benefit over existing options in terms of either efficacy or safety.
On the safety question specifically, the instance of osteoporotic fractures being used as a safety rather than an efficacy endpoint, well, I guess we would say that it should be used as both a safety and efficacy endpoint in prevention and treatment trials, and reiterate our statement that the intermediate endpoint of bone mineral density isn't an adequate measure.
And in terms of duration of study needed for assessment of safety, we recognize it's not practical to require sponsors to conduct trials that last ten or more years. I'm sure the speakers who went before me would agree with that.
But women who are prescribed drugs for osteoporosis are likely to be taking them for decades. And so we would say it's necessary to gather data on safety of such long-term use and that FDA should recognize this need by making long-term follow-up studies on these products a condition of approval, and by putting in place active systems for monitoring adverse reactions to the drugs. Thank you.
CHAIRMAN BRAUNSTEIN: Thank you.
(End of this portion of proceedings; 12:40 p.m.)
A-F-T-E-R-N-O-O-N S-E-S-S-I-O-N
(1:32 p.m.)
CHAIRMAN BRAUNSTEIN: Our next speaker is Dr. Steven Cummings. He's going to discuss the size, scope, and implications of placebo versus active control trials.
DR. CUMMINGS: Thank you. I'd like to thank the committee, particularly Eric Colman for his invitation to come and talk to you about this set of issues.
I've probably got the most diffuse task of the day, and that's to cover a lot of issues and hopefully stimulate some discussion about a few important issues and principles that I'd like to propose.
To start off with, I'd like to acknowledge a lot of help from a couple of other people in putting this together. As I am not a biostatistician, I'll be showing a lot of numbers that were generated with the help of two statisticians: Dr. Charles McCulloch, who is Chief of Biostatistics at UCSF, as well as my colleague Dr. Dennis Black, who has been involved in the design and implementation of a number of trials in the area of osteoporosis and fracture prevention.
What I'm going to cover first will be just a few comments about the rates of fractures and risks in placebo-controlled trials as a background and a sample size that we are typically using for such trials.
And then we'll move on to the alternative that I've been asked to discuss at more length, which is the non-inferiority trials designs, the samples sizes for trials that have bone densities as an endpoint and vertebral fracture outcomes, or vertebral deformity outcomes as I'll call them as an endpoint. And, we'll propose a principle for how to set non-inferiority margins, which are such a critical assumption in that.
And then finally, I'd like to make a couple of comments and a proposal regarding the duration of trials and why consideration of the duration of trials for fracture prevention may be an important issue for this panel to consider in reframing guidelines.
There have been a number of discussions about whether or not placebo-controlled trials remain an alternative for the testing of new drugs for the prevention of fractures. It really in many ways boils down to both the risk and your perception of the risk of being in a placebo group instead of receiving standard therapy.
And when we talk about osteoporosis patients, it's very important to realize that we're not dealing with just osteoporosis as a single group. It's a very heterogeneous group of people with heterogeneous risks. I would like to point out how we might in designing trials begin to draw finer distinctions between people that may allow us to do placebo-controlled trials in a more ethical fashion.
For example, a woman who has a vertebral fracture has about a four-fold greater risk of suffering another fracture, vertebral fracture and other fractures than a woman with just low bone density or osteoporosis defined by her densitometry measurement at the hip or the spine.
It's a very important principle, but we've also been able now to distinguish even finer gradations of risk among women who have vertebral fractures. And I'll point out that it makes a difference whether it's recent or whether it's an indeterminate age. It makes a difference whether there are multiple.
And although there is less data on this last point, and perhaps some of my colleagues can help me about some of the unpublished data, it makes a difference about how severe these fractures are in terms of the risk to the individual patient. So, that database is just starting to develop and I think it will be very important for designing trials.
The annual risk of having a new vertebral deformity on which we base our sample size estimates in trials -- and I'll return to in discussing alternative designs -- are in this order for new vertebral deformities. That means new radiologic events rather than clinically apparent painful events. Defined as Ken Faulkner described earlier, those with a hip density that's in the osteoporotic range have about a one to two -- in one trial up to about three percent per year risk of having those events, depending in part how they're defined.
Patients who have a vertebral fracture in an undifferentiated way have between a five and ten percent risk of suffering a vertebral fracture per year in the existing trials that have been done to date. Now because these drugs have generally reduced the risk by about 35 to 50 percent, then you can go through those and find out what the comparative risks are for someone in the placebo group compared to someone in the treatment group.
I'll return to the limited activity days at the end. But, the risk of having a spine fracture during the course of a trial on an annual basis, again a radiographic vertebral fracture depends on whether you have a spine fracture or you just have osteoporosis according to the densitometry machine. So if it's just a densitometric osteoporosis, your risk per year of suffering a vertebral fracture is on the order of one percent. And if you start off with a spine fracture, it's on the order of two to three percent per year.
I'll return to limited activity days later. However, as I was pointing out earlier, not all fractures are equal. Bob Lindsay pointed out a couple of years ago in a very nice article that defining fractures somewhat more liberally so you get a somewhat higher incidence -- in the VERT trial, there's a 15 percent reduction in risk. He pointed out that a woman who'd had a fracture in the last year had about a 20 percent fracture risk in the following year.
We've gone back to two other databases that we have, the FIT trial and the MORE trial, and have confirmed that these women have about a four to five-fold greater risk of a subsequent fracture than women whose fracture is old or of an indeterminate age.
That means to me, and roughly estimating this, that the women who have a recent fracture have about a ten percent per year. Just a rough estimate, about a five to ten percent per year are at risk of suffering another radiographic event. About a third of those will be clinically evident, diagnosed as clinical vertebral fractures.
These limited activity days on the right point out the number of days that we estimate from the FIT trial that an individual is disabled as a consequence of having a fracture or is a result of back pain.
And we've estimated that a woman who has a spine fracture, for example -- not differentiating the recently acute or multiple ones from the indeterminate ones -- those who have a spine fracture on average have about seven days of limited activity per year. In the placebo group, it would've been prevented by treatment. That's about seven days.
But for recent fractures, this probably amounts to on the order of weeks, two, three, four weeks that would be preventable by taking standard treatment instead of being in the placebo group of a trial.
One of the other issues that comes up is the risk of death. I have in many epidemiologic talks, as all my colleagues have, have said over and over that hip fractures are associated with 12 to 20 percent risk of dying in the first year. And those who have vertebral fractures have an increased risk of mortality from trials.
It turns out, of course, that those are epidemiologic associations. The reason, by close chart review or other kinds of methods, the reason that most people die after hip fractures is because they have other diseases like cancer that lead to the fracture and then cause the death. And when you get down to actually estimating how many deaths are attributable to the fracture, it's a much smaller number but very hard to figure out.
So, I've gone back to the trial databases to try to figure out whether prevention of fractures prevents death. I think that's an important thing to know.
In the fracture intervention trial, we went back to all the 6,459 women who were in that trial and suffered. Nine hundred and seven women suffered 1149 fractures, and there were 122 deaths. And, we couldn't find in the database a single death due to the fracture.
Now these are healthy women. They don't have other comorbid conditions because they'd been screened out for such as you would normally do in a trial. But I think that's an interesting statistic.
We've also then gone back to all of the trials and pooled all the mortality rates in the existing trials that you've seen, the major pivotal trials, to see if we could find evidence that reducing fractures in some way reduced overall mortality.
Again, this is not fracture related mortality because those are not reported in the papers. But in no single trial was there a reduction in risk of mortality that was statistically significant in the dozens of thousands included in our overall poolings. Again, it fails to find a statistically significant reduction of risk of mortality due to participating in the placebo group of a placebo-controlled fracture trial.
One other area is quality of life. We have relatively insensitive methods of measuring that, just questionnaires given once every six to twelve months in a couple, not all of the trials. We're fortunate that Merck has allowed us to have access to the database from the FIT trial to begin to look at some of these issues.
One of the things I did was take a look at the SF-12 quality of life instrument that was measured at baseline and then at the end. It has six subscales of functional status. What we found is that the change in functional status, that the pain in other domains in fact did not differ significantly between the placebo and the alendronate group on any measure from beginning to end of trial.
So it's not to say that osteoporosis is not an important condition, but that, in terms of global changes and quality of life for an individual participating in the placebo group, there doesn't appear to be a substantial risk of deterioration in overall quality measured this way.
We've measured it in other more sensitive ways, and that is to count the number of days, again, of limited activity or back pain that sent you to bed. Amongst women with vertebral fracture from the FIT trial, we've estimated that again you can see an aggregate here that they spent on average an extra day in bed compared to those in the alendronate group. And, they have additional days of disability that is limited activity that add up in total to about a week of limited activity or bed rest as a consequence if they're in the vertebral fracture group.
But remember, not all vertebral fractures are equal. This gamushes together all of those with recent and multiple and single indeterminate aged fractures. I haven't been able to distinguish those amongst who have just osteoporosis by density of the hip. We have a reduction of days in disability due to a fracture, but we haven't been able to find statistically significant reduction in other measures of days of limitation.
Let me just summarize. The risks of being in a placebo group in past trials of effective drugs, on average have produced limitations of activity due to the fracture in the bisphosphonate trial, the alendronate trial -- I don't know the data from the Risedronate trials -- has not measurably reduced quality of life as measured by questionnaires and doesn't significantly increase mortality. Most importantly, however, even the risk of disability depends very much on the degree of severity of osteoporosis.
So, placebo-controlled trials are becoming much more difficult to do. And many of us involved in those know that it's hard to recruit people from the United States because so many of them want to be on alternative drugs, and their doctors resist, and IRBs are difficult at times.
I want to make one other point. The placebo-controlled trials and women with osteoporosis have sample sizes. The ones that are out in the field now that I know about had sample sizes between 2,000 and 8,000. Towards the end, the 2,000 end, if the only endpoint is vertebral deformities in very high-risk women. Towards the right end, more towards 8,000, if they have more ambitious goals such as reducing risk of fracture in the first year or eventually trying to find a risk of reduction of hip fracture.
So with that as background of the current state of placebo-controlled trials, let me move on to non-inferiority trials. We're testing new chemical entities versus actual comparators, and I'm going to divide this into two parts. As we go through, I'll focus on bone density and then outcomes of vertebral deformity.
I'm not going to talk about hip fracture endpoints because the numbers will get really big. But BMD and vertebral deformity will be the two outcomes I cover. And, I'm going to address bisphosphonates for which we at least have adequate data to do some of this.
The way I develop some of the assumptions here were to avoid looking at anybody else's sample sizes and published articles first, but to try to develop all the assumptions from discussions with a number of investigators at ASBMR at recent bone meetings. Then, I had statistical models provided to me by Dr. McCulloch. And then once I had the assumptions in place, I went ahead and calculated the sample sizes so that I wasn't iterating back and forth to try to make this look good or bad for you. So, I hope that this is an unbiased estimate.
The assumptions that underlie all the models that I have given you, we're going to use confidence limits. One sided, essentially, we're going to just look at the bottom part of the confidence limit and we'll use an alpha of 0.025. If you wanted to use an alpha 0.05 for that lower confidence limit, I'd be more liberal. The sample sizes would go down by approximately 20 to 25 percent, as noted by an earlier speaker.
I'll show a couple of examples of 0.9 power, but in general, we'll stick with 0.8 power, or 80 percent power. I'm assuming that the bone density trials will last two years throughout, so I won't repeat that figure, and the fracture trials will last three. Towards the end, when we begin to talk about duration of studies, I'll also look at a one-year fracture outcome trial with a comparator.
I did not inflate any of the numbers here to take count of loss to follow-up since that will vary a lot from drug to drug and how it's done. But in general, our experience is that there should be about a 20 to 25 percent inflation of the numbers I give you to account for loss to follow-up the way most trials are done.
The key specification in all of these models is the non-inferiority margin. I have to say what this meant was still a little fuzzy to me. I hope it's not still fuzzy to me as I'm presenting this. But I would sympathize if some of you are still a little unclear about what the non-inferiority margin is. So, I was asked to try to define that a bit.
It's really kind of an accrued streetwise fashion of how much inferiority are you going to allow in a new drug and still let them get away with approving it. And so, it's a margin of difference below the existing drug that you allow in order to say that the result is comparable, sufficiently comparable. You say that you have confidence it's not inferior.
I'll just use an example to illustrate this. If this is an old drug "A" that's been around in the markets, it's proved. We know that it has approximately a four and a half percent improvement in bone density over the course of two years. This is what that would look like.
And we may decide then that to set a non-inferiority margin, that's like a confidence limit that extends down. You'll see it also extends up, but I'm just going to ignore the top part because I don't care about superiority for this. I'm interested in non-inferiority so I'm looking at the bottom part of that confidence limit in order to establish the margin by which this other drug has to perform.
So in a sense, the way I've set this up that non-inferiority means that there is less than a two to five percent change that the effect of "B" will be at least one percent worse than the effect of "A". Here's an example of where it worked. "A" and "B" have about the same mean effect on bone density. "B" may be a little bit worse, but it's still within the margin. You'll call this non-inferior.
And here's a situation essentially where it doesn't work. The difference between "A" and "B" exceeds the non-inferiority margin. You reject the non-inferiority assumption. Again, for those of us who think in other common terms, it's probably inferior by the criteria that were established.
So now, some of the assumptions that go into the bone density calculations come from the trials that we've done and that others have done in looking at the literature and from a consensus when there's been a difference of opinion, we're looking at two years for bone density. So over the course of two years, we're assuming that alendronate, which is the example I'll use first, has an improvement in spine bone density of roughly 4.5 percent over the course of two years compared to the placebo. And at the hip, it's approximately 3.0 percent at the end of two years. The numbers are bigger at the end of the trials that go on to three to four years.
We going to make the assumption in all of these that the new drug you're bringing to market is estimated to have the same effect. It's changes our calculations to assume that it's better or assume that it's worse to start with. We could do that, but I'm not going to. If you want to, we'll do it in private.
The new drug has essentially the same estimated effect. That's why you're bringing it to market. You're not trying to bring something that's inferior. And we're also making the assumption that the non-inferiority margin is one percent lower.
Again, this comes from a consensus or sort of a median of people that I've spoke to about this. That's roughly about a 20 percent difference between the placebo effect and the mean effect. But I test a range in this from 0.5 percent margin to 2 percent.
A very important assumption in all of this, which is surprising perhaps to some of you, but is the standard deviation over two years of change in bone density measurement in the whole population. Small changes in that actually make a big change in the sample size.
So, what I've done for this is I've used real data. There are lots of data points, and Dennis Black and I have tried to find what looks like the modal or the median value of the various groups that we looked at.
So for the standard deviations of change in bone density, I've used five percent for the spine, which is a pretty good estimate for what we saw in the fracture trial treatment groups. Since these are comparative trials, everybody has got treatment. So, we're using just the treatment groups for those estimates. And, a total hip of about four percent.
Now some of you may say that's surprising. Spine should be better than total hip. In the fracture intervention trial database, this is what we found. This is what's in that database for a standard deviation of change, so that's what I've used.
I don't think it makes qualitative differences in the results. But for those specifications of non-inferiority at 0.8, it looks as though you need hundreds of course. And if you want a very, very narrow margin, that actually is not 0.01, that's 0.005 due to a glitch in PowerPoint. But 2002, you want very stringent non-inferiority margins. But for the ones we assumed were reasonable, it's on the order of hundreds. Five hundred and six is the exact number for the total number in the trial that we specified.
To use total hip BMD, it actually was somewhat greater because of the smaller change in total hip BMD. If you were using that as an endpoint, the numbers are somewhat larger. And the one that fit our specified assumptions is about 788-person trial. You can see that if you're more liberal with the non-inferiority margin, you can get down to numbers that are in the 100 to 200 range.
Yes?
PARTICIPANT: (Speaking from unmic'ed location).
DR. CUMMINGS: Yes. I'm sorry. The PowerPoint, this actually is 0.03. I don't why, but I just couldn't get my PowerPoint program to read out the right numbers. So, 0.025 and that's 0.005. In other words, that's a half of a percent non-inferiority margin. I apologize that I'm clumsy with PowerPoint. But, it's fixed on later graphs.
If you use a power of 0.9, it doesn't really make a whole lot of difference to the magnitude of the trial you're planning. Again, it's in the hundreds of patients that you need for non-inferiority for BMD.
Okay, now vertebral deformities is a more challenging task. Vertebral deformities -- again, I want to make the assumptions clear. For me, this was a change in measurement or semi-quantitative grade on lateral spine films.
For the morphometry part, we defined this as a 20 percent decrease in any vertebral height. And that's been the primary outcome of trials that we've done, and then confirmed by semi-quantitative gradings or radiologic readings. We used this criteria for defining the rates of fracture primarily.
Some trials have found much higher rates like the VERT trial, but they've used more liberal criteria like 15 percent. So therefore, their rates of fractures tend to look higher. But the numbers I've used might be slightly smaller than you're used to because I tried to standardize it around 20 percent.
Now the assumptions used are that in these non-inferiority vertebral deformity trials, we'll be using a high-risk group. And those are women with a vertebral fracture, probably of indeterminate age. It's been our experience that we're tending to get somewhat lesser of lower risk patients from the United States than we used to.
The placebo rate that we started with, that Dennis and I came to for this presentation, was a placebo rate of about five percent per year for vertebral fractures. It can be done for higher as you'll see later. That means that there is a 15 percent cumulative incidence of having a vertebral fracture at the end of three years under these assumptions. That makes the numbers turn out to also look neat and easy to remember.
Now I'm going to use for the active comparator the approved drug alendronate. And I going to assume, because the numbers worked out exactly and it's very close, that that drug reduces the risk of vertebral deformities by 47 percent over the course of the three-year trials. That's almost exactly what it is.
That means that if you were to do a trial with an Alendronate group over three years, their rate of fractures, if we had the same kind of placebo group as before, would be eight percent. An eight percent cumulative rate over three years is a 47 percent reduction.
So, this is the number to which we are going to be comparing the new drugs, that eight percent cumulative rate over three years. That's the estimated effectiveness in the right-hand column. That's what we're assuming that eight percent means.
One of the problems with doing active comparator trials is you're really not sure of that. That estimated effectiveness could be 55 or it could be 60 or it could be 35 percent. But, that's one of the assumptions you have to use going into this. And this is the best estimate I've got now.
If you then accept rates for the comparison of nine percent, ten percent, eleven percent, twelve percent margins that go up by you see a percent each, the non-inferiority margin for nine percent would be one percent essentially. If you're going to accept those, then the estimated effectiveness associated with each one of those allowable rates of fractures in the new drug group are on the right-hand column.
I have trouble using figures like 20 percent reduction in the non-inferiority margin in order to understand the value of a comparator trial. It makes a great deal more sense to me as a clinical investigator and clinician to be concerned about the effectiveness of the drug I'm going to be approving.
This represents about a two-thirds preservation of effect. This is about half of the benefit or half of the effect of the comparator drug. I think we chose from the survey of people I did, there seemed to be a consensus that about a 33 to 35 percent reduction in risk was a clinically important reduction.
And if it were below that, people began to have concerns about whether or not it's worth using such a drug in practice and whether or not it's sufficiently different from placebo to be acceptable for approval. So, that's the 33 percent or the 10 percent three-year rate is what I will use as the principle assumption, the number that I'll highlight.
I think that a 20 percent estimated effectiveness or a 12 percent three-year rate in the comparator trial is just -- personally, I don't believe that this is an acceptable degree of efficacy, sufficiently different from placebo to make it worthwhile, even estimating numbers about that alternative. So, the range of numbers I've used is from eight to eleven percent.
Those are assumptions I've already given you. The approved bisphosphonate has a 47 percent reduction risk. The new drug has the same effect, and we set that non-inferiority margin at a two percent different. In other words, we'll accept it if, under this scenario, it reduces the risk of fractures by about a 33 percent.
And, I've tested the range I told you about that includes 11 and 9 percent, a 40 percent reduction, 27 percent estimated reduction, and these are the sample sizes. So if the non-inferiority margin is as we guessed is about two percent, under these sample sizes, it's about a 6,000 person trial not accounting for dropouts under a power of 0.8.
Now if you're using a more strict, a one percent margin that goes way up -- and you can see the effect of a non-inferiority margin is huge on these assumption, so we'll return to how you go about setting those. The panel will have copies of those slides.
Now if you raise the power to 0.9 instead of 0.8, it has a modest effect on the sample size estimates that we came up with. Again, about a 7,000 to 8,000 person non-inferiority trial with a two percent margin. If you're more liberal, you get down to about a 3,000 person trial.
It's possible, I mean I've been a consultant at a couple of meetings where people say we want to compare this to another bisphosphonate that is presumed to have lesser effects or weaker effects, sort of lowering the bar.
