ATDEPARTMENT OF HEALTH AND HUMAN SERVICES FOOD AND DRUG ADMINISTRATION CENTER FOR DRUG EVALUATION AND RESEARCH ONCOLOGIC DRUGS ADVISORY COMMITTEE 58TH MEETING Tuesday, September 2, 1998 8:00 a.m. Holiday Inn Bethesda Versailles I, II, III 8120 Wisconsin Avenue Bethesda, Maryland PARTICIPANTS Janice Dutcher, M.D., Chairperson Karen Templeton-Somers, Executive Secretary MEMBERS Kathy S. Albain, M.D. David H. Johnson, M.D. Kim A. Margolin, M.D. Robert Ozols, M.D. Richard L. Schilsky, M.D. Richard Simon, D.Sc. Derek Raghavan, M.D., Ph.D. E. Carolyn Beaman, Consumer Representative Deborah Cassel, Patient Representative Sandra Zook-Fischler, Patient Representative CONSULTANTS George Sledge, M.D. Carole Miller, M.D. Julie M. Vose, M.D. Steven E. Lipschultz, M.D. GUEST EXPERTS Susan Ashley Trevor Powles, M.D. Timothy O'Leary James H. Doroshow, M.D. James L. Weiss, M.D. FDA Rachel Behrman, M.D., M.P.H. Julia Goldstein, M.D. Susan Honig, M.D. Susan Jerian, M.D. Robert Justice, M.D. Kathryn Stein, M.D. Patricia Keegan Jay Seigel, M.D. John Johnson, Clinical Team Leader Julie Beitz, Medical Team Leader C O N T E N T S A.M. SESSION NDA Supplement 17-970/S-040 Nolvadex (tamoxifen) Zeneca Pharmaceuticals Call to Order and Introductions, Janice Dutcher, M.D. 5 Conflict of Interest Statement, Karen M. Templeton-Somers 7 Open Public Hearing: Marilyn McGregor 8 Carolyn Aldige 12 Samuel Epstein (letter) 15 Barbara Brenner (letter) 20 Sponsor Presentation: Introduction, Jerry Lewis, M.D. 24 Summary of the Breast Cancer Prevention Trial Results, Joseph Costantino, Dr. P.H., NSABP 27 Summary, Jerry Lewis, M.D. 47 Questions from the Committee 48 FDA Presentation, Susan Honig, M.D. 88 Comments on the Royal Marsden Study, Trevor Powels, M.D. 124 Questions from the Committee 132 Open Public Hearing: Ann E. Fonfa 149 Cynthia Pearson 154 Vincent Li 161 Helen Schiff 161 Mary Ann Napoli 169 Sharon Batt 172 Committee Discussion and Vote 175 (ODAC Discussants: Kathy Albain, M.D. and George Sledge, M.D.) P.M. SESSION Biologics License Application (BLA) 98-0369 Herceptin (trastuzumab) Genentech, Inc. Call to Order and Introductions, Janice Dutcher, M.D. 232 C O N T E N T S (Continued) Conflict of Interest Statement, Karen Templeton-Somers, Ph.D. 233 Open Public Hearing: Alice Hamele (letter) 235 Rosemary Locke 238 Elaine Doubrava (letter) 239 Marilyn McGregor 241 Philip Wyatt (letter) 244 Introduction of the Issues, Julia Goldstein, M.D. 247 Sponsor Presentation: Introduction and Regulatory History, Karl G. Trass 254 Scientific Rationale and Clinical Efficacy, Steven Shak, M.D. 257 Clinical Safety, Virginia Paton, Pharm. D. 276 Summary of Benefits and Risks, Steven Shak, M.D. 292 Questions from the Committee 295 FDA Presentation, Susan Jerian, M.D. 324 Questions from the Committee 357 Open Public Hearing: Robert Erwin 370 Committee Discussion and Vote 375 (ODAC Discussants: Janice Dutcher, M.D. and Carole Miller, M.D. P R O C E E D I N G S Call to Order and Introductions DR. DUTCHER: Good morning. In case you are in the wrong room, this is the Oncologic Drugs Advisory Committee. We are going to start a three-day meeting. Two of our committee members were unable to make it here because they live in cities that are served only by Northwest Airlines, Drs. Krook and Santana. They send their regards. As I am sure everyone in the room is aware, we have a tremendous amount of material to cover today, and our goal is to carefully evaluate the data that are presented by both the sponsor and FDA, and that is the goal of both the committee and the audience. We do have a large number of members of the public who requested to speak and participate, which we welcome. We are going to ask that everyone, including members of the committee and members of the audience, to be as succinct with their comments as possible so that we can get through what should be a very interesting and pretty power-packed day full of information. So, we hope that everyone will work together so that we are not all here until midnight. Thank you all for your interest and for your willingness to participate. We will go around the room and introduce the members of the committee. I am Janice Dutcher, from Albert Einstein Cancer Center, in New York. DR. JOHNSON: David Johnson, Vanderbilt University, Nashville. DR. MARGOLIN: Kim Margolin, City of Hope, Duarte, California. DR. ALBAIN: Kathy Albain, Loyola University of Chicago. DR. SIMON: Richard Simon, National Cancer Institute. DR. SCHILSKY: Richard Schilsky, University of Chicago. DR. OZOLS: Bob Ozols, Fox Chase Cancer Center, Philadelphia. DR. TEMPLETON-SOMERS: Karen Somers, Executive Secretary to the committee, FDA. DR. SLEDGE: George Sledge, Indiana University. DR. RAGHAVAN: Derek Raghavan, University of Southern California. MS. BEAMAN: Carolyn Beaman, consumer rep, Sisters Breast Cancer Network. DR. BEITZ: Julie Beitz, Medical Team Leader, FDA. DR. HONIG: Susan Honig, Medical Reviewer, FDA. DR. JUSTICE: Bob Justice, Acting Director, Division of Oncology Drug Products, FDA. DR. TEMPLETON-SOMERS: We have also two guests for the FDA, Dr. Trevor Powles, if you could stand up for us? Thank you. And, Dr. Susan Ashley, statistician for his group? Thank you. DR. DUTCHER: All right. We are now going to read the conflict of interest statement. Conflict of Interest Statement DR. TEMPLETON-SOMERS: The following announcement addresses the issue of conflict of interest with regard to this meeting and is made a part of the record to preclude even the appearance of such at this meeting. Based on the submitted agenda for the meeting and all financial interests reported by the participants, it has been determined that all interests in firms regulated by the Center for Drug Evaluation and Research which have been reported by the participants present no potential for a conflict of interest at this meeting, with the following exceptions: Dr. James Krook is excluded from participating in today's discussions and vote concerning Nolvadex. In addition, Dr. Robert Ozols has been granted a waiver that permits him to participate in all matters concerning Nolvadex. A copy of this waiver statement may be obtained by submitting a written request to the FDA's Freedom of Information Office, Room 12A-30 at the Parklawn Building. In the event that the discussions involve any other products or firms not already on the agenda for which an FDA participant has a financial interest, the participants are aware of the need to exclude themselves from such 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 involvement with any firm whose products they may wish to comment upon. Thank you. DR. DUTCHER: Thank you. I think you can see from the agenda that the open public hearing has been expanded due to the interest in the agents being discussed today. So, we will start with the open public hearing, which will occur before the presentation, and then following the two presentations we will have additional comments from the public. We will begin with the people who have requested to speak. The first is Marilyn McGregor. We would appreciate it if all speakers would identify themselves as well as any sponsorship, either the sponsor or otherwise. We would appreciate it if those who speak could use the podium if possible. Open Public Hearing MS. MCGREGOR: My name is Marilyn McGregor, and I am Administrative Director of the Cancer Support Community located in San Francisco. I have no financial interest in tamoxifen. There is a great longing to believe that there is a preventative drug for breast cancer. Given the dismal and long-term unchanging mortality rate of breast cancer, there is a willingness to believe that this drug could be the answer to so many people's prayers. But as scientists and as an organization responsible for the public good, I urge this committee not to approve the application of tamoxifen as a breast cancer prevention drug. The reasons I urge this decision are as follows: There is clinical trial medicine and real life medicine and media over-estimation of the benefits of any one cancer drug. There needs to be a higher level of assurance when prescribing a drug that is a known carcinogen in a healthy population or at least no discernible breast cancer. Those of us diagnosed with breast cancer have a different ris/benefit ratio, and tamoxifen may be appropriate. Clinical trials medicine defines exactly who benefits given their family history of breast cancer, as was done in the NCI trial. Real life medicine has a busy doctor in an HMO whose patient may have no family history or risk factors or the pervasive anxiety about developing breast cancer. This woman would most likely be prescribed tamoxifen as a preventative. Off-label prescription is common in cancer and is a benefit to most cancer patients, but this may not be the case with tamoxifen as a preventative in a healthy population. The NCI study does not prove that tamoxifen prevents breast cancer for the life of any one woman. The most that can be said of the NCI trial is that the tamoxifen group appeared to have less breast cancer for the short period of time of the trial, which was an average of 3 years. A woman can develop breast cancer in over a 50-year period. If this drug is approved as a preventative, the insert should say that the drug is only to be prescribed for the length of time of the trial, which was approximately 3 or 4 years. Other speakers will, no doubt, discuss the British studies which reported no benefit of tamoxifen as a preventative. I am going to discuss the Italian study of Dr. Bianco. At the May, 1998 ASCO meeting in LA, there was a symposium on HER2. Dr. Bianco discussed his 20-year study of HER2 overexpression in tamoxifen use. Bianco and his colleagues found that there was no apparent benefit in using tamoxifen for those who overexpress HER2. All other categories showed a benefit of tamoxifen use. In addition, the Italian research also showed that those women who overexpressed HER2 and took tamoxifen had an overall worse outcome. Dr. Bianco stated that his research could greatly affect the use of tamoxifen, and called for further study on this issue. If, indeed, the Italian studies prove to be accurate, this could potentially mean that 25 to 30 percent of women would have no benefit of tamoxifen either as healthy patients or cancer patients. This could potentially mean that before a woman would be prescribed tamoxifen she would have to be tested for HER2 overexpression. Many other possibilities are also possible. However, at this time we do not know the answers to these scientific questions but answers are certainly needed. Good science demands more good science. It is well-known in scientific circles that negative studies or non-U.S. studies are routinely not included in drug analysis. I urge that the Italian and British studies be considered carefully in your application. I urge the NCI to immediately conduct appropriate studies regarding the interaction of HER2 overexpression and tamoxifen use. I recommend these studies be completed before any approval of tamoxifen as a preventative for healthy women. Meta-analyses, retrospective tumor block studies and/or well-controlled trials all need to be done to ascertain if tamoxifen is beneficial to those women who overexpress HER2 in the healthy population and in the cancer population. As tamoxifen is already licensed, doctors may continue to prescribe the drug in individual cases, but the FDA and the NCI need to protect the public. Thank you for consideration of my remarks. DR. DUTCHER: Thank you. The next speaker is Carolyn Aldige. MS. ALDIGE: Good morning. I am Carolyn Aldige, President and Founder of the Cancer Research Foundation of America. Additionally, I have the privilege of currently serving a 2-year term as President of the National Coalition for Cancer Research. The mission of CRFA, cancer prevention through research and education, is fueled by the knowledge that as much as 70 percent of all cancer is preventable. We believe that prevention provides our greatest hope for reducing cancer's deadly impact. We also believe that our organization's focus on prevention is unique among cancer-related organizations. Since 1985 CRFA has directed more than $30 million to promising research, education and early detection programs that turn the promise of cancer prevention into reality. Before making my formal comments, I should note for the record that CRFA receives support from a number of pharmaceutical companies, including Zeneca Pharmaceuticals, as well as a host of other corporate supporters and individual donors. You know the challenge. Breast cancer is the most common cancer among women, accounting for one out of every three women's cancer diagnoses in the United States. Last year approximately 180,000 new cases of breast cancer were diagnosed, and nearly 45,000 women died from the disease. Only lung cancer causes more cancer deaths in women. In the face of such discouraging news, the prospect of the first effective chemopreventive agent for women at risk for breast cancer, tamoxifen, is heartening indeed. I would like to thank the Oncologic Drugs Advisory Committee for allowing me to speak today for this is, in fact, the first time I have requested permission to address an ODAC panel. Why? Because this is the first time, to my knowledge, the committee has considered approving a drug to prevent cancer. Heretofore, consideration was given only to drugs that could be used for treatment. In our view, this is a landmark event. CRFA has long supported the National Cancer's Institute decision to conduct ground-breaking cancer prevention trials. In fact, in 1993 I was pleased to have the opportunity to testify before the Senate Cancer Coalition in favor of continuing the breast cancer prevention trial. We believe that the compelling results of NSABP's P-1 study merit approval of Zeneca's application for the use of Nolvadex as the first preventive agent for women at risk for breast cancer. The trial results are, as Dr. Richard Klausner has noted, nothing less than a real advance for women with a family history of breast cancer. Women in the trial taking tamoxifen developed 45 percent fewer cases of breast cancer than those on placebo. There were 85 new cases in the tamoxifen group over 4 years compared with 154 in those on placebo. Women on tamoxifen also had fewer cases of DCIS, as well as fewer bone fractures of the hip, wrist and spine. We also note that the drug has its drawbacks -- more cases of endometrial cancer, pulmonary embolism and deep vein thrombosis. The risk for endometrial cancer in the tamoxifen groups was more than that of the placebo group, while the risk of pulmonary embolism was nearly tripled. However, these potentially dangerous side effects appear to be limited to women older than 50, and we believe these risks can be managed. While no one can underestimate the seriousness of these potentially life-threatening side effects, the case for ODAC approval remains a strong one. Approval will ensure that doctors and other health care professionals are fully aware of the drug's side effects, and can discuss them fully with patients. Approval provides the FDA with the opportunity to capture data about adverse events, rounding out knowledge of the drug. Approval means that patients and doctors will not have to seek the drug off-label prescriptions. The approval of tamoxifen is a crucial early step in the prevention of breast cancer in American women. We, at