x A Publication ¥\\ 0f the National 1 Cancer Institute i Cancer Control Obje for the Na ’ 1985-2000 ' l ctives on: 1986 Number 2 US. DEPARTMENT OF HEALTH AND HUMAN SERVICES Public Health Service National Institutem {Health New NO! Monographs Meetings previously published under the titles Cancer Treatment Symposia and the National Cancer Institute Monograph Series will be presented under one title, NCI Monographs. The monographs will be soft bound and numbered sequentially as they appear. Material presented in the new version of the monographs will, on an alternating basis, reflect meetings and reports accepted and reviewed by either the Cancer Treatment Reports or the Journal of the National Cancer Institute Editorial Boards. If you would like to be notified when issues of the NC/ Mono- graphs are available for purchase, send the Priority Announcement Service form in the back of this issue to the US. Government Print- ing Office. Cancer Control Objectives for the Nation: 1985 - 2000 Division of Cancer Prevention and Control National Cancer Institute Editors: Peter Greenwald, M.D., Dr. P.H. Edward J. Sondlk, Ph.D. NCI Monographs National Cancer Institute, Vincent T. DeVita, Jr., Director International Cancer Information Center, Susan Molloy Hubbard, Director f” Editorial Board Peter Greenwald, Editor in Chief Elizabeth K. Weisburger, Assistant Editor in Chief Associate Editors William J. Blot Kurt W. Kohn Michael Ft. Boyd Steven M. Larson Peter Blumberg Lance A. Liotta Charles C. Brown Douglas R. Lowy Joseph W. Cullen John R. Ortaldo Charles H. Evans Jeffrey Schlom Janet W. Hartley Snorri S. Thorgeirsson George S. Johnson Jerome Yates Publications Branch, Jean Griffin Baum, Chief Monograph Editors Monograph Consultant Anne McCarthy Dee Kittner LeRoy Florence I. Gregoric Editorial Policy Manuscripts from key conferences dealing with cancer and closely related research fields, or a related group of papers on specific subjects of importance to cancer research, are considered for publication, with the understanding that they have not been published previously and are submitted exclusively to NCI Monographs. All material submitted for consideration will be subject to review, when appropriate, by at least two outside reviewers and one member of the JNCI or Cancer Treatment Reports Editorial Board. Opinions expressed by the authors are not necessarily those of the publisher or the editors. Symposia or related papers in any of the following areas should be submitted to the Editor—in-Chief, Cancer Treatment Reports: surgery, radiotherapy. chemotherapy, biologic response modification, supportive care, pharmacology, mechanisms of drug action, and medicinal chemistry. Symposia or related. papers in cancer etiology, molecular biology, prevention, control, or other areas of cancer research should be submitted to the Editor-in-Chief, JNCI. The editorial offices for both journals are located at the Ft. A. Bloch International Cancer Information Center, Building 82, Room 209, National Cancer Institute, Bethesda. MD 20892. Foreword In the fourteen years since the establishment of the National Cancer Program in 1971, much has been learned about the causes and cures of the multiple forms of cancer. The source of our progress in understand- ing cancer has been a vigorous basic, clinical, and cancer control research program. The knowledge we have gained about cancer can be used now to control a significant portion of the disease which was responsible for an estimated 462,000 deaths in 1985. This report, prepared by the Division of Cancer Prevention and Control and its advisors, is an analysis and synthesis of current knowledge about the prevention and control of cancer through life-style factors, screening, and treatment. A reduction in the cancer mortality rate of as much as 50% is possible if current recommendations regarding smoking reduction, diet changes, screening, and state-of-the-art treatment are effectively applied, and if we continue to advance cancer patient survival through improved treatment. Eliminating cancer as a major cause of morbidity and mortality will require continued research programs to further our knowledge of the causes, cures, and prevention of cancer. The National Cancer Institute remains committed to the strongest possible research agenda now and in the future. It is only through research and the effective translation and communication of research results to practitioners and the public that we can continue our progress. Vincent T. DeVita, Jr., MD. Director National Cancer Institute Preface This report stems from the work and collaboration of four Working Groups convened by the Board of Scientific Counselors of the Division of Cancer Prevention and Control at the request of the Director of the National Cancer Institute. These Working Groups were composed of experts from across the United States and Canada and the staff of the National Cancer Institute. Formed by the chairperson of the Board of Scientific Counselors, Dr. Lester Breslow, the groups included the Prevention Working Group, chaired by Dr. Philip Cole; the Screening and Detection Working Group, chaired by Dr. David Eddy: the Treatment Working Group, chaired by Dr. Paul Engstrom; and the Surveillance Working Group, chaired by Mrs. Dorothy Rice. The complete membership of each group is listed following this Preface. The purpose of these Working Groups was to assist the Institute in developing quantified objectives to serve as targets for cancer prevention, screening, and treatment and, in the associated cancer surveillance effort, to identify measures of progress and evaluation strategies. By setting specific objectives, with quanti— fied indicators for measuring progress, and by recommending actions for achieving these objectives, the Working Groups present this report as a consensus statement about the potential progress toward cancer control that can be achieved by the American public assisted by the concerted efforts of public and private agencies. The goal of a marked reduction in the cancer mortality rate by the year 2000 was chosen by the National Cancer Institute as a unifying theme for the current cancer control program. Although our major efforts will be focused toward this goal, it is also recognized that other important consequences are associated with cancer and that progress toward improvements in quality of care, quality of life, and reductions in cancer care costs also must be made. Such areas will continue to be specific foci within the Division of Cancer Prevention and Control. This report follows the initiative of the Public Health Service of the Department of Health and Human Services, which formulated an agenda for the Nation in health promotion, health protection, and disease prevention. Healthy People: The Surgeon General ’s Report on Health Promotion and Disease Prevention, published in 1979, introduced a set of major goals for improving the health of Americans through the 19805. The report called attention to 15 areas for specific programmatic achievements. A year later, in Promoting Health/Preventing Disease: Objectivesfor the Nation, 226 measurable objectives were grouped under the 15 priority areas. Since then, implementation plans have been developed so that activities toward achievement of the objectives can be advanced. The knowledge to reduce a significant portion of the cancer toll exists and must now be translated into action. Further reductions in smoking rates, changes in dietary habits, increases in breast and cervical cancer screening rates, and more widespread application of state—of—the-art cancer treatment methods must be achieved. A delay in the implementation of this knowledge will result in needless suffering and loss of life. The National Cancer Institute is eager to work with those involved in cancer control. For more informa- tion, please contact Dr. Peter Greenwald, Director of the Division of Cancer Prevention and Control, National Cancer Institute, Bethesda, Maryland 20892. vii Year 2000 Working Groups Four Working Groups were convened by the Board of Scientific Counselors of the Division of Cancer Prevention and Control to assist the National Cancer Institute in developing quantified objectives for the cancer control program. The members of these groups are listed below. The editors of this report wish to express their appreciation to these people for their insight and foresight and note that responsibility for the report in final form rests with the National Cancer In- stitute. Dr. Lester Breslow Chairman, Working