And so, let's assume that an alternative, in this case risedronate -- it's best unable to tell from the VERT trial -- has about a 40 percent reduction of risk of fractures in the populations in which it's been tested. It's hard to tell whether it's as effective or less effective than alendronate because they're in different populations. But, there's about a 40 percent risk.
So if you use that, maybe then if you get a two percent inferiority margin and you get all the way down to 20 percent, you can actually get a drug that's weaker on to market by choosing as it were a lower target. Now that's one alternative we'll talk about briefly.
But I think that's where this analysis helps. I really don't want to go below this line. In fact, I'd say that if you're going to use a drug that has weaker effects, you set a narrower inferiority margin in order to test the non-inferiority of the drug.
If that's the comparator, and this is two-thirds of the effect, although that's a half of effect, under this circumstance, I will draw the line in exactly the same place about the estimated effectiveness. I'll look to test that ten percent. In other words, a one percent margin for this particular drug instead of a two percent for the other. And not test the drug that's estimated to have -- I wouldn't be interested in just a 20 percent reduction in risk.
The sample size for a non-inferiority under these circumstances would be at 23,100. That would be the preferred sample size. In a sense, you're penalized by choosing the lower bar. But if you went ahead with the two percent non-inferiority margin, you're in the 6,000-person trial arena.
So if the issue is aiming low, if a fixed margin is allowed, you know two percent regardless of which drug you're choosing as a comparator, a 20 percent reduction, then choosing the weakest comparator will tend to produce an easier, a smaller sample size, which may lead to approval of drugs that are less and less and less, eventually as you go through time, distinguishable from placebo.
I would argue that the basis for choosing non-inferiority margins should be the estimated effectiveness of the new drug as a principle. So the base margins on the minimal estimated effectiveness that you're aiming at, and amongst colleagues, that seemed to be somewhere around 30 percent. Again, a very informal convenient sample of friends has a lower limit.
Now SERMS. SERMS are interesting in this kind of analysis. I'm a little uncomfortable doing this for SERMS because they have effects on multiple organs and conditions. Non-inferiority tests just for bones to allow a new SERM on the market is problematic to me. I mean I'm a bonehead. Despite that, I'm not sure that the bone is the most important organ in the body. My colleagues I hope will forgive me for that.
We've only had one in the class that's been tested for fracture effects, Raloxifene. So, there's not a great deal of data here, which to make these comparisons. This approval on the basis of non-inferiority on bone alone to other SERM agents seems to me to be premature. Having said that, I will nonetheless go through with my assigned task of giving the assumptions and the results of this.
The assumptions are essentially here that the approved Raloxifene effect is about a two percent change in spine bone density. The new drug has the same effect. And we set the non-inferiority margin here because it's such a small difference from placebo. Setting it at one percent is reducing that potential benefit by half. It doesn't make much sense, so we tested it at a 0.5 percent lower bone density effect as the non-inferiority.
It just turns out in the MORE database that the standard deviation of two-year change in the spine bone density looks better than in the FIT trial. Perhaps, because the effect of the drug is smaller, you get a lower range of changes within the population. It's 3.5 percent for the standard deviation change in our database.
Anyway, the sample size for non-inferiority spine bone density, in part because of the better standard deviation, is modest. At 0.5 percent, it's around 800.
And for vertebral deformities, again I'm going to assume for convenience that it reduces fracture risk by about 40 percent. Therefore, in a sense the analysis we did for Risedronate really applies quite directly, and I can just skip to that. Depending on the inferiority margin you'll use, it's between 6,000 to 23,000. It's a very critical assumption.
Again, because the fracture effects are perhaps more modest with the SERM class than with the most powerful bisphosphonates, you would tend to choose somewhat larger sample sizes.
DR. COLMAN: Steve?
DR. CUMMINGS: Yes?
DR. COLMAN: For those sample sizes, those are total or --
DR. CUMMINGS: These are all total, not per group. These are total, not per group.
DR. COLMAN: Okay.
DR. CUMMINGS: Now, these are a lot of numbers. And when you come down to actually calculating this, there will be different assumptions. People will come to the table, manufacturers will come with different assumptions that will make different numbers.
So, I think that the summary really is that for vertebral fracture comparator trials, the number that you need is in the thousands. And for bone density comparator trials, the numbers will be in the hundreds. The non-inferiority margin that you accept makes a huge difference. Probably next on the list would be the standard deviation of change over time for bone density studies.
Now, what non-inferiority margin makes sense? I've suggested something for vertebral fractures as a minimum, sort of a bottom floor. But I just wanted to offer an opinion that in a sense, using non-inferiority margins of the sort that I talked to you about that reduced or allowed 20 to 30 to 40 percent inferiority makes some sense to me if the new drug has other benefits. You can accept something that's inferior if it's got other health benefits, it's safer, it's more convenient, and therefore better adherence and it's less expensive.
I know that this is not usually done in setting non-inferiority margins, but it seems to me that there would be a compelling pace for being a little bit more liberal for something that has other benefits. For new drugs with no other advantages, they should either prove superiority or the margins of non-inferiority should be very narrow, probably narrower than I've rehearsed for you earlier in the talk.
PARTICIPANT: What is standard?DR. CUMMINGS: What is what?
PARTICIPANT: What is standard? DR. CUMMINGS: No, I'm sorry. What I meant is the ones I just used. Standard came about because Eric suggested some percentages that we used that were within the range that I talked about. In other words, 20 to 30 percent differences between the effect of the drug and the placebo group. And so, I think that those are acceptable if there is something new being brought to the table. But if not, then I would propose that the criteria be more stringent.Now let's go to duration of trials, and this will be the last. We'll start with one-year duration trials. I think that there is some sense nowadays for some compounds and maybe even for bisphosphonates, where they worked so dramatically in the first year, to consider a shorter duration of trials for initial registration of drugs.
Let's first look at the one-year study. One year, a 60 percent reduction in risk of vertebral fracture can actually be seen in relatively small trials compared to placebo. So, I'll just focus here on the non-inferiority margin for a new drug and assume that the estimated reduction that we're interested in would be no worse than a 40 percent reduction in one year.
This happened in the first year, but I'll show you that in later years, the effect is not quite that strong. And so if you're going to do it on the basis of just the first year, then I don't think you allow as low as a 20 percent reduction in risk, a big range. So, I'm going to assume a 20 percent non-inferiority margin, which means a 40 percent effectiveness, and test a range from 10 to 30 percent with the assumptions you've seen before.
Two-thirds effect is a 40 percent reduction. Half of that effect is a 30 percent reduction. These are the non-inferiority margins that Chuck calculated. I'm sorry, the sample sizes for those non-inferiority margins. In other words, for that first year comparison of a drug that reduces risk by 60 percent with another new drug assumed to have the same effect, but you're willing to accept lower. That 6,154 sample size would be my preferred number.
What if the risks were higher? In a 12 to 18 month study, you've managed to get a very, very high-risk group. If we do the same thing with much higher risk groups, double the rate of events, there is a modest effect. It reduces the preferred sample size by about half to double the incident rate. So now, that's the year one-year effect.
I'd like to now talk a bit about longer durations and why there might be a rationale for being concerned about longer duration studies. It's been pretty clear, and certainly if you've been at any meetings or read any medical journals, very well advertised that all drugs, antiresorptive drugs, dramatically reduce the risk of fractures in the first year. You've heard 65, 68, 60 percent -- I saw an article recently claiming a 70 percent reduction in risk of fractures in just the first year. Again, I think that's true and there are biological reasons why that might be happening.
There is concern among some members of the osteoporosis community about the long-term effects of antiresorptive drugs. Right now, we can't do much with that concern because the placebo-controlled trials have lasted only three to four years.
There have also been long-term trials, particularly with the alendronate, with the long-term extensions going out to seven to ten years from which we've had to try to draw very indirect inferences about how long the drug continues to work. And there is some reason, I think, to be concerned that we should go a little bit longer than three to four years, coming from the data that we've seen so far.
Risedronate, for example, we know continues to improve bone density. It dramatically reduces risk of vertebral fractures in the first year. In the second to third year, we really don't know what it does separately because the sponsor hasn't released that data from the VERT trials to allow us to do those kinds of calculations.
But fortunately the sponsor for Alendronate has allowed us use of the databases and we are able to show that a similar phenomenon of dramatic improvement in the reduction of risk in the first is seen with alendronate, and there is a 40 percent reduction in the third year. It's a little bit more difficult to figure this out, exactly what that means because there was also a dosage change from five milligrams to ten milligrams between those time points.
This change or this apparent waning of effect could be due to something called depletion of the susceptibles. That is, if there is susceptible people in the placebo group that all fracture in the first year and they're gone. Then there are fewer of them around in the next year to have fractures and the drug won't look as good.
So, we've done some extensive modeling about this statistical artifact. Models with very extreme assumptions that basically assume that all the fractures happen in the susceptibles in the first year, and high rates of susceptibles with extreme assumptions really can't account for the observed declines in effects that we've seen with these drugs. In particular, because such an effect, a statistical artifact would require that you also see substantial declines in the rate of fractures in the placebo group because the people who are going to have them are gone.
And, you don't see that in the existing trials. You see pretty constant rates in the placebo group. Although it's been said that this is statistical artifact, it is not. So that means that it is a biological effect and there are two possibilities.
I think the most likely is just the dramatic first year effect on bone resorption that gives you, that preserves architecture. Then that, on top of a long-term sustained effect, results in a 30 to 40 percent long-term reduction in risk because of the improvement in bone density.
But I can't tell on the basis of the current data we have available, I can't tell that from the second alternative, which is that inhibition to bone resorption initially strengthens, and then after five, six, seven years longer than the existing data from trials or from other types of studies, longer than it's gone. So that remains a concern in some quarters of the community. So I would like to suggest something.
I would like to suggest that you might consider approving drugs for use even as early as one year or 18 months, fracture data, but then approve the drugs for use equal to the duration of the trials that you are provided as evidence. And that to get an extension in the label of how long patients be allowed to use that would be contingent on providing data that the drug remains safe and that the fracture risk is durable, that the reduction of fracture risk is durable.
Now there's a challenge with this of course, and that's that it's not feasible to continue placebo-controlled trials beyond about three years. It's just too difficult because of the environment about placebo-controlled trials.
There are a couple of alternatives to this, however, that I think are a reasonable comprise. And one is to continue your placebo-controlled trial for three years and then stop the placebo group if it works and continue the treatment for longer. Companies are doing this, but they're often doing it without adequately powering the study or planning it in this fashion in advance.
It looks like this. There's the rate of the placebo group for the first three years, and here is the year-by-year rate in the treatment group as it continues out to ten. It looks like it's continuing to work, and that's great.
Again, this is just one possibility. There are other variations on this, but this is the simplest one. You could compare slopes of lines or just year-by-year effects. For the purposes of just this one example, I'm going to suggest comparison of the rates in the treatment group at seven to ten years with the rate of the placebo group for the first three, then needing to adjust for the advancing age of the patients who are going out ten years.
And that's the comparison one. It would draw then the placebo group to the rate in the treatment group carried out longer. In this particular case, it's a statistically significant difference. But it's possible that with time we would see a loss of effectiveness, a loss of durability and that it would no longer differ from placebo. It might even cross this line and would lead to the conclusion that you should stop after three or four years rather than continue it 20, 30, or 40, or lifelong.
These kinds of trials are feasible. I won't go through the details of the sample size estimates, but we've done several such estimates. And in general, trials that had more than 1,000 per group at the baseline with less than 20 percent loss to follow-up during the placebo period and are able to retain at least 50 percent of participants out ten years will have over an 80 power to confirm a 30 percent lower risk in the last three years versus the placebo rates.
There is another alternative. I think Merck has done a very innovative thing with the FIT trials in taking the treatment group after four years and then re-randomizing them to continuing out ten, or stopping the drug to test whether or not there is benefit from continuing as opposed to stopping. I like that design very much if it's adequately powered.
With that, let me summarize by saying that placebo-controlled trials with women who have densitometric osteoporosis entail low risk to the participants and are feasible. And I think that that might also include women who have a single vertebral fracture of indeterminate age. Non-inferiority trials on bone density require hundreds, and thousands are needed for vertebral deformity comparator trials on non-inferiority, but that depends very heavily on the non-inferiority margin.
I would suggest or I would like the panel to consider at least initial registration of drugs for the duration of the evidence that you're presented with, and extend that duration with subsequent demonstration, that there is durability of effectiveness as well as safety. And, consider setting the non-inferiority margins, such as an important determinant based on some judgment about advantages, if any, of the new agent. But at the least when you're dealing with comparator trials for vertebral fractures, set those margins that preserve a minimal estimated effectiveness of treatment over placebo.
With that, with those modest suggestions, I'll stop and say thank you.
(APPLAUSE.) DR. CUMMINGS: Henry, am I allowed to take questions?
CHAIRMAN BRAUNSTEIN: Dr. Bone?
DR. BONE: Steve, thank you for a nice review. I had two specific questions, and you may even want to stay up there with your computer to answer these two questions.
DR. CUMMINGS: I don't think I'll be able to re-project, but go ahead.
DR. BONE: Okay. These had to do with some estimates that you didn't mention. One is, if you use the figure 47 percent, which is just about exactly the reported relatively risk reduction for alendronate in the trials you're referring to, what is the confidence interval around that estimate for the relative risk?
DR. CUMMINGS: For alendronate?
DR. BONE: Yes.
DR. CUMMINGS: Although I'm an author on those trials, I don't remember the confidence limit. It was relatively narrow, particularly when the two trials were pooled, those FIT-1 and FIT-2 for patients with osteoporosis.
Someone else might be able to help me with the confidence limit, but I think that it goes down no further than the high 30 percents. About 37, 38 percent is the lower limit of that confidence limit.
DR. BONE: And the other question was since it's important to know whether drugs actually reduce the risk of hip fracture -- and we've had examples of drugs which did resist the risk of vertebral fracture with no relative risk reduction at all, even not a significant one. Just no change in the risk for hip fracture. What's the sample size calculation for doing the active controlled trial for a hip fracture endpoint?
DR. CUMMINGS: That's also been done by John Kanis, and I'd refer to those. I did it one pass at it, and it's in the 20 to 50 margin.
DR. BONE: With what power?
DR. CUMMINGS: I think we used a 20 percent difference from a 50 percent reduction. Just because it's such a rare event, it's impossible to do a comparator trial for hip fractures.
CHAIRMAN BRAUNSTEIN: Dr. Watts.
DR. WATTS: You chose a 40 percent reduction in fracture for your example with Risedronate. In the two vertebral fracture trials, one showed a 41 percent reduction and the other showed a 49 percent reduction.
DR. CUMMINGS: Yes.
DR. WATTS: And that raises a dilemma as to which of those numbers you would choose if you were powering a trial. It might've been cleaner had you chosen those two numbers rather than two different agents.
DR. CUMMINGS: Yes, that's true. I could've done it the other way. What I was trying to do was not pin 40 percent just on Risedronate. But say, what if you had a drug, which you thought was less effective, and therefore you chose as a company to choose that as a lower mark to hit with your comparator.
DR. WATTS: Another trial design that you didn't mention was superiority trials.
DR. CUMMINGS: Yes, I wasn't asked to do that.
DR. WATTS: It's possible that a new agent might come out that looks like it's a lot better. And by my calculations, it takes a far smaller sample size to show superiority.
DR. CUMMINGS: Well if it's really superior, it takes a different, a larger sample size. But if it's not, if it's on the same, then that changes your sample size. Remember, throughout this I assumed that the new agent that was coming on, in fact, had the same reduction in fracture risk and the same change in bone density.
And if you change that assumption, I can show you what it does. If your drug is better and you use a non-inferiority margin, then it's an easy time. You can do it with just a few hundred patients.
CHAIRMAN BRAUNSTEIN: Yes, Dr. Temple.
DR. TEMPLE: Choosing the margin for non-inferiority trials is the subject of infinite quantities of discussion because there are a bunch of cardiovascular diseases where no one would debate the possibility of whether you can still do placebo. So, this becomes a very important issue.
It's very important in doing that to distinguish between a non-inferiority margin whose purpose is to show that your new drug has some effect compared to placebo, any, and one in which you're designing it so that you show you preserved some fraction of it.
It should be obvious, but if the situation is such that it's unethical to use placebos anymore, it's obviously important to preserve a fair fraction of the effect of the control agent, otherwise what's the point? But, those two things need to be kept in mind.
If, for example, you thought a reliable meta-analysis of the effect of some positive control agent was a 40 percent reduction, you could document superiority of the drug to placebo by showing that the difference between your new drug and the controlled drug is not more than 40 percent. Then, it would be better than placebo, which is approximately equal to what you do when you discover that something is significant 0.05.
If you don't like that well enough, if you have a mixed feeling that you want to preserve some fraction of it, then you have to do what Steve was doing, preserve 50 percent of it or something like that. Of course, the implications for sample size are spectacular.
One problem that we encounter repeatedly is where you only have one trial of something. For example, in most lipid settings, there's only one trial and it a particular setting because no one will let you do another trial once a benefit has been established. So how on earth do you pick a non-inferiority margin based on a single trial?
Well, taking the mean doesn't seem good enough because half the time the effect is going to be smaller than that so you wouldn't know what the effect in your new trial is. One thing that people have done is take the 95 percent lower bound, which if there's only one trial means the difference that you have to rule out is considerably smaller than the mean B sorry, than the point estimate of the effect, and trials get very large.
So in these situations, we've encouraged people to, either through pooling a lot of data or looking at the one drug with the most data, use that as the active control and then at least you have a number that you can rely on. You don't have to be entirely conservative.
But, we would never think that the mean effect of a single trial would be the right non-inferiority margin. It has to be, as somebody over there suggested, a 95 percent lower bound or something. Anyway, there's a great debate about exactly how to do that.
It does seem very important to distinguish between trying to show through a non-inferiority study that you're better than nothing, which might be good enough in some cases in trying to show that you preserve some clinically meaningful effect of the drug, which then enters into major debates.
For what it's worth for thrombolytics, where the endpoint is death, CBER, the Biologics people accepted a non-inferiority margin that represented retention of half of the effect of the thrombolytics based on the 95 percent lower bound of a meta-analysis. That turns out to be a little more conservative than one might do. But that's one living, breathing illustration.
Of course, there the consequence of being wrong is death. Here, as people have been saying, it's a fracture. That's not as bad as death, but it might be bad anyway.
CHAIRMAN BRAUNSTEIN: Dr. Marcus.
DR. MARCUS: A box containing Steve's slides is going around. I'd like to go out beyond the box if you don't mind of a minute.
There are two possibilities that haven't been described in ways to approach some of these issues related to fracture, particularly hip fracture. One has been proposed by Nelson Watts, and I'm going to ask that he discuss it since he is the father of that. That has to do with pooling of groups across various published trials.
The second one has to do with using as a control group, published data from ongoing, contemporary, very large-scale public health databases such as NHANES-3, where we have a very good indication of what true hip fracture rates are in this country. We can isolate the data by age, by ethnic group, by gender, and I just raised the possibility that one might be able to do a trial in which the control group could be valid public health data.
I'd like to hear the agency's response to that, if they would automatically exclude that or if they'd be willing to think about that approach. I'd like to hear Steve and some of the other epidemiologists discuss that.
As the second model, I yield to my distinguished colleague, Nelson Watts, to raise his idea.
DR. WATTS: What I've done is to look at the fracture experience in a trail in which everyone received active treatment: daily dosing, weekly dosing that had equivalent effects on bone density and bone turnover markers, and extract from another large database of a placebo-controlled trial, subject to or matched to the entry criteria for the trial that lacked simultaneous controls. I matched for key characteristics, age, bone density, years since menopause, and the percentage of subjects with prevalent fractures.
In doing that, we not only had the historical control group, but a historical active treatment group as a way of internal validation. At least in that one, the rates of fracture in the historical treatment group were indistinguishable from the rates of fracture in the study that had no control group. The difference in fractures between the historical controls and the active treatment was statistically significant.
Though it should be possible, given the large trials, to create a huge database against which to judge -- extract a control group to judge antifracture efficacy. That doesn't necessarily allow you to compare bone density or bone turnover markers or establish safety, but at least to get an antifracture efficacy.
CHAIRMAN BRAUNSTEIN: Dr. Cummings.
DR. CUMMINGS: The problem with doing that for hip fractures from databases is that there's a universal experience in trials and observational studies -- very marked in our studies and also in the WHI, that for reasons that we can't understand and that are not explained by bone density, age, or even estimates of health, there is a healthy volunteer effect on hip fractures.
So, their rates in the first few years of any study are in the order of one-tenth to no better than a quarter of the rates that you would expect from those patients matched by characteristic to databases. It could very well be that the hip fractures, of course, happen in people who are frail and don't come in to trials or to observational studies.
I don't know how to adjust for that healthy volunteer effect, so it's not a really credible, tenable way to develop controls for a hip fracture study.
CHAIRMAN BRAUNSTEIN: Dr. Temple.