CRFA, applaud your taking this step which means so much in the long term to women at risk for the disease and their families. Thank you again for the opportunity to speak today. DR. DUTCHER: Thank you. We also have two letters, and Dr. Somers will read the letters. DR. TEMPLETON-SOMERS: The first letter is from Dr. Samuel Epstein, who is a Professor of Environmental and Occupational Medicine at the University of Illinois. Zeneca's Nolvadex NDA for preventing breast cancer in healthy women "at high risk of cancer," including all women over 60 years old, is primarily based on NCI's April 6, 1998 summary report, "Breast Cancer Prevention Trial, BCPT, Shows Major Benefits and Some Risks." This report was unsupported by a peer reviewed scientific publication and was qualified by the admission that "further analyses of the data are under way." No further data have yet been released, nor has the report yet been published. Additional evidence is derived from tamoxifen's partial protective effects in rats and mice against the induction of breast cancer by 7,12-dimethylbenzanthracene, DMBA, besides other carcinogens. However, those DMBA-induced cancers which were not suppressed were hormone independent and highly aggressive. NCI's report announced that the BCPT had been terminated prematurely on March 24 in view of "clear evidence that tamoxifen reduced breast cancer risks." As indicated in the Table -- and for this I will have to refer the committee to the tables in their packets -- based on data cited in the report, tamoxifen reduced the incidence of both invasive and non-invasive breast cancer in women of all ages. However, the short term duration of the trial precludes determination as to whether the drug prevented cancer or merely delayed its onset by treating small undetected tumors. On July 11, 1998, two publications in The Lancet reported no evidence for the efficacy of tamoxifen in preventing breast cancer. A 6-year trial by the Royal Marsden Hospital, London oncologic team, based on some 2500 women with a family history of breast cancer, and a similar 4-year study by the European Institute of Oncology in Milan, based on 5400 women, reported no difference in the incidence of breast cancer in women treated with tamoxifen or placebo. An accompanying editorial warned -- this is a quote -- the failure of these trials to confirm the results of the U.S. study, however, casts doubt on the wisdom of the rush, at least in some places, to prescribe tamoxifen widely for prevention. Longer follow-up of completed and current trials is clearly required to clarify the relative preventive benefits and risks in different populations, and to confirm the BCPT findings. Most importantly, none of these trials provides reliable data on mortality, which should be the ultimate endpoint. These concerns have been summarily dismissed by NCI -- "the chance that our results occurred by chance was 1 in 10,000." However, The Lancet editorial did not challenge the results themselves, but their interpretation and significance. Serious short-term complications in the BCPT, uterine cancer, pulmonary embolism and deep vein thrombosis, were increased 2-3-fold in the tamoxifen group. These complications were only seen in postmenopausal women. Among non-hysterectomized women in this age group, the incidence of these serious complications was 2.2 percent in contrast to a 1 percent reduction in the incidence of breast cancer. It must be recognized that the short term duration of the BCPT, apart from the absence of any long-term follow-up, precludes recognition of possible further increases in the incidence of already recognized short-term life-threatening and other serious complications, and also of other, not yet reported, long-term or delayed complications. Of concern in this connection is the fact that tamoxifen induces ovarian necrosis and ovulation in a manner similar to clomiphene, a recognized risk factor for ovarian cancer. More serious still is the high hepatocarcinogenic potency of tamoxifen in the rat, as confirmed in February 1966 by the International Agency for Research on Cancer, at low doses and blood levels equivalent to those in the BCPT. Tamoxifen also binds tightly to estrogen receptors in the human liver, and induces highly stable DNA adducts in 2 rodent species. Risks of liver cancer are not precluded by the absence of such reported complications among breast cancer patients treated with tamoxifen as relatively few such women have taken the drug for over 5 years and followed up for a further 20 years before which the induction of liver cancer would be unlikely. It should be noted that senior NCI staffer Dr. Leslie Ford dismissed risks of liver cancer on the grounds that no cases were reported in the short term BCPT, and also on the incorrect grounds that carcinogenic effects in rats were only seen at high doses. Ford's logic, however, would exculpate virtually all recognized human carcinogens. Furthermore, NCI's denigration of the human relevance of the experimental carcinogenicity data on tamoxifen and its failure to warn BCPT participants of this grave risk is in striking contrast to its reliance on rodent teratogenicity data as the basis for warning against the administration of tamoxifen to pregnant women. It is of further interest to note that while some 25 cases of liver toxicity in tamoxifen-treated breast cancer patients, acute hepatitis, liver failure and deaths and hepatobiliary complications, have been reported in the U.K. by 1992, with similar evidence obtained from the FDA, no such adverse effects were noted in the short term BCPT. NCI's preliminary April 6 report on the prevention of breast cancer by tamoxifen has still not been finalized and published in a scientific journal. The advisory committee should consider the propriety of Zeneca's NDA submission as it is based, in part, on data which have not been made fully available to the public although the underlying NCI research was funded by the public. Furthermore, the claimed evidence for chemoprevention has been rebutted by two subsequent scientific publications. Of as great concern is the well-documented evidence of short-term life-threatening complications, and also risks of delayed fatal complications, evidence for which has been trivialized and suppressed by NCI. Based on these scientific and ethical considerations, the advisory committee is urged to deny approval of Zeneca's NDA. This and the other letters are available for your viewing at the registration table outside. Our second letter is by Barbara Brenner of Breast Cancer Action. It says: Dear Committee Members, based on the data currently available, both from the NCI and from the recently released European results, Breast Cancer Action opposes approval of the proposed indication for Nolvadex. Women are entitled to expect that any drug approved for the prevention of breast cancer will both actually prevent the disease and carry benefits that outweigh the risks of taking it. As far as we know now, neither is true for Nolvadex. We urge you to "just say no" to this application. Breast Cancer Action is an education and advocacy organization founded and led by women living with breast cancer, representing over 4000 members throughout the United States and beyond, we carry the voices of people affected by breast cancer to inspire and compel the changes necessary to end the breast cancer epidemic. Since our founding in 1990, we have been calling for research into true breast cancer prevention, as well as research on effective treatments. The history of the breast cancer prevention trial that led to the application that is now before the committee is well known. Breast Cancer Action long ago summarized the trials as "bad research, bad drug, bad news for women." But it is not the history of the trial that concerns us today; it is the current state of information about the drug's preventive effects and the risks associated with its use. The data currently available regarding the use of tamoxifen in healthy women point to far too many known and unknown risks to justify the approval of Nolvadex as a preventive. The risks, as revealed by the BCPT-1 data, are presumably well known to the FDA and the committee. But seeing them listed gives us and, hopefully, the committee members the overwhelming sense that this application is premature in the extreme. From studies of tamoxifen in women with breast cancer and from the BCPT-1 trial, some of the side effects of taking tamoxifen are known: Endometrial cancer, pulmonary embolism, deep vein thrombosis, eye damage, depression, irritability, vaginal dryness, hot flashes, memory loss and weight gain. Because BCPT-1 ended before the 5 years for which the trial was designed, because a number of the participants were involved in the trial for far less than 4 years, and because there is no rigorous follow-up guaranteed for the trial participants, there is much we do not know about the consequences of tamoxifen for healthy women at high risk for breast cancer. Given the recruitment of BCPT-1 participants into the STAR trial, BCPT-2, even the minimal follow-up planned for BCPT-1 participants will be of little or no value in resolving the many unknowns about tamoxifen. Among the most troubling unknowns are these: Long-term effects of the drug in terms of breast cancer risk or any other risk; appropriate duration of treatment; how long the protective effect of the drug lasts; whether and how benefits and risks vary depending on the race/ethnicity of the woman taking the drug; whether and how benefits and risks vary depending on age of the woman taking the drug; whether and how benefits and risks vary depending on breast cancer risk factors; and whether women who develop breast cancer while on tamoxifen develop a more aggressive form of the disease. While it will be argued that some of the foregoing information is known, we disagree. Either because of the trial design or because data about the trial has not been made available before now, we simply do not have the answers to these questions. All of these concerns are addressed at length in the lead article in the June/July, 1998 edition of the "Breast Cancer Action Newsletter," a copy of which is attached to this testimony for the committee's convenience. Last but certainly not least, the data that are currently available clearly indicate that, whatever else tamoxifen does for healthy women, it does not prevent breast cancer. The BCPT-1 data show only that for some small group of women the drug may delay the onset of the disease. The NCI's conclusions, even in this regard, are undermined by the recently released European results finding no benefit from tamoxifen in healthy women at high risk. Whatever else is true, if someone taking Nolvadex can develop breast cancer, then the drug is clearly not preventing the disease in any sense that the general public understands. What epidemiologists mean by prevention is not what people who are worried about breast cancer mean when they use or, more importantly, hear the word. When we finally have a drug that we know will reduce a woman's risk of developing breast cancer, with attendant risks of side effects that are both known and acceptable, we will encourage this committee and the FDA to approve it under an indication of "risk reduction," not "prevention." But, as far as we know today, Nolvadex is not that drug. For this committee to approve the indication that Zeneca is now requesting would expose millions of healthy women to the known risks of tamoxifen and to potentially grave unknown risks without any guarantee of obtaining the benefits that are being claimed. Only one word can accurately describe such an action -- unconscionable. Do not let Zeneca's drive for profit divert you from the interests of women at high risk for breast cancer. Respectfully submitted by Barbara A. Brenner, Executive Director. As a matter of policy, in order to avoid the fact or appearance of a conflict of interest, Breast Cancer Action does not accept funding from Zeneca or from any other pharmaceutical company. Thank you and, again, both letters are available for you to look at, at the registration desk. DR. DUTCHER: Since we do have time later in the morning for other comments and people are scheduled to speak, we are going to proceed with the agenda as it is printed and we will begin with the sponsor's presentation. Sponsor Presentation Introduction DR. LEWIS: Thank you, Dr. Dutcher. Good morning. I am Jerry Lewis, Senior Medical Director of Zeneca Pharmaceuticals, responsible for Nolvadex, tamoxifen citrate. [Slide] I have the distinct pleasure today of representing Zeneca, and along with my colleagues from the NSABP, the National Cancer Institute and the FDA, we will present and discuss with you the results of the precedent-setting breast cancer prevention trial. This is the basis for Zeneca's supplemental NDA for a change in the labeling -- Nolvadex is indicated for the prevention of breast cancer in women at high risk for developing the disease. [Slide] Following my introductory comments, Dr. Jo Costantino, from NSABP, will present the summary of the breast cancer prevention trial results. At the conclusion of Jo's presentation, I will summarize Zeneca's position and then be pleased to take questions from the committee. [Slide] There are a number of experts here with us today to help address your questions. From NSABP, Dr. Norman Wolmark, Principal Investigator and Chairperson of NSABP; Dr. Costantino, and Dr. Larry Wickerham, Director of Operations at NSABP. For the National Cancer Institute, Dr. Leslie Ford, Associate Director, Early Detection and Community Oncology Program; and from Zeneca there are a number of scientists that are available should they be needed. [Slide] Zeneca is very proud that NSABP selected tamoxifen to be evaluated in the breast cancer prevention trial. NSABP has been involved in cancer research for some 40 years, and has been studying tamoxifen for some 20 years. In 1991, the NSABP met with the predecessor ODAC to discuss the breast cancer prevention trial. We have with us today here Dr. Bernie Fisher who participated in those deliberations. The ODAC at that point endorsed the trial after they were convinced that the potential benefits outweighed the known risks. The trial was designed to detect a reduction in breast cancer risk of 33 percent in women at high risk. [Slide] The trial itself far exceeded these expectations. Tamoxifen for 5 years prevented 45 percent of invasive breast cancers in women at high risk, and no unanticipated toxicities occurred in the trial. For a drug with 10 million patient years of exposure, confirmation of the safety data base should not comes as a surprise. Today is a milestone for it represents the first time that the advisory committee is gathered to deliberate and vote on a drug for breast cancer prevention and, indeed, for any drug for prevention of cancer. Reaching this point in the review process as quickly as we have has been accomplished by tremendous cooperation between NSABP, the NCI and the FDA. [Slide] Let me review this time-line for you. The results of the breast cancer prevention trial were made known to investigators and, indeed, the world on April 8 of this year. Some 22 days later Zeneca filed a supplemental new drug application and the FDA granted it an accelerated review. And, here we are today, a mere 5 months later, on September 2, to consider the results of this trial