Groups Chairman, Board of Scientific Counselors Division of Cancer Prevention and Control Division of Cancer Control Jonsson Comprehensive Cancer Center Los Angeles, California *********************#* Prevention Working Group Dr. Philip Cole, Chairperson School of Public Health University of Alabama Birmingham, Alabama Dr. Kaye Kilburn Department of Medicine School of Medicine University of Southern California Los Angeles, California Dr. Lloyd Kolbe Center for Health Promotion Research and Development The University of Texas Health Science Center Houston, Texas Dr. Donald Lyman Health Protection Division California Department of Health Services Sacramento, California Mr. Richard A. Merrill School of Law University of Virginia Charlottesville, Virginia Dr. Walter Mertz Beltsville Human Nutrition Center United States Department of Agriculture Beltsville, Maryland Dr. Terry Pechacek Laboratory of Physical Hygiene School of Public Health University of Minnesota Minneapolis, Minnesota Dr. Walter Willett Department of Epidemiology Harvard School of Public Health Boston, Massachusetts ix Dr. David Eddy, Chairperson Center for Health Policy Research Duke University Durham, North Carolina Dr. Charles Cook Adult Health Services Section North Carolina Department of Health Raleigh, North Carolina Dr. Anthony B. Miller Epidemiology Unit National Cancer Institute of Canada Toronto, Ontario Canada Dr. Dan Miller Preventive Medicine Institute Strang Clinic New York, NY. Dr. Paul Engstrom, Chairperson Department of Medical Oncology American Oncologic Hospital Fox Chase Cancer Center Philadelphia, Pennsylvania Dr. Paul Calabresi Department of Medicine Roger Williams General Hospital Providence, Rhode Island Dr. Charles Cobau Toledo Clinic 4235 Secor Road Toledo, Ohio Dr. Donald Hayes Department of Health and Safety Burlington Industries, Inc. Greensboro, North Carolina Dr. J. Gale Katterhagen Department of Oncology Tacoma General Hospital Tacoma, Washington X Screening and Detection Working Group Dr. Alan S. Morrison Department of Community Health Brown University Providence, Rhode Island Dr. David Sencer Administration New York City Department of Health New York, NY. Mr. Samuel Shapiro Health Services Research and Development Center School of Hygiene and Public Health Johns Hopkins University Baltimore, Maryland Dr. Barbara Threatt Institute for Social Research University of Michigan Ann Arbor, Michigan **********#*********#** Treatment Working Group Dr. Simon Kramer Department of Radiation Therapy and Nuclear Medicine Thomas Jefferson University Hospital Philadelphia, Pennsylvania Dr. LaSalle Lefall Department of Surgery Howard University Hospital Washington, DC. Dr. David Mechanic Institute for Health, Health Care Policy, and Aging Research Rutgers University New Brunswick, New Jersey Dr. Charles Spurr Piedmont Oncology Association Bowman Gray School of Medicine Winston—Salem, North Carolina Dr. Francisco Tejada Oncology Associates Miami, Florida Surveillance Working Group Mrs. Dorothy Rice, Chairperson Department of Social and Behavioral Science University of California San Francisco, California Dr. Allen Dobson Office of Research Health Care Financing Administration Department of Health and Human Services Baltimore, Maryland Dr. Jacob Feldman Analysis and Epidemiology Division National Center for Health Statistics Department of Health and Human Services Hyattsville, Maryland Mr. Lawrence Garfinkel Epidemiology and Statistics Department American Cancer Society New York, NY. Dr. Richard Warnecke Survey Research Laboratory University of Illinois Chicago, Illinois lDr. Lubitz is the alternate for Dr. Dobson. Mr. Roger Herriott Population Division Bureau of the Census United States Department of Commerce Suitland, Maryland Dr. James Lubitzl Office of Research Health Care Financing Administration Department of Health and Human Services Baltimore, Maryland Dr. Howard Ory Epidemiology Program Office Centers for Disease Control Atlanta, Georgia Dr. Carol Redmond Department of Biostatistics Graduate School of Public Health University of Pittsburgh Pittsburgh, Pennsylvania xi xii Acknowledgments In addition to the Working Groups, many members of the National Cancer Institute staff were involved in the development and review of this document. In particular, we thank the following persons for their contri- butions: Claudia Baquet Lawrence Bergner Gladys M. Block David P. Byar James F. Callahan Louis M. Carrese‘ Kenneth C. Chu Veronica L. Conley Richard D. Costlow Joseph W. Cullen William D. DeWys Brenda K. Edwards Allen Feldman Leslie G. Ford Robert W. Frelick Lillian R. Gigliotti Anne M. Hartman Jan M. Howard Carrie P. Hunter Donald C. Iverson Thomas J. Kean Elaine Lanza Luise Light Margaret E. Mattson J. Henry Montes Barbara R. Murray Lamar Neville Robert Parks Earl S. Pollack Philip C. Prorok Knut Ringen Charles R. Smart Janet L. Sobell Rosemary Yancik Jerome W. Yates John L. Young, Jr. We would also like to thank those who contributed to the computer modeling: Dr. Larry G. Kessler and Dr. David L. Levin led the effort which included contributions from Dr. David Eddy of Duke University and Dr. Mitchell Gail, Dr. Benjamin Hankey, John Horm, and Lynn Ries. Mrs. Lillian Tauber served as lead secretary in the effort and receives our most sincere thanks for a difficult job well done. We also thank the staffs of Technassociates, Inc., Bethesda, Maryland, and JRB Associates, McLean, Virginia, for their assistance. Peter Greenwald, M.D., Dr. P.H. Edward J. Sondik, Ph.D. ' Deceased, May 1986. ALL AML BCDDP CIS DCPC DHHS DHL FOBT HCFA HIP IARC NAS NCHS NCI NDI NHANES NHIS NIH Pap PDQ RT SEER SES ABBREVIATIONS acute lymphocytic leukemia acute myeloid leukemia Breast Cancer Detection and Demonstration Project Cancer Information Service Division of Cancer Prevention and Control Department of Health and Human Services diffuse histiocytic lymphoma(s) fecal occult blood test Health Care Financing Administration Health Insurance Plan (New York) International Agency for Research on Cancer National Academy of Sciences National Center for Health Statistics National Cancer Institute National Death Index National Health and Nutrition Examination Survey National Health Interview Survey(s) National Institutes of Health Papanicolaou (smear) Physician Data Query radiotherapy Surveillance, Epidemiology, and End Results (Program) socioeconomic status xiii TABLE OF CONTENTS Page Foreword v Preface vii Year 2000 Working Groups ix Acknowledgments xii Abbreviations xiii l. Cancer Control Objectives 3 ll. Prevention of Cancer 15 III. Screening 27 IV. Treatment 35 V. Surveillance 45 VI. References 53 Appendix A: Use of Models to Project Cancer Mortality in the Year 2000 59 Appendix B: Analysis of the Impact of the Cancer Control Objectives 69 *************** A Computer-based Model for Designing Cancer Control Strategies 75 David M. Eddy A Model for Projecting Cancer Incidence and Mortality in the Presence of Prevention, Screening, and Treatment Programs 83 David L. Levin, Mitchell H. Gail, Larry G. Kass/er, and David M. Eddy 1. Cancer Control Objectives I. Cancer Control Objectives With this report, the NCI is establishing a set of quanti- fied objectives to be used in charting a course to reduce significantly the annual cancer mortality rate. As outlined in this report, a reduction in the mortality rate of 25% to 50% from the 1980 level is possible through full and rapid application of existing knowledge of cancer prevention, screening and detection, and state-of-the-art treatment methods. The NCI has set a goal of a 50% reduction in the cancer mortality rate by the year 2000. The achieve- ment of this goal depends on a reduction in tobacco smoking by 50% from 1980 levels, the adoption of a prudent low fat, high fiber diet by all Americans, recom- mended cancer screening measures, and accelerated and widespread application of gains in state—of-the-art cancer a treatment methods. A set of specific cancer control objectives has been chosen as targets to help guide the Nation’s cancer control program toward this overall goal. Linked to the objectives is a set of cancer control indicators that NCI will use to measure progress toward achievement of the objectives. These objectives, with adjustments to reflect specific regional cancer rates and cancer control problems, can also be used to direct the development of, and serve as objectives for, regional cancer control efforts across the Nation. The cancer control objectives outlined in this report incorporate and expand the disease prevention