DR. TEMPLE: Well, the Women's Health Initiative gives you reason for caution. I mean why was it done? It was because every epidemiologic study ever done showed a 50 percent reduction in users of cardiovascular events. It turned out that those rates didn't really represent the truth for reasons that remain inexplicable.
It's almost surely true that people who enter trials are not the same as people picked up by NHANES or something like that. There are too many examples to enumerate, but there are numerous.
CHAIRMAN BRAUNSTEIN: Okay. We'll turn it over to Dr. Orloff, who is going to give the group their charge.
DR. ORLOFF: And charge it is. The first thing I can think of is ladies and gentlemen, start your engines because I think there is some discussion to ensue.
Let me thank everybody for a number of very informative and careful presentations. Clearly, I don't hear any definitive answers yet. Let me start this by saying that we did not expect a consensus to come out of this meeting. I'm not hiding behind this, but I guess in part I'm saying this so that we can all leave here as friends as I think we entered the room.
We're here to frame the issues. This meeting is the first step we would hope in developing a guidance for industry for the development of drugs for U.S. marketing in the treatment of osteoporosis.
I just want to make sure that as we do deliberate -- and this again is aimed at trying to make sure we're all toned appropriately -- that guidance is just that. I think it's not universally understood that FDA guidance is guidance. It's not law and it's not regulation. It is, we hope, representative of the agency's best thinking. That's obviously with the input of our advisors and consultants on the subject issue.
So on the one hand, notwithstanding guidance, sponsors are at liberty to plot their own courses in drug development to meet the stated or implied needs or concerns of the agency. Though to the extent that the guidance is not followed, we expect sponsors to justify the alternative approach. On the other hand, adherence to guidance does not guarantee approval or otherwise mandate a particular regulatory action by the agency.
Anyway, I'm going to ramble a little bit with some thoughts that I have before I get to a specific discussion of the questions that we'd like to have you address.
From sitting here, I think we got conflicting messages in the discussions on the acceptability of bone mineral density as a predictor of fracture benefit. I think it was also clear that we need to know more about the impact of animal findings in any given specific instance as confirmatory of potential salutory effects on bone for different mechanistic classes.
The meta-analyses that were presented in brief support BMD as a good predictor for amino bisphosphonates and parathyroid hormones, to the extent that there are data for that drug, with the qualifier of course that animal data be positive or favorable even if the bone mineral density in those studies does not explain the whole fracture effect. I think we're all in agreement that that point was made.
It does sound to me as though BMD generally may be a reliable positive predictor of efficacy. It's just not a useful negative predictor. Thus, specifically it's probably not to be used to compare efficacy across different mechanistic classes for the purposes of placement in the armamentarium, which was a subject that's been raised. Whether or not it can be used for the purposes of approval, comparisons of BMD effects across classes, is perhaps another question.
I want to recognize that we did hear Dr. Marcus' comments on the need for additional histomorphometric characterization on the effects of new and existing drugs. Though the validity of some of these endpoints, as independent predictors of bone quality, remain to be demonstrated.
So, we well recognize the need perhaps to understand more about the differences between one drug class and next. Although just saying that they have different effects on a marker, doesn't necessarily mean that they are different in terms of absolute clinical efficacy.
As we, you and we, that is to say go about thinking about the problems before us, I just offer that we need to distinguish between two important needs of the system, the healthcare system with regards to osteoporosis.
On the one hand, we need to come up with standards for approval as safe and effective. We need to be thinking about placement in the armamentarium. Again this theme comes up, the latter, the placement in the armamentarium is very important but not unfortunately the driving force for standards of approval. Perhaps it should be.
Also I want people to understand that we do not, or we are not, FDA, considering new drugs in this -- or for that matter, any other therapeutic area about which we know something -- we are not considering new drugs in a vacuum. For post-menopausal osteoporosis, for example, we have tremendous or a great deal in the way of priors that allow us to put new data and development problems in perspective.
And so, for example, one might ask the simple and blunt question, which I do think we need to toss about a bit: Why wouldn't bone mineral density in an active controlled trial suffice for approval of a new bisphosphonate, or a new SERM, or estrogen or a new agonist at the PTH receptor?
It's also apparent that the ethical questions, that I guess at some level I'd hoped we wouldn't spend too much time haggling about today, have not been resolved. Two general approaches have been advanced around the table here and by others.
One is the idea of short-term placebo-controlled trials in high-risk subjects in which the fewest patients would be placed at risk for fracture for the shortest time. This is a burden-to-society argument, if you will, that does not address the irreversible morbidity and mortality standard down by the Declaration of Helsinki.
The other approach is larger, long-term placebo-controlled trials in low-risk subjects with theoretical advantages of overall low frequency of events, and the low risk in a carefully chosen population of serious morbidity or mortality in a presumably generally healthy population.
I would venture that we need to be careful hearkening to Dr. Cummings' presentation with calculations of average morbidity experienced in the trials to date. For example, average days of pain or average days of work loss or average days of bed rest.
Since this measure of central tendency, if you will, masks the fact that clearly more patients had significant morbidity in association with placebo than drug, than there must be some patients in there on placebo who had serious or significant morbidity.
Finally, on the subject of trial designs, Dr. Cummings has left us with the conclusions that hundreds of patients would be required over several years for non-inferiority BMD trials and thousands for a fracture non-inferiority trial looking at morphometric vertebral fractures.
All of these issues that I've skipped over bear further discussion. I'm not sure actually how much discussion we need to have here about placebos, but I leave it obviously to the Chair. Whether the issue of placebo versus active controls should be a question related to post approval broadening of claims, that is to say assuming that there is some consensus that bone mineral density might in many instances, studies may be sufficient for approval.
Anyway, the way we structured the discussion at least in our planning was to ask about the nature and extent of evidence from approval centered around four hypotheticals. You've seen these in your agenda. I said earlier, a new bisphosphonate, a new estrogen agonist bone, a new mechanistic class antiresorptive, and a new anabolic agent.
I just want to go through with you some of our thinking as we put together the structure of the questions that we'd like you to consider. For the establishment of efficacy, we asked you to focus on three main questions.
The first, when is bone mineral density an adequate primary endpoint, is really about how far clinical studies need to go to bone mineral density versus fractures in order to support approval. We understand full well that this decision or this judgment must take into consideration the specific results of preclinical studies. That's given. But clearly, also the confidence that should be placed in such studies. And that's why we asked for the presentations that were given to today.
So for example, going back to Dr. Rodan's presentation and Dr. Rozzoli's, such studies are all important for loose dating mechanisms of actions as well as such things as the potential for toxic mineralization effects of drugs or the existence of an apparent unfavorable relationship between BMD and bone strength indicative of poor bone quality.
I should say that as an important caveat or kind of a reverse catch-22, we also have to ask why would anyone in this day and age pursue a drug that demonstrated a poor efficacy or safety profile in animals.
The issue raised in discussion earlier, relative to extrapolation of animal studies and models of osteoporosis to other forms of bone disease associated with fracture risk and/or osteoporosis, I think is an important one. This presents a perplexing problem, particularly since the number of such patients, that is with other forms of disease associated with fracture risk will often be much smaller than those of patients with postmenopausal osteoporosis.
The reality is though, as pointed out by Drs. Bone and Rizzoli, that the other role for animal studies after pharmacology is bone toxicology. I suppose the judgment of whether some unique toxicity of bone could be anticipated, and these are others, must be made on a case-by-case basis.
With regard to the second question we then ask: If BMD is deemed sufficient, how long should trials be to establish durability of that effect -- and probably not separable, although it will be asked separately for assurance to say what duration of trial is necessary for assurance of bone and extraskeletal safety. If fractures are deemed necessary in a specific instance, do the same questions apply?
And now with regard to the choice of placebo versus active control, the question must be addressed separately if BMD are fractures that are required. Notwithstanding the numbers involved, if BMD is the endpoint, we must ask what constraints exist against the use of placebo, whether they can be addressed by escape criteria for BMD or fracture on trial, what risk categories are appropriate for such trials -- as we said before, low-risk prevention versus treatment -- and of course, whether extrapolation of efficacy from prevention to treatment populations is possible.
We must also address advantages of active versus placebo with regard to safety assessment, or disadvantages, and with regard to again, placing the drug in the armamentarium. If fractures are required, what are the opportunities for add-on, what are the possibilities with regard to active controls, and are there indeed hypotheticals at least in the list that we've given or any that you can think of for which placebo is really the only option in order to assess efficacy.
With regard to safety -- again in many instances, not separable -- I do have a question that I think has been of some confusion, at least it was Dr. Temple before I got to him, in number one, which asks about whether fractures can be used as a safety rather than as a efficacy endpoint. Although the answer may be simple, we thought we wanted to hear people's thought about it.
If a trial is examining BMD as the primary endpoint of efficacy, fracture, rather than being the measure of effectiveness, becomes a safety outcome. The question is: How should it be evaluated? Should it be evaluated based solely on ascertainment with regard to clinically apparent fractures or should there be active ascertainment as a way of monitoring patients in the trial? And, what issues does active ascertainment raise with regard to escape criteria on the one hand, but also on the other hand with regard to thoughts about essentially powering the trial for safety.
We also ask about other safety monitoring in the study, and this gets back to Dr. Grady's question at the beginning. I just want to make sure our position on this is well understood. We do not propose that the choice of efficacy endpoint necessarily impacts the scope or duration of the trials in order to assess safety. So, use of non-fracture endpoints does not necessarily imply shorter, smaller, or narrower scope trials.
With regard to other safety, we would say that those assessments of other safety issues is driven by the usual mechanisms of action, preclinical signals, early phase findings, plausibility of risks.
We also ask you about duration of trials for bone and extraskeletal safety that I mentioned earlier. And finally, what are the theoretical or real advantages and disadvantages to active versus placebo versus add-on trials for safety assessment?
With that, I hope I haven't confused matters. I'll turn it back over to Dr. Braunstein.
CHAIRMAN BRAUNSTEIN: Thank you. There's a lot of subquestions stuck in there.
I think what we'll do is ask the committee and the guests, of course, to consider and keep in mind the four hypothetical osteoporosis drugs, and we'll go through the questions even those there's obviously a lot of cross over from one question to the other.
We'll go through the questions sequentially and ask the members to ask questions of each other, to make their comments, and then when there's a lull, maybe what we'll do is go around the room and ask everybody to take their best stab at answering each of the questions.
So, we'll start off with the question about efficacy and when is bone mineral density an adequate primary endpoint. I would say in discussing this question, anybody who wants to indicate that they'd rather have fracture endpoints in place of bone mineral density for specific compounds in specific issues should mention that. So, let me open up that question to the group.
Dr. Watts?
DR. WATTS: There are three or four drugs on the market for which a relationship has been shown for gains in bone density and reduction in fracture risk. That's Alendronate, Risedronate -- you choose that as a class or agent -- Raloxifene and maybe estrogen if you take the recent bounds health initiative.
There have also been some trials that seemed adequately powered to show an antifracture effect in which a bone density change was noted, and yet the antifracture effect was not seen. And I've already posed this question to Henry, so don't answer it please Henry. But, one of these trials was with intravenous Ibandronate, which produced about a five percent gain in bone density over three years, and did not show a reduction in fracture rates.
So, I'd be interested from those of you who know the trial or those of you on the FDA side to tell me if gain in bone density in a clinical trial were the endpoint for approval of these drugs, would intravenous Ibandronate meet that standard?
CHAIRMAN BRAUNSTEIN: Yes, Dr. Khosla.
DR. KHOSLA: I guess I just caught into the Ibandronate because I remember seeing the data. And if it's the data that I've seen, the problem as I understood it was that the bone turnover markers were coming back up before the next dose of Ibandronate was given. That suggested that that particular trial didn't have a sustained reduction in bone turnover.
DR. WATTS: That was a post hoc analysis. And if you look at the marker data in the trial, the markers were suppressed. It's only by looking at marker data from other trials that it's possible to see they were much more suppressed within a week or two of the dose, and they headed back towards baseline. But, they were still 50 percent below baseline.
DR. KHOSLA: I guess the only comment I'm making is that with most other antiresorptives, there's been a sustained and consistent reduction in bone turnover. We've already heard about the importance of bone turnover as a potential additional factor that contributes to the antifracture efficacy.
I guess my only comment to this is that for classes such as "A" and "B", as I mentioned earlier, if anything, bone density is perhaps a conservative estimate of the reduction in fracture risk. Provided that changes in bone turnover are consistent with what is otherwise seen with these classes, you could argue that it may not be an unreasonable surrogate.
CHAIRMAN BRAUNSTEIN: Dr. Sampson, you had a question?
DR. SAMPSON: Actually, I just wanted clarification from Dr. Orloff for my understanding. If one came to the conclusion that bone mineral density is a primary endpoint, how would that be reflected in the indication for the compound, and would you anticipate a fracture claim being allowed if one were able to show BMD as a adequate surrogate?
DR. ORLOFF: Well in my rather naive world, as I mentioned back at the beginning, the question we're asking is whether one can rely on bone mineral density in some instances as an adequate surrogate for a reduction in fracture risk even if we can't say exactly how much of a risk it involves.
So, analogous to the approval of statins based upon LDL lowering in the absence of an effect demonstrated in a large endpoint trial with regard to reduction in the risk for heart disease. Although admittedly there is an implied claim of fracture benefit, it is not so stated in the labeling and promotion for the drug.
CHAIRMAN BRAUNSTEIN: Dr. McClung?
DR. MCCLUNG: To come back again to the issue about bone density as the alternative to fracture -- with drugs, in which that relationship, that there maybe a reasonable time when that would be appropriate, particularly if we're simply looking at other groups of patients with the same drugs we've studied, or with drugs in the same class in which the mechanism of action has been shown to be very similar, and when the dosing regimen is the same.
And, there are at least examples with bisphosphonates where a whole variety of alternative dosing regimens of different durations, which reflects on the Ibandronate data, provides a different pattern of suppression of bone turnover. And until that has been evaluated, that would be a restraint I think to not allow bone density to be the only endpoint.
And certainly to go across classes, you've already got an example of that in the lipid field. You've already used that as your example. If you are comfortable with stating the changes in serum lipid levels reflected across the class of statins, you maybe comfortable with that, but you haven't approved the use of hormone replacement therapy for the reduction of heart disease on the basis of the reduction in lipid levels that are seen with that agent.
Using bone density as the surrogate across different classes of drugs would be analogous to that circumstance.
DR. ORLOFF: Well, let me just address that for a brief moment.
The willingness to accept any surrogate -- and a surrogate by definition is imperfect because it falls short of the ultimate endpoint of interest. We concede that. The willingness to accept it is based upon not only a robust predictable repeated effect on the marker, but also on the absence of any apparent countervailing risk in the same of different body systems that you would estimate might adversely impact overall outcomes.
So, there is a judgment call always in reliance on the surrogate. We rely on them in the context of sort of a reasonable assurance of safety based upon, in many instances, very large exposures, long-term exposures, multiple -- a lot of experience, for example, with the class of drugs.
So in a case of the absence of labeling for cardiovascular risk reduction for estrogens based upon a lipid altering effect -- incidentally, they're not labeled as lipid altering drugs either. There's an appropriateness there. I can't say there's a reason, there's an appropriateness there.
Because of long standing, there has actually been some concern, doubt about either the possibility of countervailing general cardiovascular adverse effects like deep venous thromboses and/or more recently the possibility that there might indeed be coronary adverse effects. So, I don't think it makes it a non-starter, the whole issue of the surrogate.
CHAIRMAN BRAUNSTEIN: Dr. Lukert?
DR. LUKERT: One thing that would seem helpful would be to combine markers of bone resorption or bone turnover with the bone density results. Particularly when we get into intermittent dosing, I think it might be helpful to observe whether or not you're getting good -- particularly with antiresorptive drugs, to see if you're getting consistent reduction in the resorptive markers that would add to maybe the validity of bone density measurements.
CHAIRMAN BRAUNSTEIN: Dr. Bone?
DR. BONE: I've asked Dr. Watts privately, and he said now I can respond to the question.
(Laughter.) DR. BONE: I think that the particular example that he raised is a vexing one because it is a member of the class that we've been talking about. It showed a substantial and statistically significant increase in bone density in the trial that was cited.
The problem is that we have explained this ex post facto. We say now, "Well, maybe it was the markers". But, there was evidence that it also wasn't the optimal dose for bone density. So which is it? Or, is it both or is it something else?
The problem we have here is an example of a large trial with a drug in our best characterized class in which there was an increase in bone density and where there was a trend toward a reduction in the fracture rate, but it did not reach statistical significance according to the test supplied. And, I think this is very annoying.
(Laughter.) DR. BONE: I really wish that this trial had been done a little differently because I think we would be talking about this whole issue in a different way.
At the same time, it may be a lucky thing. Because, if we didn't have this trial to vex us, we would probably be much happier about accepting the idea that simply seeing a bone density increase with no fracture data and no marker data or no any other data but just the bone density increase for a drug in this class would be just fine. There was nothing to suggest this drug was making the -- the implication of that trial was that the effect was simply insufficient to have a robust clinical effect. But, it puts us in a position when we want to generalize.
DR. COLMAN: Henry, has anyone seen the actual statistics on the fracture rate data for the Ibandronate trials?
I mean because if we're talking about a p-value of 0.5 versus a p-value of 0.06, there's a huge difference there. And I for one would not be willing to say that it was a complete disaster and that the BMD fracture relationship has been permanently smeared because of that. I think if the p-value was 0.06 or 0.07 -- or if they would've added 100 patients, it would've been 0.03.
CHAIRMAN BRAUNSTEIN: Dr. Aoki?
DR. AOKI: It seems to me that we're paying a lot of attention to the numbers in the human study, and I was wondering if the same attention shouldn't be directed at the preclinical or animal studies.
The issue with the bone mineral density is a good surrogate. I think it can probably be best investigated if you looked at animals, looked at the relationship of bone volume, bone strength, and bone density in those animals.
I'd be kind of curious to see in the studies that you were referring to Dr. Bone, if you went backwards and when you found a clinical outcome in humans that was somewhat perplexing, to go back and see if those same problems were present in the animal studies. The problem may be that we don't have standard animal studies were we give a certain dose that will give rise to a certain degree of increase in bone mineral density and then extrapolate that in a stepwise fashion, both in animals and in humans.
CHAIRMAN BRAUNSTEIN: Dr. Temple?
DR. TEMPLE: Certainly in talking about surrogate endpoints generally, it's usually thought to be a bad thing if there's a well-done negative example. Now if the study was too small and other things were wrong with it, that doesn't count so much. But one of the reasons we still use blood pressure, to my knowledge, there's never been a negative placebo-controlled study of any blood pressure agent even though it may be the drugs differ from one another.
I had a related question and that is: If you establish somehow that a drug has a fracture effect in one setting, does that then settle the issue for all settings? In other words, is there a proof-of-principle thing here where let's say you do make a persuasive active control case in a very high-risk setting, does that then make everybody comfortable about the lower-risk setting? There have been people who've said changing dosage forms, changing regimens, things like that, that that's okay for BMD if it's the same drug. That sort of implies that nobody is too worried about the specific way it's used or the dose and a lot of other things, once you've established that this is the kind of drug that has not only a good effect on bone mineral density but also has a fracture effect. That could make a big difference. I have to tell you part of what I have in mind is that there seem to be circumstances in which one could ethically do an add-on study which might show a fracture effect where you might have difficulty doing a placebo-controlled study against no treatment. So my question -- and I'm obviously very interested in the answer -- is: Is this something you have to show once and then it works for all of them, or do you have to sort of show it in each setting?
CHAIRMAN BRAUNSTEIN: Dr. Cummings.
DR. CUMMINGS: I don't know that we have a large enough database to answer it across things like women and men and steroid-treated patients and other treated patients.
There is, in cardiovascular disease as you know, pretty consistent effects across classes of patients for the relative risk. So the relative risk tends to remain constant for an intervention across various classes of patients: men, women, ages, and such like that. That allows you to generalize from those to the cardiovascular drugs you're used to.
There have been at least two or three trials now finding an interaction, however. That is that women who are more severely affected seem to have a greater relative risk of reduction for non-spine fractures, and women who are in low-risk populations in fact don't have a reduction in relative risk of non-spine fractures. So with the bisphosphonate class, it may not be generalizable from high-risk to low-risk patients that it works.
And those have been reinforced by interaction terms. We don't know where that threshold is, but, no, you can't generalize from high-risk populations to low-risk populations for that one outcome. Otherwise, the database is not large enough.
DR. TEMPLE: Can you distinguish though between qualitative interactions and quantitative interactions? I mean, most people believe that in a lot of settings, yes, one group might be somewhat better affected than the other. But it would be a big surprise if it went the wrong way, which has major implications for how many of these studies you have to do.
DR. CUMMINGS: There are not enough events in the low-risk patients to say that it goes either no effect, or the wrong way, or the right way. The estimates are close to one for the low-risk populations.