and a label change for Nolvadex. [Slide] It now gives me great pleasure to introduce Dr. Jo Costantino, Associate Director, NSABP, who will present the data from this trial. These data support our new indication. Thank you very much. Jo? Summary of the Breast Cancer Prevention Trial Results DR. COSTANTINO: Thank you, Dr. Lewis. I am pleased to be here this morning to have the opportunity to provide for you the results of the breast cancer prevention trial. [Slide] I would like to begin just by answering the question why tamoxifen? Why did NASBP choose tamoxifen to be the drug to evaluate as a preventive agent for breast cancer? Primarily because of three factors. First of all, the drug has been proven to be beneficial in the treatment of breast cancer in both advanced and early stage disease. It was also shown to lower the risk of contralateral breast cancer among those patients. And, there was preclinical evidence demonstrating that tamoxifen inhibits the growth of tumors, and perhaps does this by interfering with both the promotion and initiation mechanisms. [Slide] The breast cancer prevention trial was a double-blinded, randomized clinical trial in which women were randomized to receive the planned duration of 5 years of tamoxifen or 5 years of placebo, and 13,388 women were actually randomized to the trial. [Slide] The primary objective of the study was to evaluate the effect of tamoxifen on the reduction of the incidence of invasive breast cancer. The study was powered to determine that endpoint. [Slide] Other objectives included the evaluation of the effect of tamoxifen on cardiovascular disease, bone fractures, other cancers, mortality and the risks of some other outcomes which were known to be risk factors associated with tamoxifen that we had learned from the treatment trials. [Slide] The study was designed to maintain the statistical power even if the non-compliance was as high as 10 percent per year. This is an important factor because this is something that we had planned for in advance. We anticipated there might be a large non-compliance and we wanted to make sure that we did not reduce our statistical power if there was such a fact. The analysis was based on an intent-to-treat approach. That indicates that all individuals were included in the treatment arm that they were assigned and that all events were included regardless of whether or not they took the drug. [Slide] Women got into the trial based on eligibility criteria, one of which was being at high risk for breast cancer. High risk was defined in this trial as being at least 60 years of age, being age 35 or older and having a history of lobular carcinoma in situ, or being greater than age 35 and having a 5-year absolute risk of breast cancer that was equivalent to the 60-year old woman, and that absolute risk was defined as 1.66 percent in 5 years. The determination of this breast cancer risk was based on a mathematical model developed by Dr. Mitchell Gail and his associates at the National Cancer Institute. [Slide] The factors that went into that model that helped to determine what the risk of breast cancer was for each of these women included age, first degree relatives with breast cancer, parity and age at first live birth, number of breast biopsies, history of atypical hyperplasia, age at menarche and race. The original Gail model only incorporated this first set of parameters. It did not include a factor for race. But we worked with Dr. Gail and we developed a factor to include race into the program so that we could also calculate predictive risk for non-white women. [Slide] In addition, the original implementation of the Gail model was designed to predict the risk of both invasive and non-invasive breast cancer. In the BCPT we were interested in just predicting the incidence of invasive breast cancer so we made modifications to account for that also. Almost 100,000 women had their breast cancer risk assessments performed. Of those 98,000, approximately 57,500 women were eligible based on that 1.66 percent in 5 years. Now, among those women who were eligible, there were other medical eligibility criteria that had to be met. If a woman desired to be considered for randomization, she went on to be screened and ultimately 13,388 women were randomized. The data that I am going to present to you today is based on the follow-up as of January 31, 1998. This was the data that was actually used by our data monitoring committee when they decided that the trial had met its objectives and that the trial should be disclosed. As of that date, January 31, 1998, follow-up was available for 13,114 women, and the average follow-up time was 44 months. About 73 percent of the women at that time had been followed for more than 3 years. Almost 60 percent had been followed for more than 4 years, and 21 percent had been followed for 5 or more years. [Slide] I would like to start just by quickly reviewing some of the baseline characteristics related to risk that the population had. [Slide] I will begin with age, and 39 percent of the women were in the age range of 35 to 49 at the time they were randomized; 31 percent were in the age range of 50 to 59; and 30 percent were 60 years of age or older. [Slide] In terms of number of relatives with a history of breast cancer, 57 percent of the population had at least 1 relative with a history of breast cancer; 16 percent had a history of 2 relatives with a history of breast cancer; and 3 percent had a history with 3 or more relatives with breast cancer. [Slide] In terms of the 5-year absolute breast cancer risk predicted from the Gail model, 25 percent of the women had a risk of less than 2 percent in 5 years; 31 percent had a risk in the range of 2-3 percent at 5 years; and 17 percent had a risk of 5 or more in 5 years. [Slide] A significant number of women entered into the trial with a history of LCIS and a history of atypical hyperplasia. Over 800 women in the trial, about 6.2 percent, entered the trial reporting a history of lobular carcinoma in situ and about 9.2 percent, approximately 1200 women entered into the trial with a history of atypical hyperplasia. [Slide] Now I would like to begin with the results, the primary endpoint of invasive breast cancer. [Slide] This plot is a plot of the cumulative incidence of invasive breast cancer that occurred among the participants in the trial. The black line represents the cumulative incidence for the placebo group. The red line represents the cumulative incidence for the tamoxifen group. As you can see, the cumulative incidence for the placebo group was substantially greater than it was in the tamoxifen group. In fact, there were 154 breast cancers which occurred in the placebo group compared to only 85 in the tamoxifen group. This represents a cumulative incidence of 32/1000 compared to 17.9/1000, representing a reduction of about 45 percent in the risk of breast cancer. This difference was highly statistically significant with the p value being less than 0.00001. A couple of things to note in this plot are that the difference appears to show itself very early on, and it does sustain itself throughout the whole 5 years of the plot. [Slide] Similar findings are noted for non-invasive breast cancer. This is the same type of plot only now we are dealing with non-invasive breast cancer. In the placebo group there were 59 events of non-invasive breast cancer compared to 31 in the tamoxifen group. This equates to a cumulative incidence of 12.3/1000 in the placebo compared to 6.8 in the tamoxifen group. This represents a 47 percent reduction in the risk of breast cancer. Again, you can see that the curves separate rather early, before the first year, and they continue to separate through the entire duration. [Slide] This slide reiterates the fact that this finding is consistent across time and has a lasting effect. These are bar charts, and the heights of the bars represent the rate per 1000 of invasive breast cancer by each of the years of follow-up. So you can understand the number of events that went into calculating these rates, at the top of the bars the numbers are given and these represent the number of cases. The yellow bars represent the rate in the placebo group; the red bars, the rate in the tamoxifen group. If you look across all the years, all the way through year 5, you see there is a substantial reduction in the risk of breast cancer all the way up to year 5 and even a 50 percent reduction is evident at year 5. [Slide] To give you a feel for how things look by some of the characteristics of the population, here is the rate of invasive breast cancer broken down by 3 age groups -- less than 49, 50 to 59, and 60-plus. Again, you can see in all 3 age groups that there is a substantial reduction of the rate of invasive cancer in the tamoxifen group. [Slide] Here we show the rates broken down by those who reported a history of lobular carcinoma in situ and those with a history of atypical hyperplasia. Again, there are striking reductions in both of these populations. [Slide] This chart shows the rates comparing treatment groups by categories of predicted risk from the Gail model, less than 2, 2-3, 3-5 and greater than 5. Again comparing each of these categories, you can see that there is a substantial reduction in the tamoxifen group, and this magnitude of reduction, seen here at the upper group, is about the same in terms of relative risk as it is in the lower group. Statistically speaking, there was no significant difference between the reduction observed across any of the categories of risk. [Slide] I would like to take a few minutes now to describe to you some of the tumor characteristics of the cases that were diagnosed in the trial and how they compared by treatment arm. [Slide] The first slide deals with tumor size. What we have here is the rate of invasive cancer by the size of the tumor at the time it was diagnosed, those that are less than 1 cm, 1-2 cm, 2-3 cm, and greater than 3 cm. Again, comparing the bars or comparing placebo to tamoxifen, you can see that there is a reduction in all categories but the bulk of the reduction, the most significant reduction was among tumors that were less than 2 cm in size. [Slide] This graph shows the rates by categories of nodal status, those who were diagnosed with no positive nodes, those who were diagnosed with 1-3, and those who were diagnosed with 4 or more nodes. You will note that there is a really high number of unknowns here, and this is because the majority of these women did not have axillary dissection so the status is in terms of nodes that could not be determined. If you look at the data, again, there is a striking reduction for those who were diagnosed with no nodes, and also those who were diagnosed with 1-3, but there is no difference in the rates of cancer for those who were diagnosed with 4 or more nodes. This is important to note at this point -- tamoxifen is reducing the rates of disease associated with 1-3 nodes and no nodes; there is no increase in the number of cases being detected with 4 or more nodes; and there is no increase in the number of cases being detected that are larger tumor size. It appears that tamoxifen is culling out the smaller tumors and the tumors that present with less than 4 nodes. So, the theory that cases that occur on tamoxifen are more aggressive is not being demonstrated by the data. [Slide] The last tumor characteristic is ER status, and this is an important one because there is an interaction between ER status and the effect of tamoxifen. These two bars represent women who were diagnosed with tumors that were ER positive. You see a very striking reduction in the risk of cancer based on those who were ER positive. On the other hand, there was no difference in the rates of women who were diagnosed with tumors that were ER negative. So, the effect of tamoxifen appears to be affecting tumors that would present themselves as being ER positive. [Slide] To summarize the findings in terms of breast cancer, tamoxifen reduced the incidence of invasive breast cancer by 45 percent. Reduction is seen in women of all age groups and at all levels of breast cancer risk. And, tamoxifen also reduced the incidence of non-invasive breast cancer. [Slide] I would now like to turn to other cancers that were diagnosed in the trial, starting with endometrial cancer. When we began the trial we were aware that endometrial cancer was a potential risk for women who were using tamoxifen. Indeed, from the world's literature involving treatment trials, we estimated that the risk of endometrial cancer might be elevated about 2-3 fold overall in the population. Indeed, that is exactly what we found in the prevention study. In the placebo group there were 14 cases of endometrial cancer diagnosed compared to 33 cases in the tamoxifen arm, for a relative risk of about 2.5. When this was broken down by age group, there was really no difference evident at this time between the treatment groups for women who were less than 49 years of age at the time they entered the trial. On the other hand, for women who were greater than 50 years of age when they entered the trial there was a substantial difference, 6 versus 26 cases. [Slide] It is important to note that all except for 1 of the cases in the trial were diagnosed at an early stage. All of them were FIGO stage I, 13 on placebo and 33 in the tamoxifen group. There was 1 case that was a stage IV, and this occurred in the placebo group. It is also important to note that most of these cases were picked up by a mechanism which included annual pelvic exams and every 6 months a questioning of the individuals regarding gynecologic symptoms, and stressing to the individuals that whenever gynecologic symptoms occur they should report them immediately and have them followed up. About 3 or 4 years into the trial, in 1995 actually, NSABP began paying for women who wished to have endometrial biopsies as part of their follow-up every 6 months on the trial. Some of the women did participate in that. [Slide] Only about half of the women in the trial who were eligible for screening -- and when we say eligible now, we are talking about women who actually have uteri, and I might add that all the rates that we are talking about here for endometrial cancer are based only on women who are at risk, women who had a uterus. About 37 percent of the women who came into the trial, at the time of randomization had a hysterectomy. So, 67 percent of the women in the trial were at risk for endometrial cancer, and when we calculated these rates these were based only on women at risk. That is why you see on the bottom line that a little over 4000 women in each arm were at risk. This group of women actually participated in endometrial sampling; this group did not. This is the breakdown of the total number of cancers that were detected among the group who were sampled and the group who were not sampled. As you can see, the rate of detection of cancer was not statistically significantly different, 0.6 percent in those who were sampled compared to 0.5 percent in those who were not sampled. [Slide] To summarize the conclusions in terms of endometrial cancer then, tamoxifen increases the risk of endometrial cancer. Annual pelvic exams, directed questioning regarding gynecologic problems and the prompt reporting and evaluation of symptoms can be successfully used to detect endometrial cancer in early stage. The use of endometrial