and health promotion objectives for the year 1990 adopted by the DHHS (I). In addition, the recommended actions are grouped similarly as an aid in the coordination of health promotion-related activities across the country. A summary of the major cancer control objectives is given in table I-l. Intermediate objectives, the rationale supporting the objectives, and recommended actions are presented in subsequent chapters that address in detail the topics of prevention, screening, treatment, and surveil- lance. This chapter provides an overview of the objectives, their potential impact, and a summary of possible cancer control actions, as well as potential roles for various health care agencies and organizations. FEASIBILITY OF ACHIEVING THE OBJECTIVES The current knowledge about cancer and the existing network of cancer control resources form the basis for an aggressive effort to be launched now by all concerned to control cancer. Since its inception, the National Cancer Program has emphasized research 1) to elucidate the causes of cancer; 2) to improve by therapy the prognosis for cancer patients; and 3) to control cancer, which in- cludes cancer prevention trials, smoking prevention and - cessation, demonstration, and education programs. Other programs are designed to prevent, detect, diagnose, and treat cancer and to rehabilitate and counsel cancer patients. Major gains have been made during the past decade in the understanding of cancer etiology and in development of diagnostic techniques and more effective treatment. Furthermore, the cancer control research base is now in place to improve, evaluate, and accelerate these cancer control activities. With the establishment in De- cember 1983 of the DCPC at NCI, research, development, and application programs to transfer the results of research to practice and prevention have been given new emphasis. In addition, there has been considerable progress in build- ing a network of resources that are now involved in, or have the potential to contribute to, the Nation’s cancer control effort. This section highlights major facts and concepts about the most important risk factors for cancer, the potential of cancer screening to reduce mortality, the gains in treatment that are now reflected in increased survival, and patterns of cancer occurrence that reinforce the evidence that cancer mortality can be significantly reduced. Major Causes of Cancer In 1985, it was estimated that 910,000 Americans would develop cancer and 462,000 Americans would die of the disease within the year. Yet much of cancer incidence can be prevented through changes in our smoking and dietary habits. The scientific evidence for smoking as a cause of cancer has been recognized for over 20 years. The evidence for diet has emerged over the past decade and has pro- gressed to the extent that recommendations for prudent change can now be made. Smoking About 30% of all cancer deaths (over 130,000 deaths per year) are related to smoking. Today, 54 million Americans, i.e., about 1 in every 3 adults, smoke cigarettes daily, and those who smoke 2 or more packs daily have lung cancer mortality rates 20 times higher than do non- smokers. Since 1953, lung cancer rates have increased 172% among men and 256% among women. In 1986, lung cancer may exceed breast cancer as the leading cause of cancer death among women in the United States. Ciga- rette smoking is further associated with cancers of the larynx, head and neck, esophagus, bladder, kidney, pan- creas, and stomach. However, these facts are tempered by the knowledge that 1) cancer risk returns to near normal (the risk of smokers is one to two times greater than that of nonsmokers) within 15 years after smokers have stopped 3 TABLE l-l.—Cancer control objectives: Summary Control Action Target Rationale Year 2000 objectives Prevention Screening Treatment Smoking Diet Breast Cervix Transfer of research results to practice The causal relationship between smoking and cancer has been scientifically established. Research indicates that high—fat and low-fiber consumption may increase the risk for various cancers. In 1983, NAS reviewed research on diet and cancer and recommended a reduction in fat; more recent studies lead NC] to recommend an increase in fiber. Research is under way to verify the causal relationships and to test the impact on cancer incidence. The effectiveness of breast screening in reducing mortality has been scientifically established. The effectiveness of cervical screening in reducing mortality has been scientifically established. Review by the NCI of clinical trial and SEER Program data indicates that, for certain cancer sites, mortality as shown by SEER data is greater than that experienced in clinical trials. Reduce the percentage of adults who smoke from 34% (in 1983) to 15% or less. Reduce the percentage of youths who smoke by age 20 from 36% (in 1983) to 15% or less. Reduce average consumption of fat from 37% 738% to 30% or less of total calories. Increase average consumption of fiber from 8—12 g to 20—30 g/day. Increase the percentage of women aged 50—70 who have an annual physical breast examination coupled with mammography to 80% from 45% for physical examination alone and 15% for mammography. Increase the percentage of women who have a Pap smear every 3 years to 90% from 79% (ages 20739) and to 80% from 57% (ages 40—70). Increase adoption of state-of-the-art treatment. (See table IV-I for intermediate objectives, specific percentages, and cancer sites.) for all smoking—induced cancers, and 2) this benefit ac- crues even after many years of smoking. Progress has been made in the reduction of the percent- age of adult smokers since the 1964 Surgeon General’s report on smoking and health. In 1965, 52.1% of men were smokers; in 1983, the figure was 34.8%. In 1965, 34.2% of women were smokers; in 1983, the figure was 29.5%, which represents a decrease, but not as great as that shown for men (fig. [-1) In fact, women must ac- celerate their decline in smoking or face more lung cancer than men at the start of the twenty—first century. White 'males recently experienced a decline in lung cancer in- cidence (the first such decline in 50 yr), which is a strong indicator that reduced smoking is having a significant effect. However, smoking and the use of smokeless tobacco continues to be a severe problem and an alarming trend for our youth. More than 100,000 youths aged 12 years and under are habitual smokers, and more begin smoking every day. These figures present both a promise that declines are possible and a challenge that we should reinforce and accelerate the decline into a consistent trend in smoking reduction and also stop smoking and other tobacco use among our Nation’s youth. Dletary Components In an extensive review of cancer causes, Doll and Peto (2) estimated that 35% of cancer deaths may be related 4 to dietary components, with the possible range of effect being 10% to 70%. A mid-range estimate of 35% would mean that about 150,000 lives could be saved annually through dietary changes. These estimates cannot be con- sidered definitive because the research to test the effective- ness of dietary interventions to reduce cancer incidence and mortality for particular cancer sites is only now under way. Nevertheless, even the most conservative estimate represents a potentially significant impact on cancer mor- tality. Meanwhile, a consensus is growing that certain changes in diet are prudent at this time because they may be important in reducing cancer incidence. A committee of the National Research Council has recommended a change in the typical American diet that may lower the risk of cancer. Specifically, they recommend 1) the con- sumption of fat should be reduced to 30% of total calories, and 2) fruits, vegetables, and whole grain cereal products should be eaten daily (3). The NCI concurs and recom- mends a diet low in fat and high in fiber-rich foods, fruits, and vegetables (4). In addition, a recent NIH Consensus Conference on cholesterol and heart disease noted that a diet low in fat is important for reducing heart disease (5); thus a reduction in fat consumption might simultaneously lower the risk of cancer and heart disease. On the basis of current knowledge alone, the NCI estimates that, at a minimum, 30,000 lives could be saved in the year 2000 if Americans would modify their dietary habits. If the clinical trials currently being conducted NCl MONOGRAPHS, NUMBER 2, 1986 70 60— m5 33.2% Daring J_......