In one trial that I know of, the Fracture Intervention Trial No. 2, it was done. And there are a couple of other examples where there is a much stronger effect in those with very low bone density than those without. So that proposition that the relative risk remains the same regardless of the risk of the population in the bone density, it does not seem to hold within osteoporosis for bisphosphonates.
But the major statement is that we don't have the same volume of data that you do in cardiovascular disease to be able to generalize.
DR. TEMPLE: That could imply you believe that you may have studies in severely -- in very high-risk people and that won't really tell you anything about the lower-risk people at all. So you have to do another study --
DR. CUMMINGS: The efficacy in fracture reduction.
DR. TEMPLE: Right.
DR. CUMMINGS: Again, the major point here is that the database is very limited for the other kinds of extrapolations.
CHAIRMAN BRAUNSTEIN: Dr. Grady is next.
DR. GRADY: I just want to bring us back to the bigger issue here, and I'll just give you my opinion. The question is: Is bone density an adequate outcome? And I think the question is: Is it an adequate outcome to register drugs for prevention of fracture?
So let me just first say my thinking about these things is quite different if we're talking about prevention than if we're talking treatment. So in terms of talking prevention, we're talking essentially about treating mostly women who are not symptomatic, particularly if we're just talking about low BMD.
I personally have never understood the difference between an indication for treatment of osteoporosis and management or treatment of osteoporosis. Osteoporosis basically is low bone density. It's really another surrogate outcome for risk of fracture. That's just an aside.
But let me just say when we're talking about prevention, I think we need to be much more careful that the benefit outweighs the risk than when we're talking about treatment of symptomatic conditions. I heard Dr. Orloff say that for a good surrogate, it should represent the outcome in a way that's robust, repeatable, and reliable, which nobody has convinced me that BMD is today. And, secondly, there shouldn't be much of a possibility that this surrogate could have harmful effects, which, we've been discussing that there may be in certain situations.
So I would just say I'm not convinced that BMD is an adequate surrogate for fracture prevention. And I personally think that new drugs, which are going to be registered for prevention of fracture, should be shown to reduce the risk of fractures.
Going beyond that, we get into some more difficult issues like once a drug has shown fracture prevention, should we then approve it in a different risk group. It's also in my mind a more difficult question as to whether or not, once we have several drugs in a class that all have shown fracture prevention, whether or not we need to continue requiring fractures as an outcome. I think those are questions that are much more difficult to discuss.
But in terms of new drugs -- and I guess for right now I would include the bisphosphonates because I don't think two drugs in the class is enough. I think eight statins is quite enough, but I personally don't think two bisphosphonates is. So I'll just register my opinion that I don't think BMD is an adequate outcome.
CHAIRMAN BRAUNSTEIN: Dr. Bone was next.
DR. BONE: Thank you. Dr. Cummings --
DR. ORLOFF: Make that a little clearer. No, I'm kidding.
(Laughter.)
DR. BONE: Don't be shy, Deb.
One of points that Dr. Cummings was just addressing was the generalized ability from high-risk to low-risk groups. But let's turn that around.
Steve, what would you say about our ability to generalize from relative risk reduction in the low bone density category to the patients at greater risk for postmenopausal osteoporosis?
DR. CUMMINGS: We have very little experience. I mean, I'm more optimistic about from what I've seen in the data. I think that that would probably apply, but we just don't have very many trials to generalize from.
We know, for example, we can't generalize from a reduction risk of vertebral fractures to a reduction risk of non-spine fractures because they're a lot of exceptions to that rule. So seeing a vertebral fracture reduction in a low-risk population doesn't necessarily mean that we'll see reductions of other kinds of fractures.
DR. BONE: But within the category of vertebral versus vertebral or non-vertebral versus non-vertebral?
DR. CUMMINGS: If something reduces the risk of vertebral fractures in one population, I think we've got enough consistency across these databases that I would probably believe that would work for everyone. But we just have too many exceptions to the rule for other kinds of generalizability.
So if you're asking, Henry, we see a vertebral fracture reduction risk in low-risk populations, just densinometric osteoporosis, I believe that that applies to everyone. But I think that's the only one I could buy.
CHAIRMAN BRAUNSTEIN: Dr. Watts, then Dr. Marcus.
DR. WATTS: I want to agree with Dr. Grady's conclusions but clarify some of the terminology.
The way the current labels read for agents on the market prevention is a drug that prevents bone loss in people who are normal to start with. So that's preventing someone's bone density from dropping below an arbitrary line, and that's not related to prevention of fracture. When I treat patients, I'm interested in preventing fracture and those are drugs that are currently labeled for treatment of osteoporosis.
DR. GRADY: But what's the reason we should be interested in preventing loss of bone mass?
DR. WATTS: That's a different question.
DR. GRADY: No, it isn't. It's to prevent fractures, isn't it?
DR. WATTS: But there's actually no data to show that the agents -- that the populations that have been treated for prevention of bone loss are actually protected from fracture later on. I think the discussion today is for agents that are being considered for treatment of osteoporosis, which means reduction in fracture, not just prevention of bone loss.
A second minor point -- or it may seem minor -- but I think it's unethical and impractical to do a study of osteoporosis in a low-risk population. This is relative. It needs to be a lower risk or a higher risk population. But if it's a low-risk population and they have essentially no risk of osteoporotic fractures, then it doesn't make sense.
DR. BONE: I meant minus two and a half.
DR. WATTS: I understand, but I just wanted to clarify it.
And finally, my bottom line is I'm not comfortable with BMD as the only marker. If we have a drug in class that has been shown to reduce fractures and we can show that a different regimen of the same drug or perhaps another drug in class has the same effect on density and turnover markers, I would be much more comfortable and I would be willing to extrapolate then antifracture efficacy and postmenopausal osteoporosis to antifracture efficacy and glucocorticoid-induced bone loss or in male osteoporosis provided those same surrogate endpoints, both density and turnover markers, showed a similar response.
CHAIRMAN BRAUNSTEIN: Dr. Marcus was next.
DR. MARCUS: I would like to introduce just a little bit of a commercial reality testing here.
I think there's a stakeholder in this field -- and I congratulate the agency on having representatives from all the stakeholders in the osteoporosis field. For the first time to my knowledge since my association with the panel, there have been representatives from industry invited to sit at the table, and I think that's terrific. However, there is one stakeholder that isn't here, and I think that that actually lets one point of view not be expressed. That is the third-party payors.
I'd just like to point out from my experience when Fosamax first came on the market that I spent hours on the telephone trying to convince third-party payors to allow patients to receive this drug. The resistance towards receiving that drug was astonishing, particularly in the state of California where managed-care elements were extremely potent.
I think that the tendency on the part of payors has been to define the narrowest possible group of people for reimbursement of pharmaceutical interventions for osteoporosis. If you were to define a treatment effect based on a very low-risk population, I don't think there's a snowball's chance in hell of it ever being reimbursed by third-party payors. And, furthermore, this just gives them exactly the opening they want to turn down high-risk patients because they would say "Well, you haven't shown it for high-risk patients." Therefore, I see the prospect of investing in a large-scale study in a so-called lower-risk population to be extremely unattractive to industry and something which is basically a non-starter.
CHAIRMAN BRAUNSTEIN: Dr. Rodan.
DR. ORLOFF: Dr. Marcus, first of all we appreciate your recognition of our recognition of all the stakeholders. We actually had not realized you had gone over to industry.
I'm kidding. That's why we asked you here today.
(Laughter.) DR. MARCUS: Yes, you did.
DR. ORLOFF: Could you please move to the back.
(Laughter.) DR. ORLOFF: No. I wanted to make sure that we understood you that -- are you actually saying that even evidence of fracture risk reduction in a low-risk population might not be extrapolated as proof of principle to a higher risk population as far as a third-party payor is concerned?
DR. MARCUS: As far as I can trust the insurance industry, that's exactly what I'm saying.
DR. BONE: Can I just clarify one point, and Nelson corrected me on this. When I said "low", I should have said "patients who just meet diagnostic criteria for osteoporosis". I didn't mean to imply the osteopenia population or somebody like that. I meant to refer to patients who would meet diagnostic criteria for osteoporosis but did not have a recent fracture or multiple fractures. And we could talk about whether one remote, not very bad fracture put them on one side of the line or the other, but the lower-risk or moderate-risk population.
DR. MARCUS: I think I said "lower". And it's a continuum. I think the closer you get to people who are truly in need of a drug, from above average to a very high-risk, the more likely you are to have a meaningful experiment that will result in a viable product.
CHAIRMAN BRAUNSTEIN: Let's hear from Dr. Rodan and then I'm going to try to focus some of the questions a little bit more if we can.
Dr. Rodan.
DR. RODAN: Going back to BMD, when I start a lecture on osteoporosis, I say that "science starts when you can measure something." Frances Bacon, 1700.
For osteoporosis, this started when people were able to measure bone density. A lot of data -- and Dr. Cummings contributed quite a bit to it -- showed a very close epidemiologic link between BMD and fracture risk, not only cross-sectionally but prospectively as well. And this is why before '94 this was an acceptable standard. The correlation broke down when etidronate and fluoride didn't follow this paradigm. We know very well today why this was the case.
Again, for agents that changed BMD, the tight correlation which was shown epidemiologically, was not the same. Actually, it went in the opposite direction. A smaller increase in BMD produced a larger fracture protection, as pointed out by Dr. Khosla. So there is enough science there, and physics supports that, to relate BMD to fracture risk.
Now the Ibandronate example doesn't fit this paradigm. My understanding is that Ibandronate is now submitted for approval at some agencies. It has fracture risk when given with a different regimen. So this modification, which Mike McClung suggested, that we should take into account the ways the drug was given and maybe include the suggestion by Dr. Lukert of biochemical markers as an additional criterion, may correct this problem.
Again, if you go to preclinical studies, they can indicate if the bone is normal. And if the bone is normal, then the BMD increases regardless of the weaknesses of BMD because it looks at the cross-sectional picture rather than at the structural picture. The bone risk is not going to increase. The quantity by which this will reduce fracture risk cannot be predicted from preclinical studies. This maybe can be taken care of in the label. And so this is sort of to put this in together somehow.
CHAIRMAN BRAUNSTEIN: Let me make a stab of trying to focus the discussion a little bit more. Let's make the assumption that every drug, whether it's of an existing class that's been shown or a new class of drugs coming through, will have adequate preclinical studies that will show that the type of bone that's produced is normal bone, it's got good tensile strength, and all those things that are done that are absolutely required before a Phase III study is done for the European regulatory agency.
So all that's done. Let's just make the assumption of good bone quality to start with. And then let's break down these four different classes here to two classes. One being the existing class of drugs that are on the market. For instance, the bisphosphonates or the SERMS. And the others are new classes of drugs that are not yet on the market such as new mechanistic action and a new bone anabolic agent. So those are the unknowns at the present time.
And then we have three groups of patients: those who have a low bone mineral density, two and a half standard deviations below the young adult mean on bone mineral density but no fractures; a second group of individuals who have a single fracture, age undetermined; and a third group who has either had recent fractures or multiple fractures.
So those are the three groups because those are almost three different populations, although they may be along a continuum in addressing what type of evidence we would want to see.
And then we have two other variables: the acceptance of bone mineral density as a surrogate for those groups, or acceptance of fracture as a surrogate. I would stick in with the bone mineral density "with evidence of" -- if it was an antiresorptive agent -- "depression of antiresorptive markers." And if it was an anabolic agent, simulation of anabolic markers. So that will make it even tighter.
So with those type of variables and assumptions, what I'd like to do is sort of go around the room and ask you to express your opinions on this having heard what you've heard. And if you have no opinion, just pass. This way I'd sort of like to get a feel of the group of what people think, given the state of the knowledge, would be the best way to approach the endpoints, the primary endpoints.
Sundeep, we'll start with you.
DR. KHOSLA: Well, I guess I kind of mentioned some of my views on this to begin with. But I think for new compounds in established classes, like a new bisphosphonate or a new estrogen or SERM where we understand now very well the molecular mechanisms by which these drugs act -- and you've already alluded to the fact that there's adequate preclinical data on all of these --
I think you can make a case that, provided you see the expected increase in bone density and the expected reduction in bone turnover markers, that, in fact, may be a reasonable surrogate for many situations.
CHAIRMAN BRAUNSTEIN: For all three situations, those with multiple fractures, and those with no fractures below bone density?
DR. KHOSLA: Scientifically, I don't see a fundamental difference in terms of how these drugs are going to act on bone in those three circumstances, so I don't feel uncomfortable feeling that that combination is probably adequate.
I think when you started talking about classes "C" and "D", where you've got unknown or new molecular actions or an anabolic agent where we have much less experience -- I mean the two anabolic agents that we've studied has been fluoride, where we know the bone quality was abnormal, and PTH which seems to follow the BMD fracture relationship. But I think there I'm a little less comfortable because we don't have the body of data that we do with the other two classes, and, there, we may obviously want to exercise more caution and get more reliable fracture data.
CHAIRMAN BRAUNSTEIN: So would you accept a bone density for approval but then require continuation to fracture endpoint, or just fracture endpoint for approval?
DR. KHOSLA: For classes "C" and "D"?
CHAIRMAN BRAUNSTEIN: Yes.
DR. KHOSLA: I guess I would. Provided all the preclinical data is there, you could make the case that you would accept bone density with the fracture data pending.
CHAIRMAN BRAUNSTEIN: Okay. Mike.
DR. MCCLUNG: I would alter my opinion about that a little bit. Again, I think I agree with Sundeep that the three classes of patients or the categories of patients don't influence my thought very much other than the ethical questions about which patients ought to be included in trials.
The clinical impact of an additional vertebral fracture is a function of the number of vertebral fractures the patients have at baseline. And so for patients who haven't had a vertebral fracture or have had a small or remote vertebral fracture, the clinical impact that is measured in a variety of studies is very small; whereas, if the patients had multiple fractures, an additional fracture is a substantial thing. And doing a placebo-controlled trial among patients with very severe pre-existing fractures under any circumstances in my personal view is not attractive.
CHAIRMAN BRAUNSTEIN: I set these three up in anticipation of getting to the placebo trial issue.
DR. MCCLUNG: Right. So back to the specific question about whether BMD would be an acceptable endpoint, again I think that in classes of drugs where we're confident about the relationship already between BMD and fracture risk, as long as the bone density -- and not only the magnitude but especially the pattern of change in bone markers is evident. The pattern may be at least as important as the intensity. And that accepting a bone density endpoint for registration of the drug, with the presumption that that would translate into fracture reduction, I think is adequate.
With regard to the new drugs where we don't know that relationship in clinical studies -- the relationship between changes in BMD and/or turnover and the relationship to fracture reduction -- I believe that we still need to have fracture endpoint as the primary determinant for approval for the drug. And then after we've established that in a set of trials with however many drugs Dr. Grady is happy with, then we can begin to amplify that data.
CHAIRMAN BRAUNSTEIN: Great. Thank you. Dr. Watts.
DR. WATTS: The wording of the Declaration of Helsinki was raised earlier, and I think there's one thing that's not really considered there that's relevant to the ethics of placebo-controlled trials. And that is what some call a diagnosis gap or a therapy gap, that there are many people out there who have osteoporosis who aren't identified and aren't treated. It's getting progressively more difficult to find patients who are suitable for these trials as Dr. Cummings points out.
And I think one of the reasons is that the low-hanging fruit have already been identified and are already on treatment. It's not as though we are taking patients from our clinics and putting them into these trials. I have never done that. We identify people for these trials from advertisements in the newspaper and targeted mailings and radio ads. These are often people who have not been tested for osteoporosis and probably wouldn't be tested and wouldn't be treated, were they not brought into a trial.
Having said that, I still have considerable difficulty on ethical grounds taking someone with multiple fractures or recent fractures and putting them into a clinical trial. But I feel quite comfortable in having patients with low bone mass alone or with a minor or remote clinical fracture receiving active drug versus vitamin D -- calcium and vitamin D.
It's my feeling that the only time I would be comfortable with surrogates -- and it's a combination of bone density and turnover markers -- would be in other dosing regimens or other clinical applications of drugs that have already been shown to reduce fractures. And let me quickly explain why I feel very strongly about that.
We're talking about class effects. How many amino bisphosphonates do we have on the market where this relationship is proved? Two. How many have been studied where this relationship wasn't proved? One, maybe two. How many selective estrogen receptor modulators do we have on the market where this relationship is proved? One. How many anabolic agents do we have on the market where this relationship is proved? Zero.
So I would be very happy if a drug that's been shown to reduce the risk of vertebral fractures in a lower- to moderate-risk population of postmenopausal women with osteoporosis is tested in men or glucocorticoid-treated patients and shows the same effect on bone density and the same effect on turnover markers. I would accept that in my clinic as adequate proof of fracture reduction.
I'd also like to highlight very briefly another nuance of this that has come out from Dr. Bone and others. And that is, once we've shown a reduction in vertebral fracture, we should probably be less stringent in our requirement to show reduction of non-vertebral fractures. Use a one-sided test instead of two. Use a more liberal statistical level and so on.
I think once it's out there that it's reducing fractures, as long as we see the trends for other fractures, we should be happy.
CHAIRMAN BRAUNSTEIN: Dr. Bone.
DR. BONE: Thank you. I just want to touch on a point that I touched on earlier to remind people that when we speak about placebo-controlled trials for osteoporosis, there are several things that should be born in mind.
First of all, we are really using a placebo injection as a mask for the active treatment. But there is background treatment for all the patients in all the trials. And this is a background treatment that has been shown to be efficacious in reducing fracture risk in a number of studies.
The other point is that we never withdraw patients -- essentially, in any trial that I've ever been involved in -- withdraw patients from active treatment in the way that patients are withdrawn, for example, in some other indications and replaced with a test drug. Patients who've had efficacious drugs in the past are just simply excluded from the trial unless they've been off them for years.
I am I think impressed by the consistency of the relative risk reduction for fractures, comparing vertebral versus vertebral and non-vertebral versus non-vertebral, that the high- and low-risk groups are consistent for the same drug. So I'm pretty well satisfied that in the patient who has bone density in the osteoporotic range with or without a remote, single, not very bad fracture, the information that we get about the effect on bone fragility is generalizable to patients treated with the same drug at higher risk. There is no good scientific reason to think otherwise in any case I'm aware of.
The other point I remind us all of is that in part what we're doing here is just confirming that the animals got it right, that there wasn't something that was distorting the relationship between mass and strength.
I think for initial assessment of any fracture efficacy, probably the lower-risk, placebo-controlled trial model is still very well within the boundary of ethical acceptability. And I think there is a consensus about this amongst people who take care of patients with osteoporosis as their main occupation (or very nearly so, within one standard deviation anyway, give or take), exactly who would be included for example.
But I think there is equally a consensus that the patients who are at particularly high-risk, such as those who have had recent or multiple fractures, are not in the category that we would include in such a trial.
I also think that they are probably not in the category in which we ought to initially evaluate antifracture efficacy. I think you can make a case that we ought to have at least some evidence of antifracture efficacy in the first kind of trial I was discussing before we take on the problem of the very high-risk patient.
And so I think that comparative evidence of efficacy can be obtained in an active control trial in high-risk patients in a later stage of development, and that would be my preference. There might be exceptions to this. I certainly could imagine that that wouldn't universally be my position, but sort of the first crack.
I think that what evidence we would require for initial registration is inexplicably linked to what evidence doctors need in order to make good, intelligent, clinical decisions. So we might very well distinguish between the minimum level of evidence that the agency might require to conclude that a drug is safe and efficacious and the level of information that a doctor might require in order to practice medicine in his or her clinic and make a decision about whether to use drug "A", drug "B", or drug "C" in a particular patient's case.
So I think that we will want to have evidence of antifracture efficacy for clinical decision-making even if we don't, strictly speaking, require it in every case for the initial registration. I think this is in part addressed by the current guidance, which isn't so bad after all when you read it.
In the current guidance, the U.S. provides for the ability to register the drug on the basis of clinical trials where BMD is the primary endpoint provided the fracture data is showing a favorable trend in a fully enrolled, ongoing trial. And I think this helps us with the drugs where we're not quite comfortable enough to say, okay, BMD is all we need, but where we really want to be able to move along and we're not so worried after all about those.
So it was a belts-and-braces approach when it was undertaken, and it still has some utility.
CHAIRMAN BRAUNSTEIN: So on a new bisphosphonate, you would want not only the BMD but you'd also want fracture data?
DR. BONE: I think that the fracture data will be essential, and the agency is going to have to rule on whether that is a registerable claim, whether it's the initial registration criterion or not. I'm saying it may not be as big a distinction as it sounds like in the first place in that category.
CHAIRMAN BRAUNSTEIN: And the same with SERM?