biopsy did not significantly improve the rate of cancer detection, and the small difference in detection does not justify the use of endometrial biopsy as a screening method. Consistent with these findings, when we are planning our next prevention study we are not recommending that endometrial biopsy be included as part of the routine follow-up. [Slide] Turning now to other cancers, cancers other than the breast and cancers other than endometrial, this table summarizes the complete experience of the trial. Overall, there were 88 other cancers in the placebo group compared to 85 in the tamoxifen group. You can see here the distribution by all the different cancers. It is important to note a few of these because some of these were suspected as being possibly associated with tamoxifen and it turns out that they were not. There is no difference in colon cancer. No difference in rectal cancer. No liver cancers. In fact, there is no difference in any cancer at all as you look down the list. [Slide] Ischemic heart disease was included in the trial because it was known that tamoxifen reduces levels of lipids and perhaps that would result in a reduction in the risk of heart disease. There were actually 4 different specific ischemic events that were included as endpoints in the trial. These included fatal myocardial infarction; non-fatal myocardial infarction; a category of illness we called severe angina, and that was defined as having angina that required angioplasty or coronary bypass surgery; and the last endpoint that was included was called acute ischemic syndrome, and this included individuals who had changes on the ECG but not necessarily elevated enzymes or chest pain, or individuals who had severe chest pain and required hospitalization but did not have to have surgery. This table shows the results from those endpoints. First of all, overall there were 59 ischemic events in the placebo group compared to 61 in the tamoxifen group. Dealing with just the myocardial infarctions, there were 27 in each arm. If you were to cull out those that were fatal MIs, the numbers would be 8 versus 7. In terms of the severe angina, those requiring bypass or angioplasty, 12 and 12 -- the same number in each arm. In terms of acute ischemic syndrome, the numbers were also the same, 20 and 22. So, at this time the results of the trial do not support the contention that tamoxifen does reduce the risk of ischemic heart disease. [Slide] Fracture events -- fractures were included as a possible endpoint because of the estrogenic effect of tamoxifen thought to be preserving bone. To evaluate this we included 3 specific endpoints of fractures that we identified a priori which we thought were fractures that would be more likely to represent osteoporotic type of fractures. Those 3 endpoints included fractures of the hip, fractures of the spine and fractures of the lower radial called Colles' fractures. Overall, there were 61 of these type fractures in the placebo group compared to 33 in the tamoxifen group, for a reduction of about 46 percent overall of these types of fractures. Looking specifically at the types of fractures that occurred, hip fractures were 20 versus 9; Colles' fractures were 12 versus 7; and spine fractures were 30 versus 19. These numbers don't add up exactly to 61 and 33 because there is 1 woman here who had a hip and a wrist fracture. There are 2 women here. One had a hip and spine and one had a hip and wrist fracture, and they are counted individually in that level. [Slide] Vascular events -- as I indicated before, in addition to endometrial cancer we were also aware that there were other potential risks associated with tamoxifen. We learned this from the extensive history that we had with treatment trials. These included thromboembolic events such as pulmonary embolism and deep vein thrombosis. This bar chart shows the distribution of rates and the number of events occurring for pulmonary embolism, deep vein thrombosis, stroke and transient ischemic attack. In terms of pulmonary embolism, there were 6 cases in the placebo compared to 18 cases in the tamoxifen arm. Three of the cases in the tamoxifen arm resulted in death, and this difference was statistically significant. In terms of deep vein thrombosis, there were 19 events in the placebo arm compared to 30 events in the tamoxifen arm. This difference was not statistically significant. In terms of stroke, there were 24 in the placebo compared to 34 in the tamoxifen arm. Again, this difference was not statistically significant, and there really was no difference between the 2 arms in terms of transient ischemic attack. [Slide] Ophthalmic events -- when we planned the study there were also reports in the literature suggesting that tamoxifen might have some impact on visual effects. For that reason, we did two things. First, we undertook a special study in one of our that trials, NSABP-14 and, secondly, we included questions in follow-up information in the P-1 trial to help us understand and collect information regarding the occurrence of eye toxicities, In terms of the NASBP-14 trial, approximately 300 women were called in and participated in very extensive eye examinations to determine if there were problems. The results of that study indicated that there were no problems with the development of retinal crystals -- retinal crystals is one of the things which was theorized to be one of the potential side effects. There also were no problems with macular edema or macular degeneration. However, the results from the study suggested that there might be a problem with cataracts. [Slide] In the prevention study we also found that there was no relationship between macular degeneration and exposure to tamoxifen. The actual number of events and the rates were identical between the 2 arms. On the other hand, we did find that there was a difference in the rates of cataracts. Of 483 women who came into the trial in the placebo arm without cataracts, developed them during the course of the trial compared to 540 in the tamoxifen arm. This represents about a 13 percent increase in the risk of developing cataracts. Among those women who developed cataracts, 63 out of the 483 went on to have cataract surgery compared to 101 out of the 540 in the tamoxifen arm. This represented about a 60 percent increase in the risk of having cataract surgery. [Slide] The next item I would like to talk about is total deaths. Overall, there were 65 deaths in the placebo group compared to 53 in the tamoxifen arm, 5 of the deaths in the placebo group were due to breast cancer compared to 3 in the tamoxifen arm. There was 1 endometrial cancer death. This occurred in the placebo group, and was diagnosed with a FIGO stage IV endometrial cancer. In terms of heart disease -- all heart disease not just ischemic, ischemic was 8 versus 7; total heart disease is 12 versus 12. Stroke was 3 versus 4. As I mentioned already, there were 3 deaths due to pulmonary embolism in the tamoxifen arm, and so on and so forth. If you look down at every single cause, and there are many causes in here, there are no differences between any cause of death between the arms. [Slide] To summarize the findings from the BCPT, first of all, tamoxifen use prevents invasive breast cancer among women in all age groups and at all levels of predicted breast cancer risk, and a similar effect is evident for the prevention of non-invasive breast cancer. [Slide] Rates of osteoporotic fractures were lower in the women in the tamoxifen group. The risks of tamoxifen include endometrial cancer, thromboembolic events and cataracts. No difference between the that groups was noted for rates of heart disease, other cancers, macular degeneration or other vision conditions affecting permanent vision loss. [Slide] Our conclusions then are that the BCPT was designed as the definitive trial to test the hypothesis that tamoxifen use would reduce the risk of breast cancer. The findings indicate that tamoxifen use can significantly reduce the risk of both invasive and non-invasive disease. [Slide] The weight of evidence from the trial is substantial in comparison to the recently published preliminary findings of the 2 smaller and differently designed European studies. Thus, we conclude that women who are at high risk, as defined in the BCPT, should be considered as candidates for the use of tamoxifen to prevent breast cancer. Summary DR. LEWIS: Thank you very much, Jo. Before we open the meeting to questions, I would like to summarize Zeneca's position on Nolvadex in prevention. [Slide] Tamoxifen, as given in the breast cancer prevention trial, prevents 45 percent of invasive breast cancers in women at high risk. Benefit was seen in all age groups and at all levels of risk. The safety was as anticipated from earlier trials, and is covered in our current label. The definition of who is at high risk is as described in the label and in the trial. This information has been incorporated into our current label. Having identified a woman who is at high risk of breast cancer, it is appropriate for that woman to have discussion with her health care provider to determine if tamoxifen is right for her. This discussion should include the necessity for medical care follow-up because tamoxifen is not a substitute for good medical care but an addition to it. [Slide] It is our believe that good medical care for all women includes regular examinations, mammography and pelvic examinations, and follow-up of any abnormal signs and symptoms. [Slide] Finally, we believe these data support our claim that tamoxifen is indicated for the prevention of breast cancer in women at high risk for developing the disease. Thank you very much for your attention, and I would be pleased to take questions from the committee. Questions from the Committee DR. DUTCHER: The company has given us half of their time to ask questions of them. So, we appreciate that. Dr. Albain? DR. ALBAIN: Thank you, Dr. Dutcher. I think it goes without saying that we congratulate the sponsor and NSABP for conducting this landmark trial. It struck me in the data again, presented this morning, about the courage of the over 13,000 women who consented to randomization in this trial, as well as the extensive support this trial received from the start from the lay advocacy community and breast cancer survivors. With that as an opening statement, I would like to take the discussion right away to one of the major topics of discussion out there since the data was released in May at the ASCO meetings, and that is the admittedly short follow-up at this stage for the endpoint of preventing cancers. I was wondering if you or NSABP could comment on some of the data that is out there that has much longer follow-up, those breast cancer survivors who received tamoxifen for an adjuvant therapy indication, who have now been followed much, much longer than NSABP-14 or perhaps the worldwide overview data that supports a 45-50 percent reduction in risk of second cancers. Is the maturity of that data in any way supportive of this particular indication? DR. LEWIS: I would like to call on Dr. Wolmark to make some comments on the NSABP trial itself. DR. WOLMARK: Thank you. I would like to echo your remarks on acknowledging the role of the 13,388 participants in this trial, without whose courage and perseverance and dedication and selflessness we would not be here today. Relative to your questions as far as the mean time on study and the duration of the effect of tamoxifen, Dr. Costantino showed you the reduction in relative risk over the period of years of follow-up, and that reduction was durable throughout the five years and now into the sixth year. So, even beyond the discontinuation of tamoxifen we still see a reduction. Relative to the data from B-14 where we used the contralateral breast as a surrogate marker for prevention, there too we see that the effect is not a transient one but durable. Those differences that were noted at five years were still very much apparent at ten years of follow-up. That is also true for cumulative analyses of all the NSABP trials relative to the contralateral breast, and is entirely consistent with the overview analysis relative to the contralateral breast, indicating that this is not a transient effect but a durable one. DR. ALBAIN: To follow that up, what are the confidence intervals like out at the 4- and 5-year parts of your annual hazard curve that you showed and just alluded to? We didn't see those on the slide. DR. WOLMARK: Yes, confidence intervals are a reflex response for me to call upon the statistician. [Laughter] So, perhaps Dr. Costantino would like to look up the confidence intervals to precisely address your question, and perhaps you might have another one as he is looking those up. DR. ALBAIN: I have the same question for the reduction in risk of invasive cancers by your predefined risk strata by risk. DR. COSTANTINO: I don't have the exact confidence limits here with me, but I can tell you that -- DR. WOLMARK: It was an excellent question nonetheless! [Laughter] DR. COSTANTINO: -- that the relative risk was about 50 percent. The confidence limits for that individual year approached statistical significance. But there was no indication that there was a difference in the hazard rate over time. I think that is the more important question, were the hazard rates constant over time? And, all the data that we have analyzed, including some of the data that was done independently by the FDA, indicate that the hazards are constant over time. So, there is no suggestion that there might be differences over time. DR. ALBAIN: And what were those generally, those hazards? DR. COSTANTINO: Well, about 6/1000 is what it is in the placebo group and about 3.4/1000 in the tamoxifen group. DR. ALBAIN: Thank you. DR. DUTCHER: Dr. Sledge? DR. SLEDGE: I have several questions I want to ask. If one looks at the hazard rates for endometrial cancer -- I would echo my esteemed colleagues on what a wonderful study this is in terms of its design and development, but I will tell you, as a practicing medical oncologist who takes care of breast cancer patients, I pretty much felt I knew the answer before the study was started in terms of a chemoprevention effect. I think many of us who have worked in this field for many years felt that tamoxifen was a chemopreventive drug before the trial was ever started. So, this primarily comes down to the risk-benefit questions rather than the true scientific question of whether or not it can prevent breast cancer. If you allow for that, I think a number of important questions come up. Let's start with the endometrial cancer question. If I am reading the numbers correctly, 37 percent of the women had a prior hysterectomy and 31 percent of the women were premenopausal. The figures that we were given in terms of hazard rates are hazard rates for the general population of women in the trial but, of course, if I go out to the clinic next week with a woman who is postmenopausal with a uterus, the general hazard rate from the trial is pretty useless in terms of me speaking to that patient. So, what is the hazard rate for a postmenopausal woman who has an intact uterus of getting endometrial cancer in any given year? DR. COSTANTINO: Actually, I did indicate that these are the hazard rates based on women with uterus according to their age. So, these hazard rates you see are exactly what you are asking for. So, it is 3.21/1000 women who