_I_-ii_f_ _._';_5;;:, 30 _ 137%Decllne _ 1'33 H ' .. PERCENT 20 — O Males CI Females 10 — 0 l 1 l 1965 1976 1980 1983 YEAR FIGURE I-I.APrevalence of cigarette smoking in the United States; adults were 20 yr of age or older in 1965, 1976, 1980, and 1983. Data were obtained from the Health Interview Surveys of the NCHS. show positive results over the next few years, the prospects of a decline in cancer incidence by changes in diet or use of chemopreventive agents will be brightened. Thus it is possible that current research in chemoprevention, diet, and nutrition could result in the prevention of a higher number of cases of cancer and subsequent deaths, perhaps as many as the estimated 35% or more of all cancer deaths possibly related to diet. Advances in Cancer Screening and Treatment Screening Statistics indicate that for most cancers detection and treatment in the early stage afford a much greater chance of patient survival than do detection and treatment at later stages of the disease. Consequently, cancer mortality for breast and cervical cancers can be greatly reduced through aggressive screening. The research to date on screening for colorectal cancer is not considered definitive by the NCI, although many individuals and groups, in- cluding the American Cancer Society, recommend such screening. The screening objectives posed in this report relate to screening techniques for which there is scientific consensus of proven effectiveness, i.e., those for breast and cervical cancer. Breast cancer accounts for about 18% of all cancer deaths among women, and recent data show that the age-adjusted incidence and mortality rates from breast cancer in women have not changed during the past decade (4). Yet the evidence from a long—term clinical study of over 60,000 women enrolled in the HIP of New York shows that breast cancer mortality is reduced 30% in CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 women over age 50 who are screened for breast cancer by mammography and physical examination (6). A more recent study conducted in Sweden duplicates the 30% mortality reduction found in the HIP study, but screening was less frequent (7). The fact that breast cancer mortality for women over 50 years of age has not decreased may reflect the fact that many women are not taking advantage of screening. As recently as 1983, according to a survey performed by the American Cancer Society, only 15% of the women over 50 reported receiving an annual mam- mogram. Cervical cancer screening has long been known to be effective. Use of the Pap smear can reduce the risk of mortality from invasive cervical cancer by as much as 75% according to a recent study (8). This figure, coupled with current screening figures, yields the estimate that use of the Pap smear for screening could reduce mortality from cervical cancer in the United States by at least 25%. Surveys indicate that at most 79% of the women 20 to 39 years old and 57% of those 40 to 70 (for whom the risk of cervical cancer is now greatest) follow recommended guidelines on screening for cervical cancer. Treatment The most recent figures from the SEER Program1 in- dicate that the likelihood of a person’s surviving cancer for at least 5 years from the point of detection, compared with the survival of the general population, is now over 49% for patients diagnosed in 1976 through 1981, compared with 48% for those diagnosed in 1973—75 and an estimated 38% for patients diagnosed in 1960—63 (9). These figures do not take into account the increased incidence of lung cancer; if lung cancer incidence is removed from the most recent figures, the chance of survival for more than 5 years, which for most cancer sites indicates a cure, is 56%. The figures show a steady gain in survival, and, for some cancers, the gains have been dramatic. In 1960, only 40% of the patients with Hodgkin’s disease survived for more than 5 years; the latest SEER figures show the rate to be 74%, an increase attributable to improvements in radio- therapy and chemotherapy. Similarly, survival from tes— ticular cancer has increased from 76% in 1973—75 to 87% among patients diagnosed between 1977 and 1981 because of significant advances in treatment. Of extreme impor- tance is the recent conclusion of an NIH Cancer Consensus Development Panel that breast cancer mortality can be reduced through adjuvant (postsurgical) therapies of che— motherapy in premenopausal women and hormonal (ta- moxifen) therapy in postmenopausal women. In both in- stances, the treatments are recommended for those women with positive lymph node involvement, and, for postmeno- pausal women, positive estrogen receptors (10). The survival rates for a number of cancer sites are shown in figure I-2. Although for many particular sites the increase in survival is small, a steady trend toward an overall increase in cancer survival during past decades is apparent. Analysis of data from the SEER Program in- dicates the rate of increase in the overall cancer survival 1This program monitors cancer incidence and survival in defined population areas comprising about 12% of the United States population. 100 80 E a 43 a 40 20 0 Colon Breast (females) Childhood Cancer 100 100 80 80 E § 60 60 53 G) n. 40 40 20 20 0 0 Prostate Uterine Corpus Bladder 100 100 100 80 80 80 E 3 so 60 60 E 40 31 4o 40 37 20 20 14 20 0 0 0 Non-Hodgkin’s Leukemia Kidney Lymphoma 1 100 00 100 87 so so 74 80 E 63 60 g 60 60 so a. 40 4o 40 4o 20 20 20 0 0 ‘ 0 ‘ Melanoma Hodgkin’s disease Testicular of the skin 1960 - 1963 D 1970 - 1973 1973 - 1975 1976 - 1981 m FIGURE I—2.—Five-yr relative survival rates for selected sites for white cancer patients. Five-yr survival rate is the probability that a person will escape death due to cancer for 5 yr following diagnosis. Rates for 1960—63 and 1970—73 are based on data from hospital registries and one population-based registry; those for 1973—75 and 197678! are from the SEER Program. Rates are based on the follow-up of patients through 1982. Although the data are not strictly comparable, they are indicative of change. rate is about 0.5% annually. The reasons for this steady improvement are gains primarily in treatment efficacy, earlier detection, or both, for a number of cancers includ- ing testicular cancer, melanoma, and Hodgkin’s disease. For other sites, such as the prostate, the increased survival is not fully understood but may be influenced strongly by the increased detection of asymptomatic disease that might otherwise have remained undiagnosed.2 6 2In regard to prostate cancer, both incidence and survival have increased, whereas mortality has remained relatively con- stant. Therefore, it appears that some portion of the increased survival may be due to the increased detection and treatment of early stage, less aggressive disease. Data from the SEER Pro- gram for distant (metastatic) prostate cancer, however, show some improvement in survival, but it is possible that this im- provement reflects changing definitions for distant disease. NCl MONOGRAPHS, NUMBER 2, 1986 The NCI believes that accelerated dissemination and application of existing state-of-the-art cancer therapy will significantly increase national cancer survival rates. When current cancer survival figures from the SEER Program are compared with estimated survival figures if state-of- the-art treatment were applied, the differences in survival translate into a reduction in the cancer mortality rate on the order of 10% to 15%. In addition to the gains in survival reasonably expected with aggressive application of state-of-the-art cancer treatment, further advances in research are expected. If these trends in improved survival also are taken into account and estimated through the year 2000, then the reduction in mortality from increased application of state-of-the-art treatment could reach 25%. Patterns of Cancer Occurrence and Survival LThe prospect of surviving cancer is not the same for all Americans. This fact emphasizes the potential for gains in survival through application of state-of—the-art treatment information. Survival differences among ethnic groups have been observed and, most importantly, analyses have linked differences in survival to differences in SES. Pa- tients who reside in areas with generally lower SES char- acteristics have lower survival figures for the cancer sites studied to date. If differences in survival by SES are, in fact, related to differences in patients’ access to the health care system early in their disease or to state-of-the-art treatment, then the potential for mortality reduction would be even greater