DR. BONE: With respect to SERMs, we have a more pleiotropic category of drugs here with hundreds of actions, potentially, on every single organ system practically. This is a more complicated situation. We have a pretty specific idea now about how amino bisphosphonates act primarily. Although as Graham Russell has recently pointed out, local intraskeletal pharmacokinetic differences may actually make a fairly big difference between, in certain respects, between drugs within the class.
But for SERMs, I think what is required here for us to make a good intelligent decision is how specifically characterized the mechanisms of action of the SERM would be. In other words, are all of the skeletal effects exactly those that are mediated in the way that we think they are by an estrogen-like action?
Now this something that Dr. Rizzoli has talked about, how there are some changes that may be individually drug-specific and not be entirely class-characteristic. I think this is particularly important because, of all the drugs we're talking about, this as a class has the greatest potential for use in the prevention of postmenopausal bone loss, as opposed to intervention after that has occurred.
I think that the characterization specifically here could lead to the decision that the initial indication could be, in some instances, for prevention of bone loss -- much as was the case for the registration for Raloxifene -- and that would be have to be purely on a bone density basis with no adverse safety profiling on fracture; whereas, the use of such a drug as a second-line drug for treatment of established osteoporosis, that might not be the main intervention with a particular drug in the so-called SERM category, depending on how it profiled out.
CHAIRMAN BRAUNSTEIN: Let me see if I have this straight. For bisphosphonate, you'd want bone mineral density and fracture follow-up. For a SERM, you want just bone mineral density?
DR. BONE: Well, no. What I'm saying is if we think we're going to use -- if we have characterized the SERM very precisely as having an estrogen-like action and we're proposing it to be used, not so much initially for treatment of established osteoporosis but for prevention of bone loss, then it becomes a different question. There, the only endpoint you can use is bone density.
But this depends entirely upon the preclinical characterization being bullet-proof. Otherwise, you're back to a bigger problem.
CHAIRMAN BRAUNSTEIN: If it's going to used as treatment for osteoporosis B-
DR. BONE: For the indication treatment of osteoporosis, you have a bigger problem because we've had this discrepancy between vertebral and hip fracture. And I think, there, you're forced to rely on fracture rate.
CHAIRMAN BRAUNSTEIN: Okay. So you'd want both bone density and fracture?
DR. BONE: Yes.
CHAIRMAN BRAUNSTEIN: Thank you. And what about the other two classes?
DR. BONE: Yes, I'm sorry for going on quite so long.
I think that in the case of an anabolic agent, it's practically inconceivable that an anabolic agent, given at several times the intended therapeutic dose, will produce no disturbances in the histology or mineralization and so forth of bone. So I think this absolutely falls in the category where we'll learn more about these drugs, and we may eventually change our mind about bone density, but for now, that has to be a fracture endpoint.
For the new mechanistic class of antiresorptive drugs -- in other words, something totally novel unlike any of the ones we have on the market -- I think that the current guidance is a reasonable starting point, which is what I just described a few minutes ago.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Worcester.
DR. WORCESTER: As the Consumer Rep, I'm going to try and look a little bit more at the bigger picture and particularly draw the point that, speaking on behalf of women, what women want to know is what's going to actually reduce fracture.
The idea that just because you can measure something and manipulate it, doesn't mean a whole lot to women who want to know what difference it's going to make in their lives. And, certainly, the kind of information that those of us who teach and work with consumers find out is that a lot of people are very confused with the information out there and really want the safest kinds of products.
So what I've heard today has really fit in with where I came to the meeting, with thinking that at this particular moment in history, probably we have never known better what we don't know, both in terms of what we're intrigued about in terms of the relationship of bone mass measurements related to osteoporosis, fractures, and other things, but also the whole set of products that we're talking about today, particularly how this particular issue fits in with other things.
I'm here representing not just people who may want and need treatment but also the masses of healthy women who are very confused about what to take. So I want to just comment that a lot of other organizations -- now we've heard how NIH and both the British Columbia Office of Health have in the last couple of years come out with pretty strong statements about needing a lot more information on the relationship of bone mass density to osteoporosis.
And so I think we might be before our time if we were jumping in and saying we knew more about what it meant in terms of fracture reduction. But also in terms of the Women's Health Initiative this summer, I think what it is a reminder of is even when we think we know quite a bit about products, we want to know the next step.
So I would come down to saying a couple of points. Long-term safety, looking at fracture reduction is what's going to mean a lot to the people who are going to use this. And then I want to come back to what I've heard several other members of the committee and our presenters talking about today.
I feel much less comfortable than other people grouping all the people who are going to use these products together. I think we've heard several times in a very persuasive way that the differences between different women make a huge difference.
Dr. Faulkner this morning in movingly telling us how important bone mass measurements are said, "It's not fair to be lumping all those groups together." I, in some ways, feel the same. We need to be looking at the difference between prevention and treatment.
And Dr. Watts said, "There's no evidence that bone mass density serves as an endpoint for the prevention of osteoporosis." So I think we keep hearing in different ways, we don't want to combine everything together right now.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Zerbe.
DR. ZERBE: Yes. At this point I mean it's hard to say anything new.
CHAIRMAN BRAUNSTEIN: Then you don't have to say very much.
(Laughter.) DR. ZERBE: I'll just make a general comment that, of course, we've been working for a decade looking for surrogate markers. And looking at it from the outside, this is actually one of the better documented surrogate markers for efficacy. As we look to try to simplify clinical research and bring therapies more rapidly to patients, this looks like an opportunity.
Nevertheless having said that, as the very learned people here have analyzed, I think there's really not a consensus at this point that you can accept, in any simple way, bone mineral density to replace fractures. I think the exception is in the area of bisphosphonates perhaps. I would draw the parallel to the lipid area.
As I sit down and look at it -- and Dr. Orloff made the point very well -- if you look at statins, a class of drugs approved, the approval has been based not on endpoint but on surrogate markers, cholesterol. And as you go through parallels, what was known when these additional agents were approval, what subsequently has been required, whether this explains all of cardiovascular disease, there are many parallels.
It's hard for me to distinguish why one would accept lipids, the parameters for lipid approval being cholesterol and not accept bone mineral density in a class of drugs that were understood and you had the preclinical bone mineral density data and preclinical morphology and strength to support the point.
So I guess if someone could provide clarity on that and explain the difference, it might help me understand the resistance to that.
CHAIRMAN BRAUNSTEIN: David.
DR. ORLOFF: Indeed, I might offer that we actually talked a little bit about this at lunchtime. With regard to the comparisons between the statins and let's say the amino bisphosphonates, there is actually, with regard to the mechanic of antifracture efficacy, there is probably a lot more known and there's a lot more bits of information to be garnered in any individual case under investigation in order to permit comparisons of a new drug to a drug that has established itself as an antifracture therapy than exists for possible comparison between statins.
For example, we don't have arterial biopsies for patients treated with Lipitor to compare the effect of Lipitor on the potential to reduce cardiovascular outcomes to Zocor. And yet, not that we've allowed it, but people are using Lipitor like it's -- you know. I don't want to say "water". I got in trouble for that one time.
(Laughter.) DR. ORLOFF: I think it is fair to say that the extension of that to the SERMs is not as clear-cut for many reasons that were put forward. The SERMs were designed to have differential tissue effects. So the whole idea of translating from one to another doesn't seem as solid. But with regard to the bisphosphonates, it seems that there's a pretty good case that bone mineral density could be the basis for that.
I'm a little bit less comfortable, frankly, with some of the things proposed about bone mineral density being used for formulation or indication extensions. Because the one example that was cited, as I understood it, was a formulation difference where there was dissociation between bone mineral density B-
PARTICIPANT: Different drug.
DR. ZERBE: I thought subsequently they had filed for that drug, I thought somebody said.
DR. BONE: There was a very different treatment strategy. It was not a formulation difference. It was a difference between giving a dose every three months and giving continuous dosing. So there was a majority difference in the strategy.
DR. ZERBE: So it may still hold the formulation. You can still raise the question about other indications like steroid- induced osteoporosis versus postmenopausal, just to highlight that point.
CHAIRMAN BRAUNSTEIN: Dr. Temple, you had a question or comment.
DR. TEMPLE: Well, some people have certainly said that if you wanted to go from a once-a-week treatment to a once-every-three-month treatment and you had fracture data on the once a week, it might be reasonable to use bone mineral density to go to the change.
So that means for every change like that, you've got to redo the fracture data.
DR. MCCLUNG: No. But it's not bone density by itself.
DR. TEMPLE: No, no. I understand.
DR. MCCLUNG: It's bone density and markers --
DR. TEMPLE: -- and turnovers and markers. Right. But you would have to redo the fracture data.
DR. WATTS: Let me try to clarify the issue that I raised. The two amino bisphosphonates that are approved are Alendronate and Risedronate. They've both been shown to reduce fracture risk with daily oral dosing. And within the last two years, both have been approved with once-a-week dosing on the basis of equivalent changes in BMD and bone turnover markers. I think the latter is the key to my acceptance of this.
Ibandronate is another amino bisphosphonate that is not currently on the market in this country. In the initial trial, it was powered to be fracture study with every-third-month intravenous injections that showed a significant gain in bone density. And if you looked only at the samples drawn pre-dosing, there was a reduction in turnover, and it didn't reduce fracture risk.
Almost simultaneous with that, there was an oral dosing trial that appears to show a fracture reduction and is going forward. In retrospect, the IV dosing is said to be maybe not the right dose. Maybe it should've been a higher dose, or maybe every third month was not often enough.
And there's some data from Phase II and Phase III studies to suggest that the suppression in turnover was not sustained. And so if you're dosing every third month, not only would I want to see equivalent BMD changes, but I would want to see probably at least monthly turnover markers to show that suppression is sustained to the same degree that it sustained with the regimen that has been shown to reduce fracture.
DR. TEMPLE: So that's still a surrogate. It's obviously not fractures but a more sophisticated version.
DR. WATTS: It's two surrogates in combination.
DR. TEMPLE: Just for analogies with blood pressure, we usually ask that there be evidence that blood pressure is held down throughout the day, not just at peak. So there are similar kinds of --
CHAIRMAN BRAUNSTEIN: Let me just introduce a time check here. We have an hour and 45 minutes and still a lot of questions, so I'd like everybody to keep their comments brief. We're continuing to go around the room.
DR. HOCHBERG: Can I briefly make a comment?
CHAIRMAN BRAUNSTEIN: Sure.
DR. HOCHBERG: First of all, I want to apologize for being out of the room, but I was on a conference call for an NIH project.
But this morning I didn't talk about the relationship between changes in biochemical markers and bone turnover because I was asked to focus specifically on bone density. I agree with Dr. Watts that you need to really satisfy both of these surrogates in order to say that a new amino bisphosphonate, which looks like a duck and quacks like a duck, behaves the same way probably with regard to fracture risk reduction, that you need to see sustained suppression of bone turnover and comparable increases in bone mineral density.
And then, if you have an antifracture trial, you would anticipate that you would see similar degrees of fracture risk reduction.
CHAIRMAN BRAUNSTEIN: Okay. Dr. Levitsky.
DR. LEVITSKY: I have very little to add. I think everyone's opinions mirror mine.
I believe that in classes "C" and "D", the newer agents, that one must have fracture risk as an outcome. For the bisphosphonates, I would feel more or less comfortable if they were approved with just increasing bone density and the bone marker changes. But as a prescriber, I would find it very difficult to prescribe a drug that I could not say to a patient caused a decrease in fracture rate. Therefore, I suspect that drug companies will be driven by that as much as anything else to get the additional data.
The SERMs trouble me because I don't know what the new ones will do in other tissue. So I think maybe we need more data for that.
The only new thought is I was toying with whether the most severely affected people who are already on an accepted drug for osteoporosis, whether that group might be candidates for add-on trials with a newer drug with a different action.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Sampson.
DR. SAMPSON: As a statistician, I'm always concerned about using a surrogate. I always argue that one needs strong, scientific, and statistical evidence. The question here is whether bone mineral density can stand as a surrogate for fracture rates or fracture incidents.
I think one has to be very cautious in establishing surrogacy in this regard. I think that the issues that have been identified here in terms of whether or not the compound is in the same class in which there's been a previously established -- and it's been well-shown -- relationship between bone mineral density and fracture incidents, the animal data has to be supported for the particular compound of interest.
I hear dosing regimen is of particular concern, that it not be strongly or dramatically different. I hear also that to establish surrogacy or to support the surrogacy argument, one needs to look at, in addition, marker data. I hear that using fracture data as safety, but I would still look at fracture data even if I'm looking at bone mineral density.
In summary, I think that one just needs to be very, very cautious in this regard.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Lukert.
DR. LUKERT: First, who I would enter into the study. I would be comfortable entering a woman who, by bone density definition, has osteoporosis and has one remote fracture. But I wouldn't be willing to enter someone into a placebo-controlled trial that had a recent or multiple compression fractures or serious non-vertebral fractures.
With a proviso on the ones that I would enter that there would be the safety net that you would be measuring their bone densities periodically, and if you saw a threshold fall in their bone density, they would be removed from the study. Then I would feel comfortable with that.
I pretty much agree with what has been said about the Class 1 and 2. The new mechanistic drugs, like if we're getting into the osteoprotegeran group or the bone morphometric protein to things that really are new and we don't have a lot of understanding of what those are doing in humans, I think that not only would I want fractures, but I would want histology on a subgroup of those patients to make sure we know what's going on in human bone.
In those, I don't think we can be sure that happens in the animal could be applied to the human, although I certainly feel comfortable with that with the bisphosphonates. I don't think I have anything else to add to that.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Aoki.
DR. AOKI: Well, I'm pretty much in agreement with Dr. Bone and colleagues. I'd just like to emphasize the importance of the animal studies because in this presentation so far today, I have not heard a major dichotomy that bone strength in itself is not the best predictor of decreased fractures in humans.
So if bone strength and bone volume in animals correlate so strongly with a positive clinical outcome in terms of vertebral fractures and hip fractures, then the question is: What's the relationship to bone mineral density?
It seems to me it's very important to do those studies in animals with any of these drugs that we talking about to see if, in fact, bone strength is increased, especially in ovariectomized animals that are treated with any one of these four classes; and then also to see whether the bone mineral density can serve as a surrogate by doing the appropriate bone mineral density studies at varying doses. Then I think once you enter Phase I, Phase II in clinical trials, I think that if you have any surprises, the animal studies should be looked at again to see if an explanation can be obtained.
And then, finally, I think in addition to what we've all talked about in terms of the studies themselves, I think that all of these agents should be followed at least in the Phase IV fashion, that after approval has been given, that the companies be required to continue to look at the fracture data to see if in fact it plateaus like in like five years or actually increases.
CHAIRMAN BRAUNSTEIN: Thank you. I actually ultimately want to see fracture data on any drug that comes out that's going to be used for preventing or treating osteoporosis. Although I think from a regulatory standpoint, some may be approved based on surrogate markers.
With the bisphosphonate group, there are certainly a lot of patients who are being treated with permidronate intravenously on a monthly basis for which we don't have a lot of data, good data. There have been reports of Zometa being used on a once a year basis to increase bone density, but we don't know if that's going to result in decreased fracture rates.
So even with the bisphosphonates that are coming out, I would like to see ultimate fracture data. Although I would accept bone mineral density as a surrogate for getting initial approval but would require ultimately fracture data showing efficacy.
In regards to the SERMs, I also would like to see, ultimately, fracture data. I would be inclined to accept bone mineral density for a prevention type of indication. But certainly if it's going to be used for treatment, I would like to see the fracture data for efficacy.
And for the other new classes of drugs, I would like to see fracture data, and I wouldn't accept bone mineral density as a surrogate.
Dr. Gelato.
DR. GELATO: Okay. I'd like to see fracture data for all new drugs that come to the market because I agree that it would be a hard sell to tell a patient that, yes, the bone mineral density may increase, but I have no idea whether you're going to have a fracture or not. So that's important.
For the bisphosphonates, I feel the same way. I think if it's a new drug, even though it's in that class, we should have fracture data. I would accept bone mineral density and bone markers for dosing changes within a class. I think that that's perfectly acceptable because we've already established that it does alter or impact on fracture reduction.
I would not add patients who have multiple fractures into studies. I think they need to be treated, and I think that, in my mind, it's unethical to do that.
In terms of the SERMs, I feel the same way. We need fracture data in that class in particular because there are so many different effects that they have on various tissues. We really need to know what the safety margin is in these drugs if and when they come to market.
Obviously, we need to continue to collect long-term safety data on all of these drugs. Because as we've seen with the estrogens, several years were needed before we actually were able to say one way or the other what effects they had on coronary artery disease, stroke, and so on.
And for drugs that are truly new agents with new mechanisms of action, I agree with Dr. Lukert. I think we need to have bone biopsies and look at the histology of the bone and make sure that the bone that we are increasing is the bone that we want.
CHAIRMAN BRAUNSTEIN: Dr. Tamborlane?
DR. TAMBORLANE: I basically agree with that. I think that fracture data, except for an extension of an indication for an established antifracture drug and/or dosing changes and so forth.
CHAIRMAN BRAUNSTEIN: Okay. Dr. Grady.
DR. GRADY: I also agree with also everything Dr. Gelato said. I guess I would just make two more points.
I think I might also be willing to accept BMD and other markers for new classes of patients, particularly if it would be very difficult to get fracture information in those patients such as steroid-treated patients.
And the final thing I'd like to say is that I agree with what everybody has been saying about SERMs, and I'm assuming that we're including estrogen in that category, which makes me also point out that right now, FDA's guidance makes a special exception for estrogen which I think needs to be reconsidered.
CHAIRMAN BRAUNSTEIN: Dr. Abadie.
DR. ABADIE: My recommendation will be extremely close to colleagues.
I think for all types of drugs, new molecular entities and others, for us the fracture will be monitored in the initial registration for new molecular entities probably because, as has been said, we don't know the drug, we don't know the efficacy and safety, we don't know the relation between the BMD and risk factors, so it's clear that we'd like to see fracture.
For the old entities, we could potentially accept BMD for bisphosphonate. But as far as we're concerned in Europe, there would be an impact on the labeling. And we'll end up with a magnificent paper with a marketing utilization, but the drug will not penetrate the marketplace so it will be totally worthless.
For the placebo and the low risk, the data that I show seems to go along with the fact that there is some consistency between the low-risk and the high-risk patients. I think there is a possibility to extrapolate from the low-risk to the high-risk patients with the vertebrae or with respect to the vertebrae. Although the database is relatively small, I think we can extrapolate.
The potential to use placebo in low-risk patients for me is something, which as far as the EU is concerned, these are probably acceptable from an ethical viewpoint.
And finally the problem which is the most difficult and where I don't have any clear idea, I must admit, is the problem of the hip. I think that we have to be innovative. I'm not absolutely sure that the study of hip fracture in a high-risk patient is ethical with the placebo even if we put the calcium and vitamin D in and maybe some -- I don't design, but the ways, the potential impacts on the labeling could be appropriate.
CHAIRMAN BRAUNSTEIN: Dr. Silverstein.
DR. SILVERSTEIN: Okay. I'm really going to echo a lot of what's been said.
I have difficulty putting into placebo-controlled trials people I think need treatment. So, therefore, I think anybody who is a high-risk that I would treat needs to either go into an active control or an add-on study. I couldn't justify in my own mind putting them in placebo-controlled. The low-risk, intermediate- risk I could.
As far as bone mineral density, I think I feel fairly comfortable with bisphosphonates, new classes of bisphosphonates and new indications for those drugs that have already been tested using bone mineral density and turnover markers. But for all new drugs, I think fracture data is going to be important. But even in those drugs that use bone mineral density, I would think it would be a good idea to have something similar to what the growth hormone registries are, to have long-term follow up of fracture risk.
So you don't need the fracture risk for registration but you still have that data. You'll continue to follow those patients and get those results later on. I guess that's all I have to say.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Rodan.
DR. RODAN: Yes. I agree with things which were said around the table.
I think that for agents that have a known mechanism of action and act selectively on bone and inhibit bone resorption, bone mineral density reduction in markers could be used for initial registration, with fractures being done subsequently. Osteoporosis is a continuum of risk. And so proving efficacy in low-risk patients should probably be extrapolatable to high-risk patients, as already stated by Dr. Abadie and others.
For other indications, I think this can be extended if one agent has been shown to be effective for a particular indication and other indications, like glucocorticoid-induced osteoporosis which is similar to postmenopausal osteoporosis in many respects.
That's basically it.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Rizzoli.
DR. RIZZOLI: Yes. Regarding a new molecule with a new mechanism of action to have the fracture is mandatory now. Regarding the bisphosphonate and with the long history of the assessments, the relationship between strengths and BMD and bone turnover, this particularly offers a benefit for the patient because for instance, if you had a beautiful bisphosphonate with a 50 percent bioavailability, I would find it a little bit surprising to wait three years to have the fracture data and not to benefit the patient, provided the relationship between strengths, BMD and markers is as the other compound have been demonstrated as efficacious in fracture incidents.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Turner.