have a uterus. DR. SLEDGE: Postmenopausal? DR. COSTANTINO: Over age 50 or under age 50. We used age here as a categorization for menopausal status, as an approximation. DR. SLEDGE: Okay, thank you. The second question again relates to the question of risk. The proposed indication is for women with the risk of a 60-year old and, yet, the average risk of the women entering the trial was considerably higher. Since this is largely a risk-benefit issue, what do we say to a woman who doesn't have quite as high a risk as the woman who entered the trial in terms of whether she should go on tamoxifen or not? I looked in the package insert, and the package insert basically says women went into the trial based on the Gail model. It gives us a number of scenarios in terms of who should be considered for tamoxifen, and then after all the scenarios are given it says that these scenarios only account for 17 percent of the women who went into the trial. How is the average general internist or OB-GYN out in the community supposed to decide who is going to go onto this trial? DR. WOLMARK: Well, I think obviously the information presented today is only relevant for those individuals who met the criteria of increased risk as defined by the BCPT which was, in turn, a modification of the Gail model. I think it is incumbent on us to define whether that individual is, in fact, at increased risk and meets the eligibility criteria for entry into the BCPT protocol. There have been a number of actions that have been taken to widely disseminate this information, to make it user-friendly, and also to be readily available to both the physician or to the individual who is considering the use of tamoxifen. Perhaps Dr. Leslie Ford could comment on what these efforts have been up to this point. DR. FORD: The NCI has obviously been very interested in the issue of how we communicate breast cancer risk to women, both in the context of this trial and in other work that we do. One of the things that we have been working on since the April announcement has been a user-friendly way of assessing a woman's risk of developing breast cancer based on the Gail model, and it has gone through some very early data testing but we are about to start distributing what we call our breast cancer risk assessment tool. It is available by request through our cancer trials web site. We will also be sending copies to the major medical societies, and announcing its availability in the newsletters of the major advocacy organizations and medical societies for distribution. The NSABP will also be distributing these risk assessment tools so women and their physicians can, in a sense, plug in their risk factors and determine what their 5-year time risk is of developing breast cancer and whether it was similar to the women that participated in the study. DR. SLEDGE: I think that is absolutely crucial for a drug like this because I can tell you, looking at the package insert, it is definitely not user-friendly in terms of trying to determine -- DR. WOLMARK: Is there a package insert that is? [Laughter] DR. SLEDGE: I think most package inserts are pretty simple. I think for this drug, if we are talking about adjuvant therapy for breast cancer, it would be a lot easier to describe. DR. HONIG: May I make a comment? I would just add also that in addition to those tables of risk that are in the label as it stands now, if you add in the other categories such as preceding diagnosis of LCIS or age, it actually accounts for a little over 50 percent of the profiles of the women who went on the study. It is not 100 percent, obviously, but it is a little over half. DR. SLEDGE: And that is not clear in the package insert. If I am reading the results correctly, tamoxifen is not eliminating the largest tumors; it is not eliminating the most node-positive tumors; and it is not eliminating the estrogen receptor negative tumors, the ones that we typically think of as bad actors from a survival standpoint, which I think is what patients should be interested in, in the long run. This might suggest a lesser long-term survival advantage. DR. WOLMARK: Well, I am not sure that we are not eliminating the larger tumors, or that we are not eliminating tumors with four or more positive nodes. I mean, these are the characteristics of the tumors that we see that are evolving on tamoxifen. As far as what the ultimate outcome is going to be, I think if you can eliminate breast cancer at some point in its evolution, I think we have no way of knowing whether that breast cancer would go on to become virulent and eventually kill the patient. So, I don't think that we can really comment with any degree of accuracy on what the ultimate effects are going to be vis-a-vis perhaps a less than expected impact on survival. I think the fact that we can reduce it by 45 percent will ultimately translate into a prolongation in overall survival. I don't think that there is any evidence that we are selecting out a more virulent variety of breast cancer as a result of the use of tamoxifen, and I think that we have to emphasize the fact that there has been I think a very clearly defined reduction in the overall rate of invasive and non-invasive cancer. Beyond that, I think we can't speculate. DR. DUTCHER: Miss Cassel? MS. CASSEL: I am here today as a patient representative since I am considered high risk and a target population should the drug be approved. How long would you prescribe the tamoxifen for me, so to speak, and at what age? If I have been high risk for the last ten years, at what age would you prescribe it? At forty? At fifty? And for how long? DR. WOLMARK: The duration of tamoxifen that was used in this trial was for a period of 5 years, and we think that is an appropriate interval to use. Of course, the question that comes up is how do you know that 10 years wouldn't be better? Well, the answer is we don't know since that clearly was not tested in this trial. But we do have some information from NSABP protocol B-14, where we did compare 5 years versus 10 years of tamoxifen in patients who had a personal history of breast cancer who were negative, and whose breast cancers were receptor positive. There, it was demonstrated, to our surprise, that 10 years was not only not better than 5 years relative to the index cancer but was slightly worse. But of greater significance, addressing your question, is that there was no additional incremental benefit to the contralateral breast for the additional 5 years of therapy. So, we believe that 5 years is the optimum time until data to the contrary appear. So, I would suggest 5 years. As far as when it should be started, I mean, from my perspective, I think it should be started as soon as it is known that the risk is such that it would make the patient eligible. If one has a 35-year old woman who is of such risk that she would fit the eligibility criteria for the NSABP study, I think that would be the time to initiate 5 years of tamoxifen. I see no virtue in waiting an additional 5 years to let the risk increase to start at a certain arbitrary time in the future. MS. CASSEL: I am also concerned, in talking to some of the target population, that women have a feeling that is a false safe feeling -- I have the drug, almost as a birth control pill, and I can just take it and not worry about it. I am afraid that they will forget their self-breast exam, their mammogram. This is the feeling of some of the women. DR. WOLMARK: Well, I think we have to be very cognizant of that, and I think that we have to indicate very clearly that this is not a substitute and that we have to continue to exercise the standard of medical care and the standard of screening. DR. DUTCHER: Dr. Raghavan? DR. RAGHAVAN: I have just a couple of questions. I always get a little nervous when two-thirds of the deaths on a list are listed as "other." I recognize that the other deaths from placebo are more common than from tamoxifen, but would you give us a little more information about that broad category? DR. COSTANTINO: I believe a complete list is included in the document that you were provided, but to give you some examples -- let's see, we talked about the breakdown of the cancers -- I am not sure how much detail you want me to go through. We have about 20 different causes, but these are deaths due to brain cancer, 3 versus 1; breast cancer 5 versus 3; colon, 1 and 1; endometrial 1; lung cancer 10 and 8; ovarian cancer 1 and 2; lymphatic system 4 and 1; pancreas 6 and 2. Of course, the first is the placebo arm. Moving down to heart disease, ischemic heart disease 12 and 12; stroke 3 and 4; pulmonary embolisms 0 and 3; unknown causes 4 and 4. Then there were 9 and 7 miscellaneous causes, which accounted for 11 different categories which, from the top of my head, I don't really know. But there was no indication that there was any type of cause of death which was predominant arm more than in the other. DR. RAGHAVAN: And was there a systematic requirement for autopsies where possible? DR. COSTANTINO: There was no requirement for autopsy. We did obtain the death certificates and we did obtain information from autopsy if it was performed, but there was no requirement that autopsy be performed. In other words, this is a community-based study. So, we had to accept whatever standard of care is going on in the community. DR. RAGHAVAN: You commented that there was really, I guess, an anticipated absence of ocular problems and, in fact, maybe a reduced level compared to what was expected. Did you have a mechanism where the participants were actually routinely examined by physicians looking for specific indices? DR. COSTANTINO: There was no routine examination. Our follow-up consisted of at every visit there was a series of questions that the women were asked. The first question was "have you had an eye exam since the last time you visited our clinic, and if you did, what were the findings from that eye exam?" There was also a series of questions specifically aimed at determining vision changes, asking them specifically "have you noted changes in your vision? Do you have more difficulty driving at night?" or different types of things which were included in the questionnaire. So, we have all these screening types of things. Also included as part of our follow-up was that the institutions were required to obtain discharge summaries documenting the diagnosis for all incidents for inpatient and outpatient visits. So, from these types of things there is another mechanism for us to identify women who might have had eye surgeries or eye problems that required some type of inpatient or outpatient care. But we did not have a routine eye exam. DR. WOLMARK: The data from protocol B-14, the that trial where some 303 patients were evaluated for eye changes, that too was a tamoxifen versus placebo controlled trial. That was done in a definitive manner with ophthalmologic examinations, and there I think it was noted prospectively that the changes in the retina, or crystals, or edema, or macular degeneration was not in evidence. DR. RAGHAVAN: My final question, Norm, if you look at your Gail model, the results are really very impressive for the 5-plus group, and there clearly is a difference with low level of risk, and I am also struggling a little in terms of the hazard ratios in the less at risk group. Can you talk about that a little bit? DR. WOLMARK: Well, Jo showed a slide based on risk categories, and in each category there was a reduction. How does one translate that into clinical practice? I think the report from the FDA to ODAC which summarizes our view, I think, very clearly is that it really boils down to an individual choice and an interpretation of risk and benefit, and not every individual will do that in the same way. I think we have to provide the potential participant with a clear overview of the information, given in a very definitive manner, and then I think it becomes a matter of individual choice, particularly for those areas that you allude to, where the risk is below the 5 or the 6 that you allude to. DR. ALBAIN: Just to follow that up, and then I have a new question. At least in your briefing book the hazards do cross the confidence intervals around the hazards, cross 1, in some of these other subsets. Your predefined strata for risk that you put into the randomization were a bit different than these that appear here. Could you comment on what the hazards actually are for the confidence intervals? DR. COSTANTINO: When we stratified at randomization we used relative risk. Those are categories of relative risk. Actually, the relative risk was defined as your 5-year risk relative to an individual of the same age and race but who did not have any risk factors. The reason that we decided to use absolute risk as the categorization is because absolute risk is a much cleaner mechanism to do that. Two people could have the exact same relative risk but have absolute risks which are totally different. Therefore, when we did the analysis we wanted to control for that factor, and the easiest way to do that is to stratify by levels of absolute risk. DR. ALBAIN: Then I would like to turn to some other populations at risk, in particular DCIS and the African-American population. Certainly, you were not choosing DCIS as a primary endpoint but your results are intriguing, and we are also aware you have another trial that has addressed that specifically prospectively. Do you feel that the data are robust enough in P-1 to add DCIS to the labeling, or must we wait a bit longer, and how much longer for your other study? DR. WOLMARK: I think we must wait, and I don't think we will be waiting too long. DCIS was not an entry criterion for this trial. So, I think we have to rely on the data from B-24 and B-17 prior to that, which I think will probably require a different session of this group. DR. ALBAIN: But you did show prevention of DCIS that was quite striking. DR. WOLMARK: Yes, I think to prevent DCIS -- DR. ALBAIN: That is what I mean. DR. WOLMARK: -- based on the entry criteria that we utilized in this trial, very much so. I think it decreases the rates of non-invasive breast cancer, predominantly DCIS. I completely agree with that. DR. ALBAIN: And then the African-American population, you tried very, very hard prospectively to accrue minority communities. Could you comment on that effort, and then how you feel these results could be translated to that population? DR. WOLMARK: I would like to ask Dr. Wickerham to comment on that. DR. WICKERHAM: Dr. Albain, you are right. This is an effort that the NSABP has taken very seriously from the start of the trial, and during the study we spent considerable effort to try to increase accrual from these various populations. Our goal at the outset of the trial was to have a population to reflect women at risk. Despite these efforts, we were not fully successful at that. Only about 3 percent of the women entered are women of color. That really doesn't allow us to make definitive statements relative to these results in those populations. But you should be aware that in our that trials we were more successful in entering women from those groups, 10-12 percent, 15 percent in some of our trials. B-14, which in many ways forms the basis for the prevention trial, has been evaluated and analyzed relative to response to tamoxifen in these populations, and we clearly see no difference in the outcomes. DR. DUTCHER: Dr. Ozols? DR. OZOLS: Getting