than now projected. Estimates of survival by race show that blacks have a lower chance of surv' ' ; cancer 5 years than do whites, i.e., 38% versus 50%@'reliminary analyses show that much of the dif- feren an be explained by SES; on the whole, black and white Americans with low incomes have a poorer prognosis from cancer those whose incomes are above the median level The health care community is chal- lenged to dev . stems that enable both professionals to practice at state-of—the-art technology levels and patients to have access to state-of—the-art treatment and screening. Further evidence of differences among population groups has been gained during the past 15 to 20 years because it has become apparent that substantial segments of the population (e.g., Mormons and Seventh-Day Ad- ventists) have cancer death rates far below those of the general population. For example, mortality from colorec- tal cancer among Californian Mormons is 70% (men) and 78% (women) that of the general white population. Fur- thermore, for Mormons, the incidence of cancers of the lung, larynx, tongue, gum and mouth, esophagus, and bladder is 55% lower than for the United States popula- tion (12). Also, particular counties in the United States have cancer mortality at a considerably lower rate than does the general population. As another illustration of wide differences in cancer mortality rates (adjusted for age and sex differences), about 10% of our counties have less than 71% of the average national rate. In these counties and according to age—adjusted mortality rates from 1950 to 1969, the mortality for breast cancer in women is 13.4 deaths per 100,000, compared with 25.5 deaths per 100,000 women in the United States. In addition, mor- tality in these counties for lung cancer in men is 17.5 CANCER CONTROL OBJECTIVES FOR THE NATION: 1985—2000 deaths per 100,000 persons, compared with 38.0 deaths r 100,000 in the United States (13). In the United States, many ethnic minorities have a cancer experience worse than that of whites. The excep- tions to this are certain segments of our Asian population who have better overall cancer survival and better survival rates for certain cancer sites than do whitesEAddressing the cancer needs of minorities is critical to our achieve- ment of the goal of reducing cancer mortality by 50% by the year 2000. The cancer incidence and mortality excesses and poorer survival experience of certain minorities, par- ticularly blacks, require aggressive and long-ter “catch- up” efforts if we are to achieve the overall goal. The most extensive cancer d ta available on minorities in this country are for blacks. Statistics show that blacks have higher age-adjusted incidence and mortality rates for many cancers and lower survival r than do whites for all but 3 of 25 primary cancer site able 1-2 presents a summary of the incidence, mor , and survival for the major cancer sites for blacks and whites. (All d' fer- ences are significant at the .05 level except as noted.) Figure [-3 outlines in graphic form the 5-year relative survival rates for blacks and whites by specific cancer site. It is evident that blacks experience high rates of most smoking-related cancers. Many of these (e.g., cancers of the lung, esophagus, and pancreas) have high case fatality rates in all population groups, which suggests that control of these diseases can best be accomplished through prevention and cessation of tobacco use. Cancers for which blacks show poorer survival rates compared with whites include some types that occur less frequently in blacks, i.e., cancers of the bladder and corpus uteri. Although every aspect of the cancer survival difference between blacks and whites is not completely understood, there is evidence that socioeconomic factors influence the survival differences. It is known that SES is associated with a variety of factors, including host factors (e.g., resistance to disease and nutritional status) that may TABLE l-2.—Summary of incidence and mortality rates and 5-yr survival for major cancer sites, for blacks and whites Incidence‘1 Mortality” Survival, %‘ Site / type White Black White Black White Black Esophagus 2.9 11.5 2.6 9.2 5 3 Stomach 8.0 13.8 5.8 10.0 14 15d Rectum 15.0 11.7 3.5 3.54 49 37 Pancreas 8.9 13.6 8.6 11.0 3 3" Larynx 4.6 6.6 1.3 2.5 67 59 Male lung 81.0 119.0 70.7 91.4 13 10 Breast 85.6 71.9 26.6 26.3" 75 63 Cervix 8.8 20.2 3.2 8.8 68 63 Corpus uteri 25.1 13.4 2.0 2.9 88 57 D m 120.; as. 4.32 .62 a Bladder 15.4 8.6 3.9 3.8“ 74 50 Multiple myeloma 3.4 7.9 2.4 5.0 24 27" " Incidence is given in No. of cases per 100,000 persons for 1978A81, age-adjusted to the 1970 United States population. b Mortality is given in No. of deaths per 100,000 persons for 1978—81, age-adjusted to the 1970 United States population. ‘ Survival figures are for 1973—81. ” Black~white differences are not significant at the .05 level. WHITE PATIENTS TESTIS MELANOMA EMALE BREAST BLADDER HODG KIN’S PROSTATE CERVIX UTERI LARYNX COLON KIDN EY RECTUM LEUKEMIA ,2 MYELOMA —_27 9 .‘CALLBLADDER {22:10 5 ESOPHACUS ——g 3 4 LIVER __g2 3 B— PANCREAS —fi3 E influence cancer development and treatment response. Furthermore, unequal distribution of health facilities and trained health personnel may result in poor accessibility to cancer screening, detection, and treatment among cer- tain segments of the population. Also affected by SES are educational attainment and thus occupation. These in turn may affect exposure rates to occupational carcin- ogens. For example, studies of cancer risk in the steel and rubber industries revealed that blacks work in the most hazardous (in regard to toxic and carcinogenic exposures) worksites. Certain minorities, including blacks, are dis- proportionately represented in the lower SES categories (14). A person’s lack of knowledge may result in patterns of delay in seeking diagnosis and treatment and may result in lower cancer survival outcomes. Some studies among blacks indicate an average of 3 to 12 months delay in their seeking treatment after a diagnosis of cancer has been made. A person’s knowledge and attitudes also influence anticancer risk behaviors (e.g., being a nonsmoker, per- forming breast self-examination, and seeking routine Pap smears) as well as compliance with treatment regimens (15). A possible factor in surviving cancer is one’s general health status. Low-income persons tend to have poorer health and a shorter life expectancy than do affluent persons. Therefore, public health efforts aimed at the improvement of general health status for all ma also address and improve cancer patient survival (16). j Strength of the Cancer Control Network The ability of the specialists to reduce cancer mortality depends partly on the existence and application of a num- ber of types of resources. 1) The means to provide infor- mation on prevention, screening, and treatment to other public and health care professionals is essential. 2) There must be mechanisms or systems for providing patients access to state-of-the-art cancer treatment and, as appro- priate, encouraging their participation in cancer clinical trials. The mechanism for maintaining continued research progress and for fostering new research is also essential. These needs can be met in part with the cancer control network now in place. The current network has the orga- nizational and personnel capabilities for various cancer 8 BLACK PATIENTS FIGURE I-3.