DR. TURNER: I'm going to focus my comments on the preclinical issues, which I have a bit more experience with.
I think that we've seen that the animal studies have predicted pretty well the safety issues that we face with these drugs. We haven't had a lot of surprises. In the case of ovariectomized rats in particular, we maybe overestimate the efficacy in some cases, but they've generally been predictive.
But they seemed to have missed in many cases on extraskeletal effects. This is a concern that is very important and is becoming more important, it seems, as we're learning more about estrogen. But we certainly missed on predicting extraskeletal pathologies in at least one SERM, if not a couple. There are the issues with estrogen. There are the extraskeletal effects of parathyroid hormone, which in fact were picked up by the animal studies.
So this I think needs to be considered very carefully in the clinical design as well. In fact, with some drugs, particularly SERMs and anabolic hormones like parathyroid hormone, trials may be designed with special attention to extraskeletal effects and the power calculations.
I want to hit on a couple of nuances when it comes to the very specific, potentially new molecular targets or mechanistic classes brought up here -- not necessarily just for antiresorptive agents but also for new anabolics that may follow parathyroid hormone -- and that is that if a target is very specific molecularly, it may be targeted to the human, and there may not be sufficient crossover with many animal species.
Currently, FDA guidelines require two analyses in two species. It may be in some cases that only primates or maybe only a select number of species actually respond to the treatment. So there may be some need for rethinking some of the preclinical guidelines when it comes to some of these new, very targeted compounds.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Hochberg.
DR. HOCHBERG: Well, thank you for allowing me as a guest to make some comments. Basically, I think it's difficult at the end of a U-shaped table, but I pretty much agree with the Chair. But I would like to bring up maybe a couple of small issues.
I think Dr. Rodan makes a strong point that one could generalize amino bisphosphonates into a "class effect" with adequate preclinical data and data from clinical trials that demonstrate similar changes in bone mineral density and similar reductions with continuous suppression of biochemical markers of bone turnover.
I think this is what was done with the statins for approval for the treatment of hypercholesterolemia, which is a risk factor just like osteoporosis is. And companies did not do pivotal trials showing reduction in myocardial infarctions to get approval of statin for the treatment of hypercholesterolemia.
I think you can really compare osteoporosis to hypercholesterolemia. You can also compare osteoporosis to hypertension. I'm not aware that drugs other than thiazides and maybe one or two others have been shown to reduce stroke incidence in clinical trials that have approved drugs for the treatment of hypertension.
So I think you could apply this class effect to amino bisphosphonates. I agree with you with regard to other agents.
This issue of prevention of bone loss, which is a separate indication, gets to this question of -- there's not a similar indication for prevention of hypertension in people who have intermediate blood pressure levels of, let's say, between 120 and 135 systolic and 80 to 89 diastolic. But we know that those individuals are at greater risk of having bad cardiovascular events as compared to people with low/normal hypertension from data published from Framingham, just like we know that people with intermediate levels of bone density are at greater risk of fracture than people with normal bone density in multiple variable-adjusted models from various epidemiologic studies.
So the prevention here is to maybe prevent further bone loss, like one considers preventing an increase in blood pressure or preventing the development of diabetes in people with impaired glucose tolerance or preventing hyperlipidemia in somebody with a mildly elevated cholesterol.
Obviously, we don't want to medicalize something, which is a laboratory test where a large proportion of otherwise healthy individuals fall into that group. But it seems to me that if one was going to allow approval for a prevention indication then it should be focused on the laboratory test, which is being treated, and not on a fracture outcome which requires a magnitude larger sample size in order to demonstrate antifracture efficacy for something which would need to be established first in people with osteoporosis. I agree with the comments about placebo-controlled studies in terms of enrollment of low-risk patients. The question about doing the studies to demonstrate efficacy for hip fracture reduction. We know that calcium and vitamin D are efficacious at least in residents of long-term care facilities for reducing hip fracture versus placebo.
So we really have an active comparator study if we can compare to calcium and vitamin D in the appropriate population. And we haven't really established hip fracture risk reduction in the very elderly with osteoporosis yet. So that's actually a population that could be studied.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Cummings.
DR. CUMMINGS: Thank you. I join the consensus that new bisphosphonates with the same patterns and same magnitudes should be, I think, allowed to be registered and develop new indications on the basis of that. But I don't think that's going to solve our problem. I think that that generates a lot of new me-too drugs. Unfortunately, I don't know that many sponsors are that enthusiastic about following that path.
I think that the field really needs a way to encourage the development and permit the development of new classes of agents. I think that's where the real issue is, not in the fifth and sixth bisphosphonate.
It's tough to figure out how to do that, but I think that the model of using moderate-risk patients, excluding high-risk patients, and encouraging placebo-controlled trials for those new classes -- and I would include in that new SERMs because I don't think we know enough about that class. But some ways that the agency might consider trying to make this easier might be simplifying or encouraging companies to simplify the nature of these complex expensive trials that are going on because I think that some of the costs could be dramatically reduced.
Then let's see. Anabolics. I think that that's a class where we have a unique opportunity to encourage the development of anabolics in testing as add-ons or comparison to usual care or current practice. I think we may be able to design adequately trials in the anabolic arena as add-on trials.
But my major concern is that the current guidelines really don't address the most important issue to me, looking long term. And that's that we really don't -- if this is a chronic disease that requires 10, 20, 30, 40 years of care and therapy, I think that we need to think carefully of strategies and require that, after registration, there be long-term adequately powered strategies to test the continued efficacy and safety of these drugs beyond two to three years.
Give sponsors the opportunity to get an initial registration for a year- to two-year studies, but make sure that the plan is built in and incentives are built in to make sure that we know how well patients are faring after five and ten years after use of this drug in otherwise asymptomatic, healthy people.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Marcus.
DR. MARCUS: I think it's simplistic to use the term that has been used many times around the table of "amino bisphosphonates". The fact of the matter is the two drugs which are currently approved are in fact amino bisphosphonates. The drug Ibandronate that was mentioned before is not strictly speaking an amino bisphosphonate, although it is a nitrogen-containing bisphosphonate.
There are bisphosphonates which are currently on the dock which are being shown as posters, which actually are very different in terms of their molecular structures although they do maintain the bisphosphonate linkage between the two pyrophosphate groups. I believe on is called Apomine, a bisphosphonate ester.
In other words, the heterogeneity of the compounds is going to be growing, and I would predict that the agency is going to be faced by compounds, which are not really amino bisphosphonates. For all of those, I'm afraid that I would have to support fracture trials because I can't rely on identity and mechanism of actions with currently available drugs.
I think that extension from drugs of known efficacy to different classes of patients such as glucocorticoid-associated osteoporosis, males in a drug where the efficacy has been established in women, and some other examples like that are perfectly appropriate to have a BMD surrogate endpoint.
I think that Dr. Lukert, Dr. Braunstein, and Dr. Gelato made compelling cases, though, based on moral, clinical, and other grounds that it is really more desirable to have fracture endpoint trials. Certainly for other drugs which are in the antiresorptive sphere, such as the integrin disruptors or other osteoprotegerin or cathepsin inhibitors, things which are coming along that line, it's a totally new ballgame and these have to be treated as brand new entities and have fracture-related trials.
Finally, with respect to the non-vertebral fracture, I would strong urge the Agency to permit lumping of non-vertebral fractures. I think that the nightmarish aspects of trying to do a hip fracture study have already been pointed out. In the interest of time, I won't go much more into it. But, I think it's perfectly appropriate to treat non-vertebral fractures as a class.
Thank you.
CHAIRMAN BRAUNSTEIN: Great, thank you. I'm actually going to combine three questions into the next set, and then there's going to be one final question at the end. So let me explain this and please feel free any of you who have to stretch, get up and stretch, and then come on back.
The next question is going to have to do with what duration of study is appropriate for the assessment of effectiveness, what duration of study is appropriate for assessment of safety, and what other specific safety monitoring should be conducted for those four classes of drugs that were described. So, that's one question that I'm going to ask everybody just to briefly answer.
And then the last question will have to do with the use of placebo versus active control. What types of groups of patients do the members of the panel, as well as the guests, feel would be appropriate to apply a placebo to or active control to or neither.
So let's go to the first question. Dr. McClung, what duration of study is appropriate for assessment of effectiveness, of safety, and what other safety monitoring should be conducted?
DR. MCCLUNG: Well, I think the duration for efficacy and safety are very different. Efficacy, at least with the classes of drugs that we know about, can probably be assessed in a year's time. Again, we've already made the caveat that the preclinical data are clean.
But for safety circumstances, both skeletal safety and extraskeletal safety effects may not show up nearly in that time, and there needs to, even if approval is granted, a plan to have long-term surveillance about safety type issues.
CHAIRMAN BRAUNSTEIN: So "long-term" is open-ended?
DR. MCCLUNG: I would say five years at a minimum that we need to have that duration because that's the duration of therapy that we are going to aim, with the exception of anabolic agents perhaps. At least for antiresorptive agents, long-term therapy seems to be necessary.
And then the other important piece of that is that once therapy is discontinued, we actually need to know what happens upon withdrawal to help clinicians and all of us understand how best to use the drugs. It may be that five years of therapy may protect the patient during the five years of therapy, but if the effect wanes very quickly, we need to know that because it makes a difference in whether we decide to continue therapy beyond that five years or not.
CHAIRMAN BRAUNSTEIN: Any specific safety issues, measurements outside of the usual?
DR. MCCLUNG: I think that the safety issues in terms of skeletal safety are already well put together. Obviously, the extraskeletal safety issues depend entirely upon the nature of the compound, what we learn from its mechanism of action, and what are suspicions are about what the side effects might be.
CHAIRMAN BRAUNSTEIN: Terrific. Dr. Watts.
DR. WATTS: Although vertebral fracture efficacy can probably accepted on the basis of a one-year trial, I think it's more difficult to establish the non-vertebral fracture efficacy.
I think one of the convenient things about a three-year trial of agent versus placebo is that it gives you an adequate opportunity to pursue safety, both skeletal safety and non-vertebral safety with the appropriate trap doors as Dr. Lukert mentioned. So patients with declining BMD, they leave the trial. They have an endpoint, a fracture, whether it's a vertebral fracture or non-vertebral fracture. They go on active treatment.
With the lessons of the recent Women's Health Initiative Study ringing in our ears, I'm not sure that we can really say that there's no lower limit of time to establish safety. We may never see a trial of that same magnitude that will give us the same detail of information.
But, I agree with Dr. McClung that it depends -- the length of safety for non-skeletal issues depends on your concerns about non-skeletal effects.
CHAIRMAN BRAUNSTEIN: Okay. Dr. Bone.
DR. BONE: Thanks. I think, first of all, that the observation period is likely to be somewhat influenced by the way the drug works and its pharmacokinetic characteristics and so forth. In other words, something that stays in the skeleton for a long period of time might be different, or it accumulates or it has cumulative biological effects, and might be looked at in a different way from something that acted very differently.
But with that having been said, I think, generally speaking, I'd like to have three years of information prior to registration. I think there may be instances in which that isn't the period of active therapy, and it might not be the period of blinded therapy. But that should be as a first cut the observation period.
I would certainly support the idea of extending the observation period for two more years past registration to look for changes. I think the even longer-term suggestion that Dr. Cummings has a lot of merit, but somewhere between five and ten years there may be a practical tradeoff. One of the things that can be done in that situation is a cross-over study between the placebo-controlled group and the active control, active arm, say, after three years so that everyone is treated after five years and you can look at a resolution of effect, a cumulative dosing for five years and so on, and get a lot of information in a very rigorous way from that kind of observation. I think generally speaking, that's how I would approach that.
I think the kind of data that we're collecting now in terms of density, morphometric fractures, clinical fractures, and markers of bone resorption and formation are the only ones I know about that would be appropriate. But, somebody will probably think of something else to add as we go along. For example, some of these ideas that are intended to look at structure.
The last question would be: How late in the game would you look at biopsies? You might want to consider looking at biopsies relatively late in the game if you have people who have been on a drug for five years, for example. That would be a nice opportunity to do what's being done in the continuation of the FIT trial.
CHAIRMAN BRAUNSTEIN: Dr. Worcester.
DR. WORCESTER: I think the shorter answer would be longer rather than short. In terms of finding out the safety issues, I think we may want to separate out what I will call "younger women". In anyone under about the age of 65, there might be different lengths of time that we would want to do prevention trials versus treatment.
And I also think Dr. Cummings idea of maybe tying in the approval to how long a drug has been studied and keeping it open-ended so that we can keep track of longer things but get good products on the market.
CHAIRMAN BRAUNSTEIN: Dr. Zerbe.
DR. ZERBE: It sounds like in a year to 18 months, one should be able to demonstrate efficacy, although there were some data presented that suggested the longer-term efficacy is something that does need to be considered, and therefore a controlled period extending up to three, to even longer, years probably is worth considering for the right classes of drugs. Obviously, open-label extensions are important and would have to be geared to the class of drugs.
One issue that hasn't been addressed that I think is worth pointing out is that we need to be careful that we're not too dogmatic. I think that there's a very good argument for the anabolic agents, for example, that continuous therapy for long periods of time may not be the optimal treatment, and that there needs to be enough flexibility in the guidelines to ensure that appropriate therapy, perhaps a rest period if you've got an anabolic agent would be a more appropriate and safer therapy. That needs to be also considered in the guidelines.
CHAIRMAN BRAUNSTEIN: Great. Dr. Levitsky.
DR. LEVITSKY: I agree with all and have nothing to add.
CHAIRMAN BRAUNSTEIN: Okay, great. Dr. Sampson.
DR. SAMPSON: With regard to efficacy, it sounds like the current two- to three-year duration for fracture incidents is reasonably appropriate. I would also like to agree with Dr. Cummings that, for intended long-term usage, designs or schemes be developed to monitor long-term efficacy and safety.
I'm certainly not prepared to say how many years one would want to do each of efficacy and safety. That would depend on the scheme and the plan.
CHAIRMAN BRAUNSTEIN: Dr. Lukert?
DR. LUKERT: Well, I would be in favor of two years to prove efficacy by bone density measurements just because I think you need four measurements to make sure that you have an accurate assessment of where the bone density is going.
As far as safety is concerned, I think with a drug like a bisphosphonate that's stored in the skeleton, I think there need to be monitors as long as the drug is used because we really don't know much about what those long-term exposures do to either the hematopoietic system or bone.
CHAIRMAN BRAUNSTEIN: Okay. Dr. Aoki.
DR. AOKI: I agree with Dr. Lukert.
CHAIRMAN BRAUNSTEIN: Thank you.
For BMD endpoints -- although my comments about fractures still hold -- but for BMD endpoints, I would want a minimum of a year, probably about 18 months to show an increase; for fracture endpoints, a minimum of three years; for safety endpoints, a minimum of five years, especially considering the wide distribution of these agents to a large number of the population who are also taking other medications.
I think the more information for the longest period of time, the better. I would encourage the companies to follow what was suggested about keeping at least a large cohort and keeping information on the large cohort for as long as the drugs are out in order to obtain continuous safety information and provide that. Again, I think that's very reassuring to both the doctors and most especially to the patients.
DR. GELATO: I don't really have anything to say different about the duration for safety and efficacy. I think two to three years for fracture is appropriate.
I would just like to make two points that I think were already said. But to reemphasize them, I think we need data on, when we stop the drug, what happens to the patient, what happens to the bone mineral density. Does it stay sustained? Do they lose? That data I think is very important.
And I really like Dr. Silverstein's idea about having a registry like there is for growth hormone where these patients, who as long as their on drugs, they're continued to be monitored and information gets put into a database that is accessible to people about efficacy as well as safety issues and adverse events and so on.
CHAIRMAN BRAUNSTEIN: Okay. Dr. Tamborlane.
DR. TAMBORLANE: Yes. I'd just like to follow up on that because I know that we've talked on this side of the room about how it's not so easy. You know, how would you interpret the long-term efficacy/safety data. I think it would take some good thinking to try to design it in a systematic way without a control group per se.
You're then talking about some substantial investment of time and effort if this is going to be done in a conscientious way, and it raises the question of: Are there ways that industry could be incentivized to really go after this in an effective way?
CHAIRMAN BRAUNSTEIN: Well, I think patients often are drawn to drugs that appear to be the safest of the class, and that information gets out and gets disseminated fairly rapidly.
Dr. Grady.
DR. GRADY: I'm in favor, I think, for fracture outcomes of three-year randomized trials. So a continuation of the randomized, blinded comparisons for three years perhaps for registration but then follow up for an additional two years, still with randomized, blinded design, for safety issues and then again some follow-up beyond that.
I think the idea of having a registry is a good one. But it's very problematic to make comparisons to a registry, so I think there should be some effort put into to trying to develop models perhaps using the placebo rates in the individual trials or coming up with some sort of decay rate based on multiple placebo groups and multiple trials to try to get a handle on this. I think Dr. Watts brought this up and, of course, so did Steve.
Finally, I think that companies should be required to make available major outcome data, and this would vary a little bit by the class of drug. But for SERMs, for example, there should be available numbers and annual rates of coronary events, stroke, venous thrombolytic events, et cetera to allow people to do systemic reviews and meta-analyses, which I think could be very helpful but are almost impossible to do with industry-sponsored trials because the data are not made available.
CHAIRMAN BRAUNSTEIN: Dr. Abadie?
DR. ABADIE: I think, although we can have efficacy data in probably a shorter duration than three years, I would probably vote for a three-year's data because I think that, apart of efficacy, safety is also important. Probably in three year's time we will have more data on safety and potentially also on efficacy, although I'm not sure.
With respect to the registration or to the post-marketing commitments, I have mixed feelings about the registry because I have already some examples of that in Europe. I think if we set up a registry, it raises for us a certain number of problems. We need to know exactly what we are looking for.
And the examples of registry that we have so far show us that if it is a fishing expedition, it's not that cost-effective, I would say. It's okay for something which goes beyond the marketing authorization with respect to efficacy and safety, but, please, let's know before what we are going to look for.
CHAIRMAN BRAUNSTEIN: Very good. Dr. Silverstein.
DR. SILVERSTEIN: I agree with duration of two to three years. I'd like to just touch on the registry issue.
I agree that you need to have certain adverse events, fracture incidence, et cetera, that you'd want to put into the registry. But, in addition, other adverse events could be added and then if people noted that they were seeing a lot of patients with pancreatitis or something like that, they could then query the registry to see how many other adverse events were in the registry.
I think that's really an invaluable thing for drugs that require long-term use, as these will. So I still want to make a plug for that.
I would also like to agree with whoever it was who said that certainly with the anabolic agents and possibly some of the newer agents, there should be a cohort perhaps five years down the road that had bone biopsy studies to see what they were doing.
CHAIRMAN BRAUNSTEIN: Okay. Dr. Rodan.
DR. RODAN: I agree with bone biopsies.
(Laughter.)
DR. RODAN: -- for agents that act selectively on the skeleton by known mechanisms and inhibit bone resorption. It's acceptable to register such agents initially based on bone mineral density. Probably two years would be sufficient with a proviso that fracture data would be collected over the next two years or three years, which means four years all together.
Now the two-year data could, as now in the guidelines, already show a trend for prevention of fractures, depending on the group that you selected for study. This will be very similar to the current guidelines but would reduce from three years to two years. Just remember that the three years is there because of etidronate problems.
I fully support what Dr. Cummings suggested, a long-term follow-up of patients, dependent on the agents, the agents that thin bone. We really have a responsibility to find out what's happening over time, and if mechanisms are different as well.
Now extraskeletal pathology, every agent has to be, by law, evaluated for its toxicity on all organs. This should be part of the package with this. However, for agents that we know based on mechanisms that act on other organs -- SERMs, estrogens, and so on -- there could be this additional burden of proof, which we now added on bone agents, to have additional toxicology for bone, which is basically what we're discussing.
It could be expanded for other target tissues or for agents known to have effects other tissues. For example, sex steroids and derivatives, maybe in reproductive tissues and so on.
CHAIRMAN BRAUNSTEIN: Dr. Rizzoli?
DR. RIZZOLI: The duration of the study might take into consideration the capacity of the drug to restore the strengths. And maybe with a very strong anabolic, duration of 18 months would be enough, as compared with something working more slowly. But in terms of safety, certainly, the observation should last far beyond the time of treatment.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Turner?
DR. TURNER: I think there's considerable evidence that a year or 18 months is enough to see efficacy. But there's very little information, at least to my mind, of how long you have to wait to really understand safety.
One issue that's of great interest to me is this waning effect of the bisphosphonate treatment that Dr. Cummings brought up. I should point out that we don't know what the long-term consequences are of reducing bone remodeling repair for 10 or 15, 20 years. Certainly at our institution, we're spending a lot of NIH's money to try to better understand this. I don't think anybody really knows how much you can reduce remodeling safely and how long you can reduce it without causing some potential adverse effects.