back to the risk again, the 2 major side effects, endometrial cancer and thromboembolic disease, and the 3 deaths in the that group with the pulmonary emboli, can you get any better profile on which women, you know, may be at risk for those 2 toxicities? The traditional risk factors associated with endometrial cancer -- diabetes, hypertension, obesity, are those heightened by tamoxifen? Likewise, can you identify anybody who may be at higher risk for developing pulmonary emboli? DR. WOLMARK: Well, we obviously examined that, and we are not able to come up with a profile that would identify a subpopulation that would be at inordinate risk, such that they could be eliminated from entry into this trial. We did, however, a priori eliminate those individuals who had a previous personal history of deep vein thrombosis or pulmonary emboli. But examining the actual data of the population that was entered we could not define characteristics that would be associated with increased risk for those events. DR. DUTCHER: Just to follow up on that, about 25 percent of people who were screened and met eligibility actually entered the trial, and a comment was made about medical reasons for exclusions. Was it medical exclusion or was it logistic exclusion? What was the drop-off between those that met the eligibility and those that actually entered the study? DR. WOLMARK: Following the risk assessment, those who were eligible from the eligibility and those who were actually randomized. DR. COSTANTINO: I think the biggest reason for the drop was that women were not interested in participating in the trial. They did not go forward to have the full-fledged medical evaluation. A little over 14,000 women actually went to that level to be medically evaluated to come into the trial, and out of that 14,000-plus 13,388 were actually randomized. So, the major reason for the drop from 57,000 down to the 13,000 was because women were just not interested in being a participant in the trial. DR. MARGOLIN: I have what I think will turn out to be 3 questions. The first one is sort of a biology question and it pertains to the question that Miss Cassel asked earlier on about the best timing for intervention in patients who are identified as subjects at risk. It is just hard to imagine that 5 years of that at basically any time in a woman's life is going to infer a permanent change in her likelihood of developing invasive or non-invasive breast cancer. I am wondering, based on preclinical models or based on any biology that anybody knows, whether, say, early treatment and then some period of time off therapy and then reintroduction of therapy, or if we can somehow improve on what we are trying to do here to prevent breast cancer. DR. WOLMARK: I think we are really limited by the data that we have, unfortunately. I mean, we would like to know where tamoxifen acts in this situation. We would like to know what the molecular mechanisms are. Yet, this was a clinically driven trial and we are left with clinical data. Is there an optimum time at which the intervention should be undertaken that would be better than just starting it when the relative risk becomes apparent? If one were to undertake such a trial clinically, it would require enormous numbers of participants with an enormous amount of support from the agencies, to whom we are forever grateful -- the NCI and Dr. Ford -- and I don't think at this time it is a practical endeavor. I mean, we would much rather go on and determine if we can find drugs that perhaps have the same efficacy with fewer side effects which would make that issue moot to a certain extent because they could be given longer and with greater degree of definitive intervention. DR. MARGOLIN: Thank you. My second question is I believe the study was noted as being insufficiently powered, or at least was closed at a point where it was insufficiently powered to detect survival differences. Is it expected that after a certain number of events have occurred, after a certain follow-up, that we will be able to see a potential survival difference, or is that just not going to be possible with this database? DR. WOLMARK: If we were to have primarily done a survival endpoint, I think we would have required an additional 10 years of follow-up and a considerably greater sample size, but I would like Dr. Costantino to comment on what it would have taken to have configured this trial for a survival endpoint for breast cancer. DR. COSTANTINO: We never did design the trial to be able to have the power to detect a survival difference because it would have required doubling the sample size and much longer follow-up, as Dr. Wolmark indicated. We do plan to continue following those women. We will learn more information about survival benefits, but it is highly unlikely that we will ever have statistical power to show a significant difference in survival. It requires larger numbers and a longer follow-up period. DR. MARGOLIN: I have one additional question, whether there are plans to go back and do some genetic studies of subjects enrolled in order to detect potential interactions with BRCA 1 and 2 or other genetic risk factors. DR. WOLMARK: Yes, that is I think an important commitment and those trials are about to be launched. Certainly, that is a very important issue. We have collected serum and lymphocytes from the women who participated in this trial, and we will start to analyze BRCA 1 and 2. Mary Clare King will be doing this in the very near future. We will be able to determine definitively what the benefit is in those individuals who have BRCA 1 and 2 abnormalities. Additional comments? DR. SCHILSKY: A quick comment and a question. It is striking to me that the leading cause of cancer death in this study is lung cancer. It is too bad tamoxifen doesn't prevent that. DR. WOLMARK: Oh, there wasn't a reduction in that? [Laughter] DR. SIMON: The question I guess has to do with how the participants in the study have now been informed of the results, whether women who were randomized to placebo have been advised to take tamoxifen and, if so, how might that confound the future interpretation of the results with continuing follow-up? DR. WOLMARK: We have a covenant with the participants that they would be among the first to know the data, and we did not want to repeat the unfortunate events of some of the earlier episodes that affected this trial where the participants learned what was going on from the newspaper. Despite our diligent efforts to avoid that, we were not entirely successful in this trial since the data were previewed in a well-known newspaper prior to the time that we were able to transmit that information through a widely publicized press conference that I believe took place on April 4. The participants have been formally apprised. That process was in place as the data were being disseminated, and those individuals who were on placebo are given the opportunity to go on 5 years of tamoxifen. Zeneca has been very gracious in providing that medication to these participants. Also, those individuals who did not complete the 5 years of tamoxifen who were randomized on this trial will have the opportunity to complete the full 5 years of tamoxifen. As far as what does that do to our ability to continue to monitor the differences between tamoxifen and placebo, clearly those are attenuated in that this trial has been unblinded and that we will now have crossovers, but to what extent we do not know as yet. We will obviously continue to follow these patient cohorts and, certainly, those that are on tamoxifen will continue to provide data, and we believe we can continue to model the events in the placebo arm. So, I think it will provide useful information but the primary endpoint of the trial is obviously affected by the unblinding. DR. DUTCHER: Dr. Simon? DR. SIMON: I have several questions. One, several people have noted the concern about the limited follow-up. There is not a whole lot that can be done about that, but you have basically presented data that was available to the data monitoring committee last January. Can you give us updated data on number of events in the placebo and tamoxifen group for the 3 age groups for invasive breast cancer? DR. WOLMARK: Obviously, you know, the data provided to this committee are the data that are going to be utilized so I would rather not go into the data for the updated analysis, only to tell you that the differences are even more compelling. DR. SIMON: Why do you not want to give us the updated data? DR. WOLMARK: I think that we had a cut-off that we all agreed to a priori; that this was submitted to this committee for their review; and I think that is the data set that is going to have to be used to make the decision. DR. SIMON: Well, typically, you know, when you present data to a data monitoring committee that is not up to date to that minute anyway. You know, there is a distribution of time since patients were last seen and evaluated. So, that data actually may be a year old at this point really in terms of what it represents in terms of when patients were last seen. Well, let me go on to my next question. Do you have information about the hazard rate over time for the ER-negative cases, particularly in the tamoxifen arm? DR. WOLMARK: Jo, the hazard rate for the ER-negative cases in the tamoxifen arm? DR. COSTANTINO: Over time? DR. WOLMARK: Over time. DR. COSTANTINO: I don't have that with me. DR. WOLMARK: The answer was no, he does not have it with him, and he wondered why you were asking the question. DR. SIMON: Well, because really, you know, one question is whether you are treating with tamoxifen in subclinical cases that might have materialized as ER-positive tumors -- by the selection process will materialize as ER-negative tumors, and whether you will see that there is some trend of that happening in later periods of follow-up. DR. WOLMARK: Jo? DR. COSTANTINO: I can tell you that we didn't see that kind of trend. If you consider that the ER-positive tumors were 80 percent of the tumors and we did see hazard rates over time that were constant, we would suspect that just taking out those majority of things is not going to change the pattern, but I didn't see the type of pattern that you were suggesting. DR. SIMON: I have a couple of other questions. One is that I have some concern about what we are supposed to conclude in terms of what group of patients these results apply to. One, it is one thing to say what the eligibility criteria were and that is not to say what patients actually entered the trial. In terms of communicating these kind of results in terms of who these results apply to, it is really not an issue of even simplifying in a user-friendly way the Gail model. The real issue is what women went into this trial, because there may be women who were eligible according to the Gail model but if they are not well represented in this trial then we probably can't have much confidence that the results apply to them. I guess I haven't really seen a clear explanation of what the women looked like who went into this trial. I guess the second issue is that it is one thing to say that the risk of breast cancer of a woman is equivalent to that of a 60-year old woman, and it is something else to say that the results actually apply to a 60-year old woman. Most of these women, I think two-thirds of them or something like that, were under the age of 60 and they got into this trial because they had other risk factors. So I think we have to be somewhat careful in assuming that because the Gail model said that their risk factor was at least the risk of a 60-year old woman that the results actually apply to a 60-year old woman. The only basis we have for that is, you know, where you break it down by age. You know, that is a relatively small subset. It looks like the effect is just as great for them as it was for the other women. But I think we really have to be very careful in trying to sort out who the results apply to. That is sort of a comment, not a question. I do have one other question, and I would like to sort of get your general medical interpretation of it. There were 69 fewer cases of invasive breast cancer on the tamoxifen arm, but there were 19 additional cases of endometrial cancer. There were 39 more cases of vascular events on the tamoxifen arm, and there were 38 more cases of cataracts requiring surgery. So, how do you make that risk-benefit equation? DR. WOLMARK: Well, I don't think it should be up to me nor any other physician or someone who delivers health care to compel anyone to go on tamoxifen or not go on tamoxifen. I think that it becomes an individual decision after the risk and benefits are thoroughly reviewed and after that information is transmitted in a very clear and well-defined manner. Having said that, and since you asked for an opinion, I think that there are categories that, from my perspective, clearly fall out where the benefits unequivocally outweigh the risks. I think those subsets would include those women who are under 50 years of age where the excess of adverse events is small; those women who are over 50 years of age who have had hysterectomies, and in our patient population that accounted for a substantial proportion; those women who have had a personal history of lobular carcinoma in situ; and those women who fulfill the eligibility criteria and also have atypical hyperplasia. I think in those instances, from my perspective, the benefits clearly outweigh the risks. I think in the other categories it boils down to an issue of personal choice and personal decision. I think what some people would consider as inordinate risks others would gladly accept. DR. DUTCHER: Miss Beaman? MS. BEAMAN: Would you reference the data that you have for the women who were taking tamoxifen and developed breast cancer as to whether this cancer was of the more aggressive type? DR. WOLMARK: I think in the slides that were presented relative to the distribution of women who did develop breast cancer while they were on tamoxifen there certainly was no evidence, from a nodal standpoint as well as a tumor size standpoint, that the tumors that developed on tamoxifen were more aggressive or more virulent than those tumors that developed in women who were taking placebo. So, there is no evidence that tamoxifen culls out a more virulent subset of breast cancer while suppressing the more benign forms of breast cancer. I think that appears in the slides that you have in the handout. DR. DUTCHER: Dr. Johnson? DR. JOHNSON: Actually, I want to follow-up, if I may, on what I think Dr. Sledge addressed earlier. Certainly, nodal status is one of the most important, if not the most important, prognostic factors and size as well but there is no mention about tumor grade here which clearly has an impact. If all 154 tumors that appeared on placebo were low grade and all 85 on tamoxifen were high grade tumors there might, in fact, be a difference in outcome even though the other factors were identical. I wonder if maybe you have some data regarding grade. I didn't see any of that information. DR. WOLMARK: No, we have no data on grade. I think it would be nice to know what the HER2 status of the tumors was, and is, and will be, but we don't have that information. DR. JOHNSON: Is that information that we can expect will be forthcoming in the future, or is it simply something that won't be followed-up upon? DR. WOLMARK: We are collecting slides and