—Five—yr relative survival rates, %, for selected cancer sites, white and black patients, SEER Program data, 1973—81. interventions (table I-3). A network of cancer research centers has been developed, although some of the areas of the country remain underserved. Programs are being de- veloped and evaluated that will increase the pace of clinical research and bring the benefits of clinical research to communities (the Community Clinical Oncology Program and the Cooperative Group Outreach Program). A com- puterized information system for physicians (PDQ) with information on cancer treatment has been introduced and is available to the clinician through the National Library of Medicine or through commercial information systems. The capability of investigators to monitor progress in can- cer control has been increased through the expansion of the SEER Program, which tracks cancer incidence and mortality in 12% of the United States population. The CIS maintains a nationwide toll-free telephone network for providing immediate answers to cancer-related questions from cancer patients, their families, the general public, and health professionals. All these resources, when taken together, form the framework of a network that speeds the results of cancer research to cancer control application and tracks their impact. Despite the extent of these resources, they alone are not sufficient to reduce cancer mortality. Gaps in the network exist, and it is imperative that these gaps in information transfer, practice patterns, research capability, etc., be recognized and filled to meet local cancer control needs. This and the succeeding chapters of this report contain suggestions as to how the local cancer control network can be completed or improved to progress toward the changes that can reduce the cancer toll. POTENTIAL IMPACT OF THE OBJECTIVES The Impact Though the knowledge and resources for some impor- tant cancer control measures are available, to date they are not being fully applied. The NCI estimates indicate that full aggressive application of present knowledge plus a reduction in smoking to one-half the 1980 levels by 1990 and a continued advance in cancer survival could result in a 50% reduction of cancer mortality by the year 2000. This report identifies those aspects of cancer control for which a fuller application and impact can be realized and NCl MONOGRAPHS, NUMBER 2, I986 TABLE I-3.—Netwark of cancer control resources Community Clinical Oncology Program A major research initiative intended to involve community physicians in clinical trials according to NCl-approved research protocols, thereby increasing use and transfer of state-of—the-art therapy. Cooperative Group Outreach Program A program to upgrade the skills of community physicians and other health professionals in treating cancer patients through use of well- defined protocols at affiliated hospitals, quality control of patient management, and continuing education activities Cancer Centers Program A program to coordinate grants for the support of projects in cancer research, education, and control at educational and research institutions in the United States. The purpose is to extend knowledge and understanding of the causes, mechanisms, prevention, detection, diagnosis, and treatment of the multiple forms of cancer through the development of either specialized or broad multidisciplinary programs in basic and clinical cancer research. Clinical Cooperative Groups A program of several multi—institutional consortia of clinical investigators to conduct clinical trials of cancer treatment. Physician Data Query A computerized information system about cancer treatment and research protocols designed for physicians, accessible by personal computers through the National Library of Medicine and through vendors of commercial information systems. Surveillance, Epidemiology, and End Results Program The major component of NCI’s system for tracking cancer incidence, patient survival, and mortality, consisting of 11 population-based cancer registries serving 12% of the nation’s population. Cancer Information System A national, toll-free telephone service providing immediate access to answers to cancer-related questions from cancer patients, their families, the general public, and health professionals. presents objectives to direct and measure progress toward this full application. The purpose is identification of the actions and coordination necessary for a decrease in can- cer mortality during the next 15 years. By a coupling of data on past trends with assumptions on the impact of specific changes, one can outline the potential contributions of achieving the prevention, screening, and treatment objectives to the reduction of cancer mortality. Table I—4 summarizes the objectives and the total percentage reductions for each control area. The specific methods and assumptions are detailed in the Ap- pendixes and the two papers following them.3 The model, developed by NCI staff, was based on a comprehensive model of cancer control impact developed by Dr. David Eddy of Duke University. Two parameters are prominent in the analysis. The first is the level of reduction achieved in adult smoking prevalence, of critical importance be- cause smoking is responsible for 30% of all cancer mor- tality and the impact of a change in smoking prevalence would be observed within a 10-year period. The second parameter is the degree to which advances in cancer treat- ment will continue and improve the prospects of cancer survival. Table [-4 shows the percentage change in the projected age-adjusted cancer mortality rate achieved by the year 2000 under various assumptions about smoking reduction and the annual percentage gain in survival of cancer patients. With an aggressive smoking reduction pro- gram and with an increase in cancer survival after diag- nosis for all major cancers of 1.5% each year over the pre- vious year, the age-adjusted cancer mortality rate for the year 2000 can be reduced by 50% from the 1980 age-ad- justed mortality rate projected to the year 2000. Prlnclples for Formulating the Objectives The NCI adopted several principles in formulating the objectives. 1) It was believed that a single overall goal was 3See last two companion papers by Eddy and Levin et al. in this monograph. TABLE I-4.—Estimated reduction by year 2000 in cancer mortality rate“ Objective Method Estimated reduction, % Prevention Diet 8 Fat reduction to 25% of total calories Fiber increase to 20730 g/ day ' Smoking Reduction in adult smoking prevalence to 16% If achieved in year 2000 8 If achieved in year 1990 15 Screening” Achievement of objectives in table H 3 Treatment Application of current state-of—the-art treatment for specific cancer sites (table IV-l) With no future changes in state-of-the-art therapy 10 With current trend in state-of-the-art survival (0.5% per yr, all sites) 14 maintained With accelerated gains in state-of-the-art treatment (1.5% per yr, all sites) 26 Total range of mortality reduction, % 25—50‘ ” Reduction is calculated from the projected rate for the year 2000 and is based on achievement of the objectives. All rates are age-adjusted to 1980. b Only females were considered for screening. " Range accounts for interdependence of objectives, e.g., the effect of breast cancer screening is reduced due to prevention. CANCER CONTROL OBJECTIVES FOR THE NATION: 1985 2000 necessary to serve as a unified target for cancer control, a goal that could be directly linked to gains achieved in prevention, screening, and treatment. As noted, this goal is to reduce the cancer mortality rate by 50% of current projections by the year 2000. These current projections take into account the increases in cancer incidence primarily in the smoking-related cancers, particularly lung cancer, observed over the past 30 years. Even if all smokers were to cease smoking today, lung cancer rates would likely continue to rise from the cumulative effects of past smoking. However, as shown in Appendix A, the impact of smoking cessation is dramatic and can lower the risk of lung cancer by at least one-half within 10 years after cessation (17). 2) The baseline for developing NCI’s objectives was the reduction of overall cancer mortality rates, age-adjusted to the year 1980.4 Mortality rates are a better yardstick of progress than the total number of deaths because the population is growing and aging. These population factors will almost certainly mean more actual cancer deaths as time passes, which would mask progress achieved in the reduction of the cancer mortality rates. 3) The objectives were based on the data and trends at hand. For example, the impact of future changes in cigarette smoking habits were extrapolated from existing data for estimates of the potential future effect by methods outlined in this monograph. 4) Trends in survival gains due to treatment were to be considered and alternative trends assessed. Although both major breakthroughs and advances in treatment are pos- sible, such breakthroughs were not anticipated when trends were predicted. 