So, there are some mechanistic issues in the long-term that I don't think we fully understand. There are probably other perils that apply to anabolic agents, other agents as well. I don't mean to pick on bisphosphonates. But I think there are some major unknown mechanistic issues there.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Hochberg?
DR. HOCHBERG: I was just telling Dr. Cummings that I don't have a lot to add.
I think the data showed that you can demonstrate antifracture efficacy for vertebral fractures within 12 months. But if you're looking for non-vertebral fractures, it depends on how many people you want to study. That's the trade-off, as opposed to how long you want to study them because these are time-to-event studies. So, you just have to accrue enough fractures to be able to show that your reduction is statistically significant.
I think, as everybody has said, the longer you follow patients, the more safety data that you accrue. But even in the trials -- which are an order of magnitude larger than those that I see my colleagues do in rheumatology in patients with rheumatoid arthritis -- we still may miss some rare adverse events, which is why it's important to collect data post-marketing.
In my experience with various companies who have done trials in rheumatoid arthritis, they generally have not followed through on the Phase IV commitments to monitor patients long-term. There are mechanisms set up to do this, through patient self-report and observational studies, in order to collect these kinds of data.
CHAIRMAN BRAUNSTEIN: Dr. Cummings?
DR. CUMMINGS: I would flip things around. I would start with -- Where the Guidelines I think right now are weakest is in the planning of long-term observations, long-term trial -- long-term data about safety.
If there were a solid plan that had a biologically relevant duration of follow-up, five or ten years, then I think that in that context I would be happier about seeing a drug approved after one year of fracture data. I'm not sure whether it's one year or two years or three years -- Gideon and others have pointed out that that depends on the biology of the drug.
I would really resist the notion that you would set up a single three-year, four-year, or two-year arbitrary time point. I would say that that's where flexibility is warranted; it's at one, two, or three years, depending on the biology of the drug. And in the context -- the most important thing is that you establish a context for a biologically relevant long-term period of observation compared to the initial control experience.
When it comes to registries, the best registry is the treatment cohort within the randomized trials that you've got. Just registries of people on drugs are not -- they are really problematic to analyze. So I think that that long-term plan is best established as very long-term, biologically relevant long-term follow-up of the treated groups within the randomized trials, with whatever is necessary.
And one last thing, what's the incentive for industry in doing this long-term? I mean, there are precedents. But I mean, with Tamoxifen, you've got a five-year indication. You can use Tamoxifen for five- years, right?
DR. TEMPLE: Yes, but do you know how they got that?
(No response.)
DR. TEMPLE: They randomized people with -- they continued therapy versus stopping therapy --
DR. CUMMINGS: That's right.
DR. TEMPLE: -- a randomized, controlled trial, not a registry.
DR. CUMMINGS: Okay.
DR. TEMPLE: Not a registry, but an adequate plan. Needless to say, to link the approval to the duration of evidence that's in front of you. In other words, approve it for -- where in the label this is approved for two or three or five years or ten years use or indefinite use. I think it depends on the nature of the evidence you've been provided, and I would link those two in some fashion.
CHAIRMAN BRAUNSTEIN: Dr. Marcus?
DR. MARCUS: With respect to safety, I agree with Steve and Marc Hochberg that really for a vertebral fracture, if that's what you want your indication to be, you can have a good endpoint within 12 months.
That being said, I also agree that to get non-vertebral fracture data is probably going to take a longer period of time. Furthermore, I agree completely with Dr. Gelato with respect to the off rates.
There are some drugs whose effects are lost relatively quickly, such as estrogens. There are other drugs in which bone mineral density as well as antifracture efficacy appear to be enduring. They're robust even after a period off drug. Teriparatide is one that appears to be like that. Therefore, it would be sensible to have in a trial a period of -- maintained on the trial after termination of the drug. So, I can imagine that somebody might be on drug therapy for two years, with then a third year of follow-up, of continued examination of efficacy.
With respect to safety, there are free safeties and strong safeties, as you all know. The strong safety is what you get within the course of your randomized-controlled trial. You get all the up-front safety events related to the treatment.
The more distant safety events are very hard to get. And with all due respect to Dr. Grady's idea of keeping people on randomized drugs out to five years, I think that that's a non-starter. It's a non-starter because number one, it would be unethical after you showed efficacy of a drug out, say one or two years, to keep somebody on the alternative treatment, out yet another three years.
Number two, I think that it doesn't really get to the safety issues that you'd really want to get, which are those events which are relatively rare and don't just emerge right during your first pass during the controlled trial.
I'm a neophyte in the pharmaceutical industry, and I've heard of this "rule of three". That is, if you have something like a rare event, something that occurs 1 out of 10,000 times, that you really need to have 30,000 people to make it possible even to see. You'll never see those kinds of events within a controlled clinical trial, no matter how long you sustain that trial.
So, I think that out to three years for a planned safety event within the trial is appropriate. And then, I certainly agree with the sentiment of the community here, which has expressed its desire to see longer-term assessment of safety, and long-term meaning at least five or more years. Thank you.
CHAIRMAN BRAUNSTEIN: Thank you. Okay, we'll try to tackle the last question now.
DR. TEMPLE: Can I just make one comment before?
CHAIRMAN BRAUNSTEIN: Yes, sure.
DR. TEMPLE: I want to join Eric Abadie's skepticism about registries.
If you're looking for very rare events, too rare to show up in your trials, and if they're conspicuous, pancreatitis or something, the spontaneous reporting system is rather good at that. If what you're looking for is an increased risk of something that is otherwise existing -- heart attacks, strokes, things like that -- finding those in registries is not very likely.
After all, why did we think estrogens reduced the risks of heart attacks? Because data from registries -- Framingham and things like that -- said it did. So, anything but randomized trials is not very good at these small subtle and yet potentially important risks, and the sort of obvious stuff usually comes out.
Maybe there are exceptions to that, but it's very hard to have a registry that's big enough to do what you really want it to and that finds the subtle things, because there's no really good control group.
You're in epidemiology, and small risks are not easily detected, epidemiologically. I mean, what did the WHI show? It showed very small increases, but they were considered very important. So if one really wants these things, I think long-term active controlled trials are extraordinarily difficult but are at least possible. Registries doesn't seem like the very likely way out to me.
CHAIRMAN BRAUNSTEIN: Well, the other issue also is that the incentive for doing a long-term active controlled trial goes away once something is out of the strong regulatory environment. So if it's going to be done in the long-term, it's probably set up best at the beginning, before it's marketed.
DR. TEMPLE: I guess I wanted to make one exception. If there's some particular thing you're worried about, a tumor of some kind or something like that, that may be suitable for registries. Where there's something very focused, that can be done.
DR. GRADY: Could I make one more point?
CHAIRMAN BRAUNSTEIN: Yes.
DR. GRADY: I think the Tamoxifen example is a really good one. I think right now, not only is there no incentive for companies to continue some sort of structured evaluation after getting their drug registered -- because if they do, they stand to prove that it's only good for five years. That's what happened with Tamoxifen.
Tamoxifen probably would've been used lifelong, had the NCI not done a study in which they randomized women to stop or continue after five years. So it was that study that got the drug limited to five years. So I think the whole idea that we might approve drugs for the duration of use that they've been proven to be beneficial in randomized trials could be an incentive for extending the length of those trials. Right now the incentive is to do a short trial and don't do anything after that.
CHAIRMAN BRAUNSTEIN: Thank you. Okay, for the final question -- this will concern the use of placebo versus active control, when is placebo appropriate, when is active control appropriate, both for efficacy and safety. And if you could define the populations, if you think that there's a difference. That is, women or individuals whose T-scores are less than two and a half standard deviations below the young adult mean, without fractures, those with one fracture, those with multiple fractures, new fractures.
DR. MCCLUNG: Well, with that opening, let me say that we actually know how to stratify patients in the gradients of risks based on bone density and other risk factors. Using strictly a BMD T-score cutoff actually is too naive.
We could factor in age and the presence or absence of fractures, and there are ways to assess absolute fracture risk. Perhaps rather than choosing T-scores, we can choose levels of absolute risk to categorize patients into a very low-, medium-, or high-risk.
For very low-risk patients, patients with normal bone density values, there's no ethical concern about the safety of a placebo-controlled trial, but there -- in my view -- is a real question about why one would do it. Because it's less clear to me as time goes on about the need for pharmacologic intervention in very low-risk patients.
For intermediate risk patients, we've talked about bone density values and age and absence of vertebral fractures that put them in a modest risk category. I am perfectly comfortable with those being patients involved in placebo-controlled trials and think that that's the right group in which to do the studies.
The group again which I think we've all agreed shouldn't be involved in placebo-controlled trials are the patients whose individual risk over the short run of the study is so high that there is ethical concern, patients with recent or multiple vertebral fractures.
The active controls studies, I think, would fall into two categories. One would be to document superiority of one drug over another. I'm not a big fan of equivalence or non-inferiority studies, particularly with bone density as an outcome. Because while bone density is a very strong predictor of fracture risk among untreated patients, the relationship between changes in bone density in response to treatment and subsequent -- and its relationship to fracture risk is less strong.
So, demonstrating superiority would be a reason to do it. Or more practically and more interestingly, to look at the effect of combined therapies when the mechanisms of actions of the two drugs are clearly different. I think there isn't justification for combining sets of antiresorptive agents, but there is strong interest and rationale for combining anabolic and antiresorptive agents.
And they are high-risk patients, those with vertebral fractures could be the subjects in the study, both of them receiving our current best regimen, and then one group receiving the additional drug, and the other to compare with our standard best therapy. Thank you.
CHAIRMAN BRAUNSTEIN: Thank you. Dr. Watts?
DR. WATTS: I don't think it's ethical to do placebo-controlled trials in osteoporosis. I want to be clear that when we say "placebo", and Henry has mentioned it several times, we really mean everybody gets calcium and vitamin D, and one group gets an agent that masks the fact that they're not getting the active agent.
I think those trials are perfectly appropriate for patients who are at lower risk of fracture, that is with low bone density -- and I don't know whether it's B2 or B2.5 -- and age gets factored in there. I think it's appropriate for patients who have one vertebral deformity of indeterminate age. I think it's not appropriate for patients with clinical fractures or multiple fractures.
I think the active control trials are best suited, as Dr. McClung says, for superiority or a unique situation of add-on. I think that equivalence or non-inferiority trials are almost worthless, either with a BMD endpoint or fracture endpoint. I think there's a danger in looking at BMD as a good surrogate, particularly when it comes to these novel agents or combinations of antiresorptive and anabolic agents, because I can give you scenarios in which a combination might produce less of a gain in BMD than one of the agents that has the most effect on BMD. Yet, the combination might have a better effect on bone strength and reduce the risk of fracture.
I'm not sure if the agency is ever going to get into that issue, but it's already a thorny one in the bone field and we don't have the anabolic agents on the market yet.
CHAIRMAN BRAUNSTEIN: Dr. Bone?
DR. BONE: Thank you. I concur with my colleagues that in patients where the background, in trials where the background therapy is calcium and vitamin D and we're comparing an additional agent versus a masking placebo tablet or other injection of whatever, that that kind of trial can appropriately be carried out in the patient who meets the criteria for osteoporosis diagnosis but does not have higher risk characteristics.
This would be the B2 or the B2.5 standard deviation patient. It would be the patient who has no more than one remote fracture. That patient might take into account other risk factors, as Dr. McClung has added, that might bump that patient out of the trial. But certainly, recent or multiple fractures would put the patient out of the placebo-controlled trial category.
I think that we have to think about taking the higher risk patient into an active comparator trial at an early stage of development. If we don't have established efficacy for the test agent, I'd rather get that in low-risk patients than go into high-risk patients with an unproven drug. So, I would reluctant to consider an active comparator trial of the non-inferiority type until I had some evidence of antifracture efficacy with the novel drug in a low-risk background situation.
I think we have to be a little careful when we talk about add-on trials. It sounds very attractive. But we've never done one, except for the calcium and vitamin D background therapy in which we've demonstrated additional efficacy, that I can think of.
And I think we should be pretty careful about making regulatory policies out of something that we've only imagined and never tried. There was some evidence presented based on bone density data at the recent meetings of an add-on trial with an anabolic agent and a antiresorptive agent. It looks to me like you might well have missed the effect, based on the evidence that was presented at the meeting. I may have wrongly concluded -- drawn a wrong inference, if that were the only evidence.
So, I think we should be very cautious about depending on that as the primary test of efficacy. I think it's certainly a reasonable thing to find out more about drugs with, and to determine whether adding the two drugs together would effect -- The latest combination therapy is actually better than mono-therapy with one or the other. But I don't think we can rely upon that as our primary evidence of antifracture efficacy.
If you did see antifracture efficacy in that situation, it might help you to conclude that you had an antifracture effect, but then you would still be plagued in many cases by the question of whether it was specific to that combination, and could be generalized to mono-therapy. So, I think there are some problems with that. There might not be fatal problems in every case, but we shouldn't regard that as so easy as it might have first sound.
So I think our best and really arguably the most ethical approach is to get our primary evidence of antifracture efficacy in the comparatively low-risk osteoporotic patient in a placebo-controlled trial with background therapy of calcium and vitamin D.
CHAIRMAN BRAUNSTEIN: Dr. Zerbe?
DR. ZERBE: Did Dr. Temple want to add something?
DR. TEMPLE: I know you're going around the table, but I have to ask this question, because it has to do with the things before.
I just want to mention what my credentials are for asking this question. For the agency, I have been on the attack on the Declaration of Helsinki, arguing the importance of continuing to do placebo-controlled trials when there is no irreversible risk to the patient. I've even been abused slightly for that. I feel very strongly about it, and I'm well aware --
CHAIRMAN BRAUNSTEIN: But you got some very nice articles in the Annals out of it.
DR. TEMPLE: -- and I'm very well aware of the difficulties with active control trials. However, the ICH E10 document and others uniformly agree that where available therapy -- and one might add available widely-accepted therapy -- produce a death or some irreversible morbidity, you really can't continue to do those trials.
So what I can't figure out is why it's okay to treat people whose irreversible morbidity you are depending on -- otherwise you won't succeed in this trial; there has to be more fractures in the untreated group or the trial doesn't show what you want it to -- why that's okay.
DR. BONE: Because that doesn't constitute irreversible morbidity, Bob.
DR. TEMPLE: Well, not in everybody.
DR. BONE: Not generally. When we're talking about vertebral deformities measured by millimeters in these patients, we are not talking about irreversible morbidity or mortality.
DR. TEMPLE: Okay, I still have one more question.
It seems to me what the standard of care is is very important to this. And I think Dr. Silverstein said this before. If the consensus among experts like you guys is that people with a certain condition ought to be treated to prevent those things, I think it's very difficult to say leaving them untreated is easy.
For what it's worth, in the international arena, I've argued -- but with no support from anybody -- that it's okay to go to a country that can't afford a drug and do trials there against placebo, even though you wouldn't do that in your own country. Believe me, that is not a welcome position. Nobody buys that.
So I still -- that raises the question here. If you all believe that a certain thing is a standard -- now maybe you don't, in which case I understand the position -- how does this add up?
CHAIRMAN BRAUNSTEIN: Dr. Watts?
DR. WATTS: I was sorry that I hadn't mentioned, when it was my turn, that if we have a consensus about placebo-controlled trials two to three years duration in lower to intermediate risk patients, that someone with authority needs to clarify this for IRBs.
And the two points that I would make -- Henry has made one, and I'll make it again -- that the likelihood of a permanent and serious harm is low in the population groups that we talked about, and there are ways to minimize that. I mean, there has to be a difference in fracture number.
But the second point --
DR. TEMPLE: Low but real? Or low?
DR. WATTS: It is measurable. But the second point that I think is not accounted for in the Declaration of Helsinki is the fact that I mentioned earlier. That is, we are not taking these people out of our clinics. We are not taking people off of effective therapy. We are going out and looking harder and harder and harder for people who haven't been tested, haven't been diagnosed, and are very unlikely to receive treatment during the course of these trials.
Now I realize the ethicists would say, well, that's a healthcare delivery problem and you can't use that as an ethical justification for including these people in trials. But I think that's short-sighted. I think if we're not doing these trials, those people are not going to be identified and they're not going to be treated.
So, the likelihood that a huge number of people will suffer harm if we don't do these trials is greater than the likelihood of patients in the trial getting placebo suffering.
DR. TEMPLE: I think it's very important to develop that part of the argument. We had a case where someone wanted to leave a 2B3A inhibitor out of the treatment of someone was undergoing -- who I guess had acute coronary syndrome. And we initially said, you can't do that trial. We have data that shows it prevents heart attacks.
And what they were able to show was that the serious cardiovascular community was worried about the bleeding, was worried about the cost, and were not using it. And we though "okay". But it seems to me, those arguments are very critical here to explain why, when the fractures -- I mean after all, the difference in fractures is the endpoint. They're going to get more fractures, or you lose.
Why is that okay? And also, what do you tell them as they enter? The second part of it. That seems very important. Just saying the risk is low doesn't really make it.
CHAIRMAN BRAUNSTEIN: One could also argue that it's unethical to let a drug go on the market that hasn't been unequivocally proven to be efficacious.
DR. TEMPLE: I totally agree with that, but that's not usual. The desire and need for a study is not usually considered sufficient reason to allow patients to come to harm. You know, these are all delicate and difficult matters. But the dogma is, you're supposed to think about the people in the study, not the benefit to the community primarily.
DR. ZERBE: The individual versus the population.
DR. TEMPLE: That's correct. That's the usual standard. I mean, all these things can be debated.
CHAIRMAN BRAUNSTEIN: Dr. Zerbe?
DR. ZERBE: I yielded my time, I guess. It was worthwhile.
It sounds like there's a consensus around placebo-controlled trials, and the more modest would be the exception. I think the issue is the more severe. And just to underscore the point that -- I think it's going to become more and more difficult, if not impossible, to do those trials unless there is a pretty active and unified argument that says that they should be done as placebo-controlled trials rather than -- if that's the view. And I don't think it is even around this table. So it's effectively ruled out. A placebo-controlled trial in severe cases --
CHAIRMAN BRAUNSTEIN: Severe. I think we're actually keying in on the less severe.
DR. ZERBE: Yes. I understand that, but I'm just stressing the two categories that there is consensus around. You can argue about the ethics for a long time about the severe. And there are population issues with regard to ethics, but the reality is, they are getting to be impractical, if not impossible, simply because of IRBs, unless there's a clear statement and a consensus, which there does not appear to be around this table.
It's a simple fact. You're going to have to do positive comparator trials in the more severe cases, I think.
And the other point I would emphasize is the add-on model is probably only not very complicated to interpret, but as I understand it, the data that are emerging are, that in fact adding on anabolics with non-resorptives, are not at all additive. I think that was mentioned earlier by somebody.
DR. BONE: There are some preliminary data about that. That isn't final by any means.
CHAIRMAN BRAUNSTEIN: Dr. Levitsky?
DR. CUMMINGS: Could I?
CHAIRMAN BRAUNSTEIN: Did you want to comment?
DR. CUMMINGS: I need to tell Bob or respond to Bob in one sense. The gradient of risk at which something becomes acceptable to test is a really slippery slope, and it's hard to define. But it's not so much -- the principle that seems to get lost is that it's not so much when a trial is ethical but when a decision about that trial being worthwhile or ethical switches from the patient who makes that decision to us.
And so, I think that there's a certain small level of risk where it's acceptable from my point of view that I could recommend therapy to the patient, but it's okay if she refuses on the basis of information about her absolute risk and joins a trial.
And I think it's not so much that we, as a community, wouldn't recommend treatment to someone whose bone density is below a B2.5. I would recommend it, but I would accept her refusal, an informed refusal to say to me "that's fine, but I'd rather be in a trial" because I don't think the risk is sufficient for me to overcome her right to make an informed decision. And that's different --
DR. TEMPLE: Then the consent form would tell her right now we've got serious people in this condition.
DR. CUMMINGS: Yes. But see --
DR. TEMPLE: It should be urged that treatment should be used in people like you.
DR. CUMMINGS: Absolutely, Bob. But right now the consent forms, unfortunately they create problems for the IRBs that the FDA could help with.
Now, I'll say that drugs reduce the risk of fracture by 50 percent, and your risk of dying from fractures is 10 to 20 percent, and they do not put things in terms of the absolute risk and benefit for an individual patient who's looking at the trial. Those absolute risks of transient disability are modest to small for the patients we've been talking about. And for permanent irreversible disabilities -- I haven't calculated those, but those are really tiny for the class of patients that we're talking about.
And informed consent needs to be much clearer about absolute risks rather than the relative risks that we have used to promote the importance of the disease.
DR. TEMPLE: That's obviously clear, but it also needs to be clear on what the standard recommendation is, that there are therapies that will ultimately work in this, and other things that I imagine would have an effect on --
DR. CUMMINGS: Yes, they need to see their physician and hear that too. But I think the FDA could help out with the problem of IRBs who really only understand the relative risks -- and I think if the FDA said that this is an acceptable class of patients in which to design trials, I mean that would help clarify things a lot for people who are confused about absolute and relative risk.
DR. TEMPLE: You probably don't want us to make the uniform determination on it. You probably want to leave it local. I'm just guessing, but --
CHAIRMAN BRAUNSTEIN: Dr. Levitsky?
DR. LEVITSKY: I will ask a question, rather than stating an opinion, for those of you who know more about the pharmacology of these compounds. In these very low-risk people who would be entered into a placebo trial, is there a way to do a time-to- fracture study so that you decrease the exposure of these patients, rather than --?
DR. MARCUS: The answer is "yes". There are many designs one could use. You could say that statistically you'd need to get a total of 75 fracture, let's say, in order to have the power to distinguish one group from the other, and then you'd just continue your trial until that 75th fracture is achieved. And then "wham", you'd cut it off. There are other models too, but you're quite right. You don't have to stick to a three-year model.
DR. LEVITSKY: That would diminish the risk for any individual person --
DR. MARCUS: Of course.
DR. LEVITSKY: -- if that was your approach.
CHAIRMAN BRAUNSTEIN: Dr. Temple?
DR. TEMPLE: I'm sorry I didn't think of this before. We have urged in settings where controlled trials are difficult -- such as seizure studies and recurrence of atrial fibrillation -- just what you suggested, because in a certain sense everybody gets one event, and not more than one event. That's a little -- it's truer if it's something like seizures where everybody is going to have one in the first month.
But, it does reduce the burden on the people who aren't treated. They at least don't get multiple fractures or stay on it a very long time and accumulate the risk. So that does seem worth thinking about, time to first event.
CHAIRMAN BRAUNSTEIN: Okay. Dr. Sampson?
DR. SAMPSON: I'm not prepared to address in terms of patient populations that would be suitable for placebo, but I was thinking back again to Dr. Cummings' excellent presentation this morning in terms of fracture incidents and non-inferiority.
And he and his colleagues used a certain -- left a sample size impression, a very large sample size. And they used a delta, to me about 25 to 33 percent of the difference between placebo and active. And they also assumed that the test compound was equal in efficacy to the active -- and I'm just talking about efficacy. And those are rather stringent assumptions, I think, in some ways.
Dr. Temple alluded to the fact that larger deltas might be acceptable -- in the range of 50 percent, maybe even larger. And if you were to use one of the less efficacious comparators and assume that you were more efficacious in that comparator in powering the study, that would in effect further reduce the sample size.
So at least when people have the choice of using an active comparator, I think there might be less severe sample size considerations than maybe the presentation left this morning.
CHAIRMAN BRAUNSTEIN: Dr. Lukert?
DR. LUKERT: I think I already stated what my parameters would be for placebo-controlled trials, a person without a recent fracture. And with the safety net, which sort of addresses what you say, they wouldn't be allowed to have more than one fracture. Even if their bone density starting falling more than your predetermined amount, they would be removed from the study.
As far as the active controls, I think the only place I would consider it practical would be combinations. I think eventually those are going to have to be studied, the combinations of anabolic and antiresorptive agents.
CHAIRMAN BRAUNSTEIN: Dr. Aoki?
DR. AOKI: I basically concur with Dr. Watts and Dr. Bone's opinions regarding the placebo and the comparator studies.
CHAIRMAN BRAUNSTEIN: Yes, so do I. That low-risk groups -- all groups should get vitamin D and calcium and exercise. In low-risk groups, I see no major ethical problems with carrying out a placebo trial to either a fracture endpoint or BMD, depending on the class of drugs.
As far as the individuals who are at high-risk with multiple fractures or recent fractures, I think an active control study is reasonable and that a placebo-controlled study in that setting is not.
Dr. Gelato?
DR. GELATO: I agree with pretty much everything that's been said, except that I think in the high-risk group they should not be given a therapy until it has been proven to be efficacious. I agree with Dr. Bone, Watson, and McClung. I think that until you know it works, they're not a group that you should really take a chance on, because the risk is too great for them. There's not a good risk-benefit ratio there.
CHAIRMAN BRAUNSTEIN: Dr. Grady?
DR. GRADY: I find this a very difficult issue. I mean, I think you can pick out specific language from the Declaration of Helsinki, but the sort of intent of it is generally to say that the risk to participants in a trial should definitely not outweigh the potential benefit of what we're going to learn scientifically, or the public health benefit.
And I think in this case that that's how we justify it. I think that there is some risk for people in the placebo group, but that it's small. And hopefully we stand to learn something that outweighs that small risk.
That said, I still think we should think harder about non-inferiority trials. I mean, I think what continuing to do placebo trials leads to -- or is going to lead to -- is eight or ten bisphosphonates on the market, with clinicians not really having a good idea of which one is better than any of the others. That really doesn't do a service to science or society either.
And I think the main reason we don't do equivalence trials is because they have such practical problems, and it is also problematic to interpret them. So if you don't really learn anything from them then you haven't met the requirements of the Declaration either.
CHAIRMAN BRAUNSTEIN: Dr. Abadie?
DR. ABADIE: I think placebo may be ethical in patients who are usually not treated, where the drugs are not seen as widely available medications. And that's the case in Europe, at least in certain countries, as in my country, in patients with low risk, that is, without any pre-evident fractures at the beginning.
And I would probably strongly echo the time-to-event statistical analyses. In fact, we have already thought about that for the multiple sclerosis, where in fact we have exactly the same problem of active control trials. Before, we'd think that the placebo in these particular populations may be also acceptable with the time-to-event approach.
With respect to the active control trial, certainly not pivotal unless it's a superiority trial -- because we would like to see superiority on fracture either versus active control but it will be probably difficult -- or versus placebo.
For the rest, I would say, such as a new dose, new formulation, I would probably go along with the active control trial in a non-inferiority setting with BMD as a first endpoint, but only -- only I would say -- in case of new dose, new formulation, new pharmaceutical formulation.
CHAIRMAN BRAUNSTEIN: Dr. Silverstein?
DR. SILVERSTEIN: I agree with everything you said. As a clinician, one of the difficult things -- as Dr. Grady said -- is, why should I choose one drug over the other? And in the absence of active control trials, we really don't know. And so, I think that there is a role for them, in this particular instance and in many others as well.
CHAIRMAN BRAUNSTEIN: Dr. Rodan?
DR. RODAN: I agree with the Chair regarding low-risk patients. The proviso that patients who fracture -- for example, it was shown in clinical trial that Alendronate reduced by 100 percent the occurrence of more than two vertebral fractures. So there is a way to prevent really significant deleterious outcomes in patients involved in the trials, based on existing therapy.
CHAIRMAN BRAUNSTEIN: Okay. Dr. Rizzoli?
DR. RIZZOLI: Yes. Calcium and vitamin D we all agree -- it's more than a placebo, particularly if it's a full dose given, which is not always the case in several trials, in which was given just the minimal dose.
Second, the risk should be defined on the absolute risk base, not only on BMD but other risk factors in the presence of multiple fractures. And finally, I'm not sure that the high-risk patient should not enter a placebo, calcium, vitamin D controlled trial if, for instance, he or she is in a class of age in which there is no well-established treatment, or if we ever design a little bit less stringent, like a time-to-fracture or shortened study, with the possibility after one year to be switched to the active drug.
CHAIRMAN BRAUNSTEIN: Dr. Hochberg?
DR. HOCHBERG: I guess I agree with a lot of what's been said before, both by the Chair and by my colleagues to the right and by my colleagues across the wide gap between the table.
I comment on a little bit of this question about add-on studies.
One of such designs is the issue of taking so-called partial responders, or non-responders, to therapy and then randomizing them to either continue on their therapy if they're a partial responder, or add on a therapy. We know that patients who get active drug in this situation have fractures. They just have fractures at a lower rate than the group that receives the placebo -- everybody getting calcium and vitamin D.
So, one question would be: Could you design an add-on trial where you would continue patients on active drug and then add on something, as compared to adding on placebo in that situation? That obviously might be a possibility for this so-called high-risk population.
I think the other comments people have made are all very reasonable. I don't want to trivialize things. I think often times we consider when we enroll patients in trials whether the so-called "mother test" -- this is relevant to osteoporosis -- as to whether you would enroll your mother in the trial. I think, given the low-risk in the overall group for serious outcomes and looking at absolute risk as suggested by Dr. McClung at the beginning of this discussion -- is a reasonable way of making those decisions.
CHAIRMAN BRAUNSTEIN: Dr. Cummings?
DR. CUMMINGS: Nelson Watts said it beautifully. I couldn't add anything to what he said. So Nelson, if you wouldn't mind restating it. That would be my comment.
(Laughter.) DR. CUMMINGS: I also agree with everything that Bob Marcus is about to say.
(Laughter.) DR. CUMMINGS: And Bob, the most serious comment is actually about the international scope of the trials. I think that Bob's -- I actually support not only the ethicalness but I think the desirability of doing trials in places where people are not getting adequate access to -- it's a small -- we should talk more about that. In large part, because the trials that you have approved, or you've let go -- not right now -- for registration at the FDA, are almost all being recruited outside of the United States.
And so we may believe that we shouldn't include patients who have multiple fractures or severe or recent fractures in trials like this, and may decide that for the United States. But in fact, 90 to 95 of the patients that are being recruited right now to fracture prevention trials are being recruited outside the U.S., many of them outside of Europe, mostly in places where there isn't any alternative for treatment of osteoporosis.
I don't know how you consider that when you are considering the design of these trials. But regardless of what we're saying about what's applicable in the United States, the numbers of fractures -- most of the fractures in these trials are going to come from areas where there isn't adequate medical care, and they're being assigned to a placebo or the active drug.
DR. TEMPLE: I basically think that's okay.
DR. CUMMINGS: So do I, but --
DR. TEMPLE: But CEOMs, the National Bioethics Advisory Committee, and damn near everybody else does not.
Just so you know, there's a great international debate about such things. My own view is that if it's not available, you're doing good for all the people in the trial, because at least some of them are getting the good stuff. But I can tell you, that's highly controversial.
DR. CUMMINGS: I'm just wondering -- it needs to be thought through clearly. Whatever -- if there's a revision of the Guidelines here and you put in some suggestions about or limits as to who should be in the trials, how does that influence or affect the design of trials that are then done in a unified protocol around the world?
DR. TEMPLE: Other countries determine who can be in trials. We don't tell them what to do.
DR. CUMMINGS: You don't tell them, but there is one unified protocol that is usually based on discussions with the FDA, not with -- It's a counter-party issue.
DR. TEMPLE: As a general matter subject to debate, if another country decides that something is ethical, they're a country. It means you get to decide if you're a country. It's one of things countries can do.
There's controversy about that, too. But, I think that has generally been our position, unless there is something really just obviously awful.
CHAIRMAN BRAUNSTEIN: We'll let Dr. Watts make a brief comment, and then we'll go on to Dr. Marcus.
DR. WATTS: Very brief. I think the ethical objection to doing a trial in a country where these drugs are not available is that the population of that country does not stand to benefit from the results of that trial. And therefore, it's not ethical to take the trial there.
CHAIRMAN BRAUNSTEIN: Dr. Marcus?
DR. MARCUS: I have often looked up to the opinions stated by my three gray eminent friends over there ever since I was young and in training.
(Laughter.) DR. MARCUS: And I'm prepared to be persuaded to some degree by them today -- as well as by Dr. Lukert, who isn't so gray -- that there may be a problem with people who have multiple fractures and recent fractures.
The interesting thing is -- and this is a new twist that I haven't heard before or seen before in print -- the concept of new fracture. The Helsinki Declaration, the European -- I've forgotten what the acronym is, but the European position paper on placebo-controlled trials and osteoporosis -- just talk about prevalent fractures and they, to my knowledge, they don't address the recency.
But I'm persuaded by you, Nelson, and Henry, and Michael, and Barbara, that the recency may actually provoke an additional risk which is not justifiably undertaken.
That being said, I want to make a strong plea for placebo-controlled trials. And if we have an education problem among our colleagues and IRBs and in universities, well so be it. We need to educate them.
I think it is unconscionable to subject a person to the risks of a trial where you are not going to get the accurate estimate of efficacy or safety. End of story.
I think the Amoxicillin story with Otitis Media, where people got into recognizing the efficacy of that drug for so much for so long that that became the standard comparator. Everything fell apart when a new placebo-controlled trial showed that efficacy wasn't there. That's how it was presented to me in the New York Times.
Bob Temple may want to correct that, but the concept is still, I think, an important one when you don't really know the efficacy or the safety of a drug in absolute terms. You're exposing people to an unmerited risk, and so I think there is still a strong role for placebo-controlled trials.
CHAIRMAN BRAUNSTEIN: Yes, Dr. Bone?
DR. BONE: I'd like to add one comment which -- I think it's obvious, but it probably needs to be stated for the record.
And that is, I think everyone would accept enrolling a patient who was at comparatively high-risk, if that patient either had contraindications or categorical refusal to take any of the drugs that had established any fracture efficacy -- I just wanted to have that in the transcript.
CHAIRMAN BRAUNSTEIN: Great. Okay --
DR. LEVITSKY: Can I?
CHAIRMAN BRAUNSTEIN: Yes, Dr. Levitsky.
DR. LEVITSKY: I asked you a question before, and the reason I asked it is because I've been sitting here applying the "mother test" while you all talked.
If a placebo-controlled trial can be conducted in women where the outcome will be measured only radiologically, and that is by small micro fractures at the vertebrae that will not involve any disability. That is probably a time-to-fracture trial. I think it is acceptable.
If people who enter this trial have three days of disability because of their fracture -- when there is a drug out there which is perfectly appropriate this disorder and which they should be receiving treatment with -- I think any argument about availability in the community, any argument about what is available in the country, is not an ethical argument.
CHAIRMAN BRAUNSTEIN: Yes, Dr. Watts?
DR. WATTS: I want to clarify that. I think what Dr. Cummings said about disability was that the average disability for the patient who fractured was measured in days.
And so, if we're looking at a relatively lower risk population and the fracture rate is two percent in the treated group and one percent in the placebo group, then the difference is one percent of people in that trial might have a day or two of disability, not that everyone in the trial would have days of disability.
DR. CUMMINGS: Can I clarify that? There are some people who have the average fracture that occurs, carries with it in our data so far of approximately 100 days of disability. This is three months.
The reason it's an average -- and this goes back to what David pointed out -- is that when you average things like this, you're really not taking account of the fact that for an occasional patient there is a more prolonged -- and the average is about three months of disability. It's because that happens infrequently -- to one to two percent of the people per year -- that you end up with that average of three. It's not spread out over everybody.
And it's really interesting that, in most of medicine, we have bent over backwards to allow our patients to make informed decisions about treatments -- even if we believe they are beneficial to them. It's in the world of clinical trials where that trend seems to be going just in the opposite way, where we are taking on the decision-making about what's acceptable to our patients.
And I think that the right threshold for the discussion is: At what point do we continue to allow patients to make informed decisions about this? And I that there is a degree of modest disability where that is allowable, where I wouldn't allow my mother -- I would allow my mother into the trial, but would I allow her to make her own decision? Damn well right, because she makes her own decisions. She doesn't ever call me.
So, where is that point where you allow informed decision-making to be made on the part of the patients? And I'm not sure about that. But I think that that's the issue, not where we think personally that line is drawn.
CHAIRMAN BRAUNSTEIN: Dr. Silverstein?
DR. SILVERSTEIN: I think Dr. Bone made a very reasonable suggestion, which is that drugs of unknown efficacy should first be tried in a low-risk population, in placebo-controlled trials to demonstrate efficacy. And then for the high-risk people, be in comparative studies, superiority.
You didn't say that, but that's my --
DR. BONE: Yes. My additional comment to that was, I thought that it should come after some initial evidence of efficacy in a lower-risk patient population.
DR. SILVERSTEIN: Right, then the higher risk.
DR. BONE: Bear in mind here, as Dr. Abadie pointed out, in places where there are -- let's say, social consensus -- about what medications are provided by social health schemes in, say in Europe and actually in Canada, the patient with a low bone density and no fracture is not considered a patient for whom treatment will be provided -- if one of these active agents would be provided, beyond calcium and vitamin D.
So there's -- we're not talking about something where there's a compelling consensus that active pharmacological intervention beyond that kind of support is compelling.
CHAIRMAN BRAUNSTEIN: Dr. Levitsky?
DR. LEVITSKY: Well, I think that the issue of social consensus is a real one. But the social consensus in this country is probably somewhat different right now. And if we can define an appropriate social consensus that would allow us to study a low-risk group that would not otherwise be treated, that's one thing.
But if we have a relatively low-risk group for whom the medical consensus presently is that they should be receiving Alendronate once a week, or something like that, then it's very hard to see not offering them that, unless they are among that group who cannot take this drug or who are very much aware that that drug is available and do not wish to take it and wish to join that trial. That is a very difficult road to walk when you're consenting patients.
CHAIRMAN BRAUNSTEIN: Dr. Zerbe?
DR. ZERBE: I just want clarify one practical issue related to informed consent.
I think it's fair to say that companies would be very reluctant to in any way to put together or approve informed consent, which viewed in retrospect, could be considered as inadequately informing. So I think there's a lot of care in presenting for protection, if nothing else, a very conservative description of the risks that are associated with that.
CHAIRMAN BRAUNSTEIN: Dr. Temple?
DR. TEMPLE: I think some of the discussion of what the consensus is is very important to this. I must say, I think that's the argument that works.
The distinction between Amoxicillin and this is perfectly clear. There was no evidence in fact that Amoxicillin was useful for a middle ear infection. So, it's obviously perfectly reasonable to use placebo-controlled trials. Also, nothing much happens to you if you don't get treated right away.
In this case, everybody believes that at least certain drugs, not the one that's in the study or the placebo, actually work. That's a different situation. You're denying someone therapy that you all believe works.
But, if that's not really the ordinary practice -- there have been ethical discussions about this by Benjamin Freedman and others, who was not a fan of placebos, that said if there's real data in the community or uncertainty about whether it's worth even on economic grounds, then that's a legitimate area to continue to carry out studies.
I just think it's very important to make those parts of the arguments, because I'm worried about this.
CHAIRMAN BRAUNSTEIN: Dr. Lukert?
DR. LUKERT: Well, I just want to second what Dr. Cummings was saying about taking about away patient autonomy. I've been astounded at the discussions that we've had in several different venues on this subject -- of how the patient's decision to enter a trial is just totally disregarded. I mean, it's as if they shouldn't make that choice because we know that treatment is better for them.
And, there are a lot of people who will make the decision to enter a trial, even with knowing these are your risks for fracture, the pros and cons of entering this trial. A lot of people are still -- there must be some altruism still present in the world, and I think we should not take away the autonomy of an individual to make those decisions.
CHAIRMAN BRAUNSTEIN: We'll give Dr. Marcus the last word.
DR. MARCUS: First of all, I think that it's not really accurate to say that we're denying treatment. Any patient, who has ever been in a clinical trial that I've been involved with who said that they would like to drop out of the trial and be on drug "X", went with my blessings and with my goodwill and agreed to be part of the follow up studies.
So, we are not denying anybody therapy.
I was a trial investigator for Merck's FIT trial for the NIH PEPI trial, as well as for a Lilly's MORE trial, so I have personally consented well more than 1,000 patients in osteoporosis clinical trials over a span of 12 years. I have been overwhelmed by the number of women who say they want to participate in the trial because they want to do something which will ultimately help their daughters and help other people.
The altruism gene is very strong, and the ethics community in this field seems to disregard it entirely. And I think that's a very bad precedent.
CHAIRMAN BRAUNSTEIN: We'll give Dr. Orloff the chance to make some final comments.
DR. ORLOFF: I want to thank everybody for an incredibly thoughtful morning and afternoon, and a lot of hard work. I think it's quite remarkable that we were so successful in convening this group on relatively short notice.
As I said earlier, Henry Bone deserves a lot of credit for his input, as does Eric Colman, of course, who headed up the FDA side of things. Kathleen Reedy did the organization, and Dr. Braunstein clearly isn't particularly rusty after all these years out of the game.
I also wanted to say that if you move, make sure we have your phone number. But even if we don't, we will find you, because this is not the end of this discussion.
We've got a good group. I'm sure we'll call on you again. It'll take us quite a while to go through the transcripts and decide where to proceed next. But, we've got a good start, so thank you all very, very much.
CHAIRMAN BRAUNSTEIN: Thank you.
(Applause.)
(Whereupon, the above-entitled meeting was concluded at 5:31 p.m.)