blocks, which the protocol has mandated, on all events that occur in this study and, hopefully, that information will eventually be forthcoming. DR. JOHNSON: And, if I may follow-up with one further question, the death rate from breast cancer, as has been pointed out, is really rather small in the trial overall and it is similar in the 2 arms which, actually, is sort of interesting given the fact that the number of overall cancers is twice as great on the placebo arm. So, do we have any information about the status of those women who have developed breast cancer at this juncture? Again, just to take the extreme, if all 85 of the women on the tamoxifen arm now have stage 4 disease and all 185 on the placebo arm have stage 1 disease there may be an indication of a difference in the aggressiveness of the tumors. Do we have that data? DR. COSTANTINO: We are collecting information regarding recurrence, and we do have that but it is not complete at this stage. Dr. Paik is in the process of reviewing all the pathology slides as we speak but we do not have that information accumulated as of yet. DR. WOLMARK: But, David, why would you think that there would be that disparity? DR. JOHNSON: Well, because unlike George, I don't have the ability to see the future. [Laughter] He predicted that this drug was going to work and I just didn't realize it was going to work. So, I was really happy that the study was done. So, you know, data is what really drives my decision-making, or I like to think it does. So, I don't believe that is the case. Just because I asked the question doesn't necessarily mean I believe that is the answer. I think I would like to know, and I am sure everyone sitting over here as well as in the audience would like to know those data as well. And, if it were to turn out that way, then it would be disturbing. DR. WOLMARK: George, perhaps you could save us a lot of time by telling us raloxifene versus tamoxifen for the STAR trial? [Laughter] DR. SLEDGE: I would be glad to tell you that afterwards. DR. DUTCHER: Dr. Margolin? DR. MARGOLIN: I have one question and one comment. The question is that -- unless I have missed it and it has already been presented -- we have heard at various times well before this meeting that the subjects who were accrued or registered to this trial turned out to have higher at least relative risk of breast cancer than was expected and was planned for the original accrual. I am curious to know, at the end of the trial, at least based on the placebo arm data, whether the incidence of breast cancer reflected what was expected based on that revised accrual estimate. DR. COSTANTINO: Indeed, it did. It was about double what we expected. DR. MARGOLIN: Thank you. My comment is that if the drug is approved for this indication, and I think that the world, certainly the U.S. but the world really worships what the NSABP says and does, and the NCI as well, and it would be very crucial that very firm guidelines be given in terms of selecting subjects for that with this type of intervention. DR. ALBAIN: I have a question for the sponsor. Your choice of wording in the indication, using the word "prevention." Typically, when that word is used you have the luxury of long-term follow-up, in particular like we do in B-14 and the worldwide overview for prevention of contralateral breast cancers. Would you consider perhaps softening that statement to say reduction in risk of occurrence of first cancers because that is really what we have seen quite dramatically by this data? DR. SIMON: Right. Actually, that is what was seen in the overview also in the contralateral breast, and we would agree with you. I think "prevention" means it doesn't occur and it also means risk reduction. What we are looking for here is a way of getting the message across to the average person. It means something to us. The trial was called prevention; Dr. Leslie Ford's group is prevention. We are not preventing all breast cancers. Clearly we are not. But this is a major step forward, and I would like to think that we could retain the term "prevention," describing it as it was described in the manuscript which states that it is a reduction in the number of breast cancers that are anticipated. DR. DUTCHER: Dr. Simon? DR. SIMON: The women over the age of 60, were they a representative group of women or did they have high risk features? DR. COSTANTINO: Actually, if you look at the hazard rate in the placebo, you can see that their risk of breast cancer was among the highest of all the women in the trial. So, it is true that being over 60 was an eligibility criterion and so you got in, so your risk could be as low as 1.66 theoretically but, indeed, the rate and the hazard in the placebo group was over 7/1000. So, they were essentially comparable in risk to the women in the other groups. As far as being representative, I think you mean representative of the general population? DR. SIMON: Right. DR. COSTANTINO: I don't think any of these women are representative of the general population because they have been selected out to be at high risk for breast cancer. DR. SIMON: So, how do we know -- DR. COSTANTINO: They volunteered for the trial. DR. SIMON: So, how do we know that these results apply to a typical spectrum of women over the age of 60 in the United States? How do we know who these results apply to? DR. COSTANTINO: We know the results are consistent across all categories or risk -- DR. SIMON: No, but you say age and you show over 60, but these are not a representative group of women over 60. DR. COSTANTINO: That is true of any clinical trial. I think the best we can say is that within the trial we were not able to demonstrate any population which did not show benefit from the treatment; that we can think of no reason to conclude that the women in the trial, as far as effect is concerned, would not represent the general population. So, I don't see that we can do more than that. DR. SIMON: So, what is your proposal for who you are recommending this for in terms of an indication? Is it women whose risk would satisfy the Gail model? And, if that is the case, that would include all women over the age of 60 but we don't really have any indication that these results apply to typical women over the age of 60. So, I find that a real inadequate specification of who these results apply to. Can you clarify it for me? DR. COSTANTINO: I just don't understand your argument, Richard. DR. SIMON: Well, maybe I can clarify it. You have a study of high risk women and this study seems to have shown a benefit of tamoxifen for this group of women, and now we are trying to figure out who this group of women are before we wind up recommending this drug, with its side effects, for all women. If we recommend it for all women whose risk is at least equivalent to that of 60-year old women, then we recommend it to a large group of women over the age of 60 in this country who probably may not have been represented at all in this clinical trial. You may have gotten a result that worked because of some genetic features that these women had that gave them other high risk features that really don't have anything to do with your typical -- DR. COSTANTINO: The eligibility criteria for this trial was simply being over 60. It had nothing to do with the Gail model. We are recommending that those same type of criteria be applied to women who are considered candidates for the drug. Simply being over 60 makes you a candidate so that you can go forth and make these kind of comparisons of the risks and benefits and decide. For women who are under 60, we are recommending using the Gail model just as it was applied as eligibility criteria. So, our recommendations are pertaining specifically to the exact same type of women who were deemed eligible for the trial. DR. SIMON: Well, that is what I was saying. There is a difference between being eligible and who actually got into the trial. I think when you make recommendations as to who the results of the trial apply to, you have to look at who was in the trial, not who was eligible for the trial. DR. WOLMARK: Richard, I really agree with Jo on this. I think that this is an inherent problem in every clinical trial you do. You set out the eligibility criteria and whoever actually enters the trial may or may not, you know, fulfill the entire spectrum of the eligibility criteria but that does not justify anyone from going back and retrospectively culling out a subset to say that this is more representative of those individuals who actually entered the trial. We don't have the power to do that, from my perspective and, more importantly, I don't think we have the right to do that. DR. SIMON: I am just trying to figure out whether the results of this trial apply to the typical woman over the age of 60 who doesn't have other high risk features. DR. DUTCHER: Dr. Honig? DR. HONIG: We were concerned about that also with that particular age group, and I don't have the numbers with me but we looked at women over 60 to see if, for example, all of them had positive family histories, or a significant proportion had LCIS or atypical hyperplasia, and that was not true. At the time though we did not have the risk disc so we couldn't run the Gail models, but most of them appear to have a combination of the other factors that went into the Gail model, if that helps answer your question in part. DR. WOLMARK: Yes, I think we apply, you know, what we believe and we will be using the criteria for the next NSABP trial, the study of tamoxifen and raloxifene as they were used for the BCPT, with the exception that this will be limited to postmenopausal women. I think it would be a mistake and somewhat disconcerting to try and fine-tune the characteristics of patients who would benefit from tamoxifen based on a subset analysis of the NSABP population. I think we should use the criteria as they were applied to the BCPT. DR. LEWIS: I would just like to comment that there was one other criterion that was left out, and that was a discussion with the patient, and we plan, working with the National Cancer Institute, to stress this as a critical part of a decision for a woman which empowers the woman to elect to take tamoxifen. Certainly it is not Zeneca's intention to take all women at 1.66 and say that tamoxifen is right for them. As a matter of fact, in our label there is a sentence which says tamoxifen is not right for all women at high risk -- something to that effect, and we do plan to handle that responsibly. DR. DUTCHER: Dr. Sledge? DR. SLEDGE: I would like to get back to Dr. Albain's comments a few minutes ago. On this question of prevention, I think it is reasonable to ask whether what we are seeing in this trial is true prevention. If we look at the risk ratio by year, a great point was made that it was pretty consistent over the first 5 years. It is real hard for me to believe that what I have always thought of as chemoprevention, that is to say, the transition from an earlier to a later place along the stage of development of cancer is what you are seeing when you don't see a cancer in the first year of a trial, and I think we have to assume that what we are dealing with in the early years is chemosuppression of existing invasive tumors rather than true chemoprevention. DR. WOLMARK: I would have no argument with that. And what are we seeing in later years? DR. SLEDGE: Well, presumably the later out you get, the more likely you are to be seeing chemoprevention. I don't think there is any argument about that, but I think to say that all of these early cases where we are seeing a difference represent chemoprevention just simply probably isn't true. DR. WOLMARK: Yet, the ultimate effort is to reduce the incidence of invasive cancer, reduce the incidence of non-invasive cancer and its clinical consequences. DR. SLEDGE: I think we can agree on that. DR. DUTCHER: Dr. Margolin? DR. MARGOLIN: I would like to know whether the data on the effect of tamoxifen in this cohort reduced the expected risk of breast cancer down to that of an age-adjusted woman with no additional risk factors, or if it is still higher by some relative risk. DR. WOLMARK: Jo, do you want to comment? DR. COSTANTINO: Let me make sure I understand your question. Your question is -- DR. MARGOLIN: Does tamoxifen normalize the risk of breast cancer? DR. COSTANTINO: Did it bring it back to essentially no excess risk? I have not specifically done that analysis to compare the women on the tamoxifen arm to what would be expected, but without having done that I can say I am sure it did not take it all the way down because in general the rate on the tamoxifen arm was about 3.4/1000 consistently across all ages, and I know that is not the baseline rate for women who don't have any risks. DR. MARGOLIN: So, that information would be a necessary part of the counseling in terms of the subjects on this treatment. DR. WOLMARK: It reduces it 45 percent overall in this analysis, but not back to baseline. DR. DUTCHER: Dr. Albain? DR. ALBAIN: It is not preventing ER-negative cancers essentially, among a few others probably. DR. WOLMARK: I think one can theorize that is the case. DR. DUTCHER: What is the long-term follow-up plan on this study? DR. WOLMARK: We will continue to follow these patients as long as we and they continue to agree to be followed. DR. JUSTICE: I would just like to follow-up on that question. I think there was perhaps some mis-communication about the updated data. I think I agree with what Dr. Wolmark says, that we don't want the updated data presented today. We will certainly ask to see it, and we will certainly ask to see follow-up data on both efficacy and safety, but we haven't actually seen the updated data and we don't want a lot of different numbers floating around. So, I think that is the hesitation Dr. Wolmark had, not that he is not willing to provide us with it. DR. WOLMARK: I think that was very elegantly stated. DR. DUTCHER: Are there any other questions for the sponsor? If not, we are going to take a 15-minute break. We will be back here at 10:20. [Brief recess] DR. DUTCHER: We are going to begin the FDA presentation. FDA Presentation DR. HONIG: Thank you. I will be presenting the FDA analysis of tamoxifen for prevention of breast cancer in women at high risk. [Slide] In every FDA presentation you see a slide similar to this one, but I would like to emphasize that for this review in particular it was truly a collaborative effort to be able to review this much material, on this many patients in such a short time frame. I wish I had time to detail everyone's contributions but I would particularly like to mention several people. Donna Griebel, another medical reviewer in our Division who reviewed and analyzed the case report forms for stroke; Karen Johnson who, prior to her departure from FDA, reviewed all of the case report forms for invasive and non-invasive breast cancer; and Alison Martin, who analyzed and reviewed all the case report forms on the endometrial cancer patients in this study. I would also like to spend a minute talking about the administrative time line because it was certainly a challenge for everyone involved in the application to be able to process and submit this much data, and also to review it in a timely fashion. [Slide] As you have heard, on April 2 the NCI and FDA were notified that there were significant efficacy results in this trial. On April 23, there was a pre-sNDA meeting designed to facilitate the submission of the application in a timely fashion. Along those lines, FDA agreed to accept the report that had been prepared for the ERSMAC committee and the BCPT technical report in lieu of a study report, and also asked that the draft manuscript of P-1 be submitted as soon as possible to us. Of course, the NSABP was busy with a number of other commitments, as well as trying to write that manuscript which was submitted. We waived the requirements for the integrated summaries of safety and efficacy, and it was agreed that the data would be submitted electronically. On April 30, 1998 the SNDA was submitted and, as you can see, we are here 4 months later at ODAC to discuss the results. [Slide] When we initially received the electronic database tables, it was clear that there were some limitations. We didn't have primary data which is usually the type of data that we review. For example, the primary endpoints were at first listed yes/no without dates, and we didn't have any of the characteristics of the breast or endometrial cancers that occurred on study, and we didn't have a complete list of risk factors. But we had multiple discussions with NSABP. We ironed out some of the technical problems in transferring the data and, as you can see, we, in fact, worked out a way for these to be submitted. [Slide] We got the first additional set of requested elements on July 23, and the last set of data was submitted to us on August 4 so that we were able to go through the majority of this data in time for ODAC. [Slide] As is usual in a clinical trial, we requested specific case report forms on participants in the trial. We requested those for participants who had died during the study, who had developed both invasive and non-invasive breast cancer, endometrial cancer, DVT, PE and stroke. With those listings we received approximately 625 that were submitted and reviewed in detail by the members of our team. [Slide] What I would like to do during this presentation is cover the following topics. I don't want to go over details that have already been presented by the applicant, and I would like, instead, to concentrate on areas that perhaps we have a slightly different interpretation of or some additional information. [Slide] As you have already heard, NSABP P-1 is a large randomized, double-blind, placebo-controlled trial of tamoxifen for 5 years. Again, you have already heard about the number of participants on study. Most or the data are with reference to this denominator, however, 13,118 had additional follow-up and this is the denominator for certain of the adverse events, such as hot flashes, that we will be discussing later. [Slide] I want to spend just a few minutes on the requirements for trial entry because this has a bearing on how this drug will be used in clinical practice if it is approved. In the trial a multistep procedure was required for entry. In the recruitment phase women were first seen and given information about the trial and had the opportunity to ask some questions about the study. Then if they chose, they could fill out a risk assessment form that listed the risk factors for breast cancer that would be entered into the Gail model. This risk assessment form was then forwarded to NSABP, and in a separate second protocol eligibility assessment the participant returned, having read materials at home, was able to ask and have more questions answered, and was able to discuss the actual risk assessment form generated by NSABP. If at that point she wished to continue with study entry and she was eligible on the basis of breast cancer risk factors, she signed an informed consent and then proceeded with the staging studies required for entry. It is worth noting that all eligibility factors were reviewed by the NSABP as well as by the local institution, and that includes both breast cancer risks and medical conditions. [Slide] In a third, separate visit at the study enrollment phase, the results of the studies were reviewed. If the participant were still willing to go on study and eligible, her informed consent was reaffirmed, and she was then randomized with a number of prospectively specified stratification factors. So, this was a multistep process. There were at least 2 institutions involved in ensuring that the participant was informed and eligible, the NSABP and the local site. I mention this only because it is unlikely in a busy clinical practice that practitioners are going to be able to devote 3 separate visits to this level of detail. So, it is very important that we all develop patient education materials that will allow women to make an informed choice about whether they wish to take tamoxifen or not. [Slide] There were many protocol amendments during the course of the study, however, I would say that there were probably 2 major protocol amendments and you have heard about some of these already. One was that on September 24, 1994 a requirement for baseline and annual endometrial screening for newly randomized participants was added. Participants who were already on the trial were offered screening but they could decline and continue on study. Also, in October of 1996 there was a formal decrease in the sample size based on the higher than expected number of events. This had been prospectively specified in the protocol though. It had called for an interim analysis to calculate sample size. [Slide] In terms of on study conduct, the protocol said that you could be unblinded and know your treatment assignment if you developed invasive breast cancer, or if your physician felt that there were medical conditions that warranted knowing the treatment arm. As you might expect, there were some non-protocol specified unblindings. However, whether these occurred because the participants wished to know or the physicians wished to know based on a variety of medical conditions, these were all balanced between the 2 arms. The "other" category is not other medical conditions but, rather, a separate category and you can see overall that there was really no difference. We have examined all of these reasons in detail. They were supplied by NSABP. And, there was no difference. [Slide] In terms of non-allowed medications on study, this is the information that the NSABP had in its database. Hormonal medications, which was an accumulated group of estrogen, progesterones, androgens, as well as a full complement of hormone replacement therapy. They also collected information on oral contraceptive use; the use of open-label tamoxifen and raloxifene. No one on study used raloxifene. I really wanted to spend the time on the top line. As you can see, if you look at the women on placebo and the women on tamoxifen who are listed as using hormonal therapy at any time during the study, the number looks large. But if you really restrict it and look at the number of women who used it while they were taking the study drug, it is a relatively small number. Less than 1 percent actually used these medications. This is in distinction to the European studies, which we will talk about later, which allowed the use of hormone therapy in various forms. There were a few limitations of the database that I will review. We did find some instances in the case report forms where women used hormonal therapy that were not in the database. These were relatively few instances. Also, the database was designed to capture the date of the first use of these medications. So, we don't have duration of use and we don't have multiple events of use. Overall, our impression from looking at the case report forms is that it was still a relatively small number of women who used these medications for short times. I think a good example would be women who had episodes of dysfunctional uterine bleeding who took short courses of Provera several times, for example. [Slide] Compliance, as you have already heard was very high. For women who started their therapy and subsequently discontinued therapy, the most common reasons are listed here -- hot flashes, anxiety, vaginal discharge. The hot flashes and vaginal discharge are consistent with what is already known about tamoxifen and its side effects. [Slide] I am going to move on to the endpoints of the study. First, it is important to note that all events in all participants were reported unless the participant withdrew consent or was lost to follow-up. In oncology treatment trials I think we frequently think about events being reported on drug or within 30 days of stopping drug. This is not the case here. Overall, participants who were followed had all of their events recorded in the database. In our review of case report forms we could find potentially 1 breast cancer that was perhaps not captured in the database. We are still discussing this with NSABP. The NSABP also set the rules up prospectively that the worst event per participant would be recorded. So, if you had angina and then subsequently had an MI, the MI would be recorded but not each individual related event. Similarly, if someone had a TIA and a CVA, it would be the stroke that would be reported in the database. This was true for nearly everything except fractures. That was in distinction where all of the fractures per participant were reported, not just the first fracture. [Slide] So, with regard to invasive breast cancer, you have already heard that there were 154 cases on placebo and 85 on tamoxifen. We looked at the number of cases that were diagnosed on each arm after stopping the study drug, and you can see that there are relatively comparable numbers on each arm. Within the follow-up available to us on the study, there was no evidence of a rebound increase in the number of cases after the study drug, tamoxifen, was stopped. We also saw reductions in the number of breast cancer cases in all the prospectively defined subgroups, specified by the sponsor. In fact, we found reductions in every retrospectively defined subgroup that we could think of at FDA. The reductions were seen in participants who had a family history, regardless of the number of affected first degree relatives. We were able to carry this out for none and then 1 through 4, I believe. We also saw reductions in participants who did not have a family history. At first we were concerned that perhaps all the risk and all the benefit was being seen in a subgroup of women who were at risk because of a family history, and that was not true on our review of the database. [Slide] The only subset in which this beneficial effect of tamoxifen was not observed was in women of color. There were 486 non-white women entered on the study despite the really aggressive attempts on the part of the NSABP to recruit more women of color, and there were 9 cases, 3 on placebo and 6 on tamoxifen. We looked at these women in detail. The risk profile of these women didn't difference from that seen in the general population of women entered on the trial. The characteristics of the 2 groups were not any different either. They were not more aggressive or less aggressive. At this point, I suppose you could say that it is unknown whether there is a differential effect in non-white women but we would favor the interpretation that overall women of color made up a small subset of the population and had relatively few events, and that we just don't have the statistical power to make any comments about that. [Slide] In terms of the case report form review of the invasive cases, we agree that all the cases that were reported were, in fact, invasive breast cancer. We assessed 2 additional cases of invasive cancer on the placebo arm. These had previously been categorized as non-invasive breast cancer, and 1 on the tamoxifen arm as invasive cancer. This was a woman who, after several reviews by NSABP, was ultimately assigned to the category of cancer of unknown primary and after our review we felt it was likely that she had breast cancer. We also reviewed the assessed tumor size based on the original pathology reports. We disagreed with the assessed tumor size for 3 cases on placebo and 1 on tamoxifen. It resulted in minimal stage shifts for these participants and, as I mentioned before, we may have found an additional case that we are still discussing with NSABP. Overall though, even with these shifts in cases, it doesn't change the primary conclusion of the sponsor, which is that tamoxifen did result in a significantly decreased number of breast cancer cases on the tamoxifen arm compared to placebo. [Slide] This shows you the tumor size and nodal status distributions. You have already seen this so I don't want to spend a lot of time on it except, again, to reiterate Dr. Costantino's point which was that tamoxifen was most effective in tumors that were less than 2 cm in size and in cancers that were either node negative or had 1-3 positive nodes. [Slide] This shows the stage groupings for all of the cancers that were identified on study. As is consistent in the general population, most of the women diagnosed with breast cancer had node-negative disease. Some women had node positive. There were 10 cases of inflammatory breast cancer, and 2 women who had either probably or confirmed metastatic disease at diagnosis but, again, you can see that they were not significantly different between the treatment arms. [Slide] Again, you have already seen this slide showing that tamoxifen appears to have the greatest effect in reducing the number of estrogen-receptor positive tumors. [Slide] So, overall we would conclude from our review that there was a significant reduction in the number of breast cancer cases with tamoxifen regardless of the subgroup. At the beginning of this trial and throughout the conduct of the study there had always been concern about use of tamoxifen in younger women. We looked at them specifically. We did not see an excess number of cases in young women, nor did we see more aggressive appearance to the tumors in young women. We would say that at this point there is an unknown effect in women of color, simply based on the small number of participants in this trial; again, that it appears to have an effect on ER-positive but not ER-negative breast cancers; and we would like to point out that we did not see an excess number of ER-negative cases. [Slide] As you have already heard, it is most effective against cancers that were earlier in the course of their development. You saw this information from NSABP, not in my presentation, but we also independently calculated the time to event. We did it by 6-month intervals, and there was a reduction in the number of cases diagnosed in the first 6 months and then within every 6-month block afterwards, including at the 60-month time point. [Slide] In terms of the non-invasive breast cancer endpoint, NSABP reported 59 cases on placebo and 31 on tamoxifen. When we reviewed the case report forms for these participants, 28 of these non-invasive cancers were actually diagnoses of LCIS, 21 on the placebo arm and 7 on tamoxifen. An additional 2 cases consisted of atypical hyperplasia without a component of invasive or non-invasive cancer, 1 on each arm. [Slide] When we looked at the women who had been diagnosed with LCIS during the course of the study, 12/28 women on placebo and 6/7 on tamoxifen had a prior diagnosis of LCIS as part of their eligibility criteria to enter the study. The seventh participant on the tamoxifen arm had a diagnosis of atypical lobular hyperplasia at entry. When she subsequently had her b