5) The analysis of the impact on the United States of meeting the objectives reflects solely the impact on cancer mortality rates. Other benefits would most likely accrue, such as a decrease in heart disease from reduced cigarette smoking. The analysis does not address the economic benefits or costs of achieving these objectives because the cost of cancer control interventions will vary in response to various forces in the health care system. For example, the cost of mammography varies widely today, from as little as $25 to over $200 per examination. It is clear that changes in economic market forces, the organization of health care services, and technology could change con- siderably the cost of implementing mammography. The analytic effort toward assessment of the potential impact of meeting the objectives was designed to identify the potential benefits (reduced mortality) that would accrue. Issues of identification of the least costly ways by which these objectives can be met will be addressed as part of the NCI Cancer Control Research Program and through the efforts of other organizations and state and local govern- ments. SUMMARY OF RECOMMENDED ACTIONS In addition to developing objectives, the working groups and NCI staff developed recommendations for actions 4Age-adjustment to a base year weights sex- and age-specific rates according to a population distribution for the base year; therefore, changes in the age or sex distributions of the popula- tion do not affect comparisons over time. IO aimed at preventing cancer or detecting it in earlier stages. The result of such actions should be a reduction not only in mortality but also in morbidity, inasmuch as cancers prevented generate no cancer morbidity, and cancers de- tected earlier should, generally, be more successfully treated. In addition, because the purpose of the treatment recommendations is to improve the speed and extent of bringing state-of—the-art treatment to the practice setting, the NCI believes that such recommendations would apply to rehabilitation and continuing care as well. Recommended Acllons tor Intervening Agents The specific actions recommended for prevention, screening, treatment, and surveillance are listed in the respective sections of this report. Here they are broadly summarized for all the cancer control topic areas accord- ing to the type of intervening agent that would seem ap- propriate: the NCI, other federal agencies, state agencies, local government, private industry, professional organi- zations, voluntary organizations, and the media. These potential intervenors are being encouraged by the NCI to accelerate and intensify their actions toward cancer control as part of the National Cancer Program and, especially, as part of the effort to achieve the year 2000 goal. National Cancer Institute The role ofthe NCI in cancer control is threefold: 1) guide and support basic and applied research in cancer preven- tion, screening, diagnosis, treatment, and public educa- tion; 2) provide information and technical assistance to other agencies and organizations interested in conducting cancer control activities; 3) conduct public and profes- sional education programs. The institute should continue and intensify these efforts, especially for proven screening and diagnostic techniques, treatments, and prevention in- terventions relevant to smoking, diet, and occupational exposures. Other Federal Agencles Federal agencies are encouraged to sponsor appropriate research and data collection, publicize information na- tionally, offer technical assistance to program planners, promote and participate in collaborative efforts, and to set and enforce appropriate regulatory measures. State Agencies State agencies are encouraged to integrate cancer con- trol techniques into their health care delivery and health promotion programs, develop survey and surveillance capabilities, collaborate on an interstate basis in forming coalitions and implementing programs, coordinate pro- gram planning among the numerous state agencies (agri- culture, environmental protection, health, aging, etc.), and develop policies that promote cancer prevention. Local Government Local governments are encouraged to promote health education and cancer prevention education in schools, develop local health promotion programs and coalitions, provide technical assistance to community organizations that are planning health promotion programs, cooperate with state and federal agencies to provide survey data or capability, and to offer screening programs through local health clinics. NCI MONOGRAPHS. NUMBER 2, 1986 Private Industry Private industry is encouraged to offer health promotion programs and screening programs to employees, collab- orate with employee groups to promote worksite health promotion programs, monitor employee use of measures to prevent exposure to carcinogens in the workplace, offer in their food service units choices consistent with guidelines for cancer prevention, and to develop insurance policies that reward risk—avoidance behavior. Food production industries are encouraged to offer healthful food products to the marketplace. Prolesslonal Organizatlons Professional organizations are encouraged to incor- porate cancer control knowledge into basic training cur- ricula and continuing education, increase questions about cancer prevention and control on licensure examinations, include more articles in journals about cancer control and the role of the health professional, counsel patients about prevention steps they can take for themselves, and to provide assistance to other organizations and agencies developing cancer control programs. Voluntary Organizations Voluntary organizations are encouraged to increase their offerings of health education and screening programs at the community level, form coalitions for cancer control, and to collaborate with federal agencies in disseminating health information and materials nationally, especially to those populations who are difficult to reach. Media The media are encouraged to increase coverage about cancer causes, prevention, and control (especially about tobacco and diet) and communicate more closely with scientists and health professionals. Survelllance Actions The Surveillance Working Group and staff developed a set of recommended actions to track the progress of cancer control as well as the set of indicators (table 1-5) to measure progress toward achieving the objectives. All indicators are to be indexed by sex, race, age, and geo- graphic region. To track these indicators, the Surveillance Working Group also recommended that the NCI implement the fol- lowing actions systematically: 1) Identify sources of baseline information for all in- dicators in the objectives. 2) Establish cooperative working arrangements for data collection from all sources. CANCER CONTROL OBJECTIVES FOR THE NATION: 1985—2000 TABLE 1-5.#Factors and indicators to be used in measurement of progress Factors Indicators Mortality Deaths per 100,000 persons, total deaths Incidence Cases per 100,000 persons, estimated total cases Smoking Smokers (adult and youth), %; years since quitting Diet Fat and fiber in the diet, %; obesity, % Screening Eligible persons screened, % Treatment Distribution by stage of cancer at diagno- sis, cancer patients treated by state-of— the-art methods, cancer patient survival, resources Occupation Workers exposed, %; workers screened in the workplace, % Knowledge, attitudes, and Risk factor-related profiles among the beliefs general population, ethnic groups, and health professionals Costs Indirect and direct 3) Monitor and report annually on changes in the indi- cators through the year 2000. 4) Evaluate progress in detail toward the objectives in years 1990, 1995, and 2000. CONCLUSION This report provides the background for the NCI pro- jection that a significant fraction of cancer mortality can be eliminated during the next 15 years with widespread aggressive application of existing knowledge and expected advances in knowledge. Achievement of this potential de- pends on the effective coordination of cancer control ac— tivities at the federal, state, and local levels and the active participation of volunteer and professional organizations. The actions outlined in the chapters of this report are only the starting points. Public debate and discussion will lead to the formulation of effective strategies to achieve this reduction of cancer mortality. The NC] will continue to focus its research program on the causes, prevention, treatment, and control of cancer. In addition, it will monitor, through a comprehensive surveillance plan, the progress of cancer control and make reports on a periodic basis. The surveillance results will be used to update objectives and, as appropriate, redirect the cancer control resources of the NCI. The Institute welcomes collaborative efforts with government, industry, voluntary organizations, professional organizations, and the media to achieve the full measure of reduction of cancer mortality that is now possible. I] 11. Prevention of Cancer ll. Prevention of Cancer RISK FACTORS FOR CANCER For several decades, it has become increasingly apparent that most cancer is caused or promoted by life-style and environmental factors and that only a fraction of cancer is completely genetic in origin. Some genetic or susceptibility factor and chance must be involved because most people similarly exposed to environmental factors do not develop cancer. Today, however, research data support the esti- mate that life-style and environmental factors are related to the development of roughly 90% of cancer incidence and, therefore, support the conclusion that, in principle, most cancer is preventable. In this report, the prevention component of cancer control refers to the lowering of cancer incidence and, therefore, cancer mortality, by changes in those life—style and environmental factors that influence the occurrence and progression of cancer. Types of Risk Factors Three types of factors, individually or in combination, increase an individual’s risk of developing cancer, i.e., life- style, environmental, and genetic factors. For convenience, life-style and environment are treated as separate factors, although life-style may be considered a subset of envi- ronment. Life-style factors are behaviors over which the in— dividual has some control, especially tobacco use, diet, al- cohol, excessive exposure to sunlight, sexual behavior patterns, and general personal hygiene. Environmental factors include both occupational exposure to carcinogens and exposure to carcinogens and radiation in medical pro- cedures, as well as factors that are naturally occurring or man-made causes of cancer that contaminate water, air, and earth. These factors are largely beyond an individual’s control and thus require broad social actions or system changes to achieve effective control. Genetic factors are conditions inherited at conception. Control of these fac- tors is at present largely not feasible for cancer except through genetic counseling activities. These categories are neither absolute nor independent because the relationships between the individual and so— ciety are complex and changing. For example, clean air is primarily a social concern, but some might argue that per- sonal behavior plays a role in cancer prevention through choice of place of residence or by participation in the po- litical process. On the other hand, smoking is a behavior of personal choice, although social elements influence that choice. Whereas workers may be able to influence the conditions of their occupational environments, this ap- proach has a number of constraints. Controllable Risk Factors Given the numerous complex causes of cancer, an effec- tive prevention program must be highly focused if the greatest prevention impact is to be guaranteed. Therefore, this report highlights and recommends actions to prevent those factors that cause the greatest number of cancer deaths and that are the most controllable, e.g., tobacco use, diet, and exposure to carcinogens in the workplace. Tobacco Use Use of cigarettes, cigars, pipe and chewing tobacco, and snuff is the most well-known cause of cancer mortality in our nation today. It accounts for over 30% of cancer deaths that could readily be prevented by a major reduc- tion in tobacco use. Dietary Factors These factors are significant in cancer occurrence, pos- sibly as significant as tobacco use. The consensus is that as much as 25% to 35% of cancer mortality could be related to dietary factors. This estimate is based on a large number of studies, although uncertainty surrounds the exact magnitude of the association and the biologic mech- anisms involved. The relationship between diet and cancer is under study, and more data should be available early in the next decade. Occupational Exposure to Carcinogens Exposure to carcinogens through occupation is included as the third factor for preventive actions for three reasons. l) Occupational exposures are in principle controllable through the exercise of caution by the individual and through enforcement of the wide range of existing laws and regulations. 2) These exposures at substantial levels affect relatively narrow segments of the population but, for those persons, have the potential for a high impact. 3) The populations who have the greatest occupational exposures also have the highest smoking rates (and possi- bly dietary risk factors), a combination which increases the risk for cancer. Occupational exposure to carcinogens is estimated to be responsible for up to 4%—6% of all cancer mortality. This report places heavy emphasis on the above three controllable risk factors. Although other risk factors are also important and will be given attention, these three will receive the greatest emphasis. Any mortality reduction yield for activities related to other controllable risk factors is not included in the plan. Although NCI is focusing for the year 2000 on those factors that cause the most deaths and that are most subject to prevention, it is important that current efforts by government, industry, and others to IS monitor and control other carcinogenic factors continue. A complete review of factors related to cancer in the United States has been provided by Doll and Peto (2); their findings are summarized in table 11-]. In this chapter, each of the three factors mentioned pre- viously is discussed in more detail. For each, this report includes a set of recommended prevention activities. These recommendations, which typify actions that could signifi- cantly reduce cancer mortality in the United States, repre- sent a statement of potential. Health care professionals and the general public are invited to add to them and to act on them, thereby becoming part of a process to reduce cancer mortality substantially by the end of this century. TOBACCO USE Thousands of epidemiologic and animal studies have provided conclusive evidence that tobacco use increases a person’s risk of developing cancer at a variety of sites; in particular, smoking causes lung cancer. The weight of the evidence linking tobacco use to cancer is so uniformly per- suasive that the Surgeon General of the United States has stated that “Cigarette smoking is the chief, single, avoid- able cause of death in our society and the most important public health issue of our time.” This section highlights the major quantitative relationships among smoking, tobacco use, and cancer. Contribution of Tobacco Use to Cancer Mortality and Risk The 1982 Surgeon General’s report (18) estimated that 30% of all cancer deaths are attributable to smoking. Cigarette smoking is the principal cause of lung and laryngeal cancers, a major cause of oral and esophageal cancers, and a contributory factor in the development of bladder, kidney, and pancreatic cancers. Some association has been observed between smoking and stomach cancer, and a possible relationship has been noted with cervical and liver cancers. The risk is greatest for cigarette smokers, with risk increasing with duration, amount of smoking, and amount of tar exposure. The latter factor is principally a function of the tar yield of the product, the presence or absence of filters, and the depth of inhalation. Pipe and cigar smok— ers and users of smokeless tobacco (chewing tobacco and snuff) are at less risk than cigarette smokers (although this may be due to the amount chewed or smoked); neverthe- TABLE ILL—Causes of cancer mortality“ All cancer deaths, % Best Range of Factor or factor class estimate acceptable estimates Tobacco 30 2540 Alcohol 3 2—4 Diet 35 10770 Reproductive and sexual behavior 7 1-13 Occupation 4 2—8 Pollution 2 1—5 Industrial products 1 1»2 Medicines and medical procedures 1 0.573 Geophysical factors 3 2—4 " See (2). 16 less, the use of any of these forms of tobacco is associated with an increased risk for cancer, particularly for oral cav- ity tissues and the throat. For example, in a recent study of North Carolina women, Winn et al. (19) reported that the relative risk of oral and pharyngeal cancer associated with snuff dipping among white nonsmokers was 4.2, and among chronic users the risk approached 50-fold for can- cers of the gum and buccal mucosa. The carcinogenic action of tobacco is enhanced syner- gistically by exposure to other carcinogenic or co-carci— nogenic substances. Consumption of large amounts of alcohol and concurrent cigarette smoking substantially increase risk (over a simple additive effect) for cancers of the mouth, larynx, and esophagus. Occupational expo- sures to carcinogens in conjunction with smoking also multiply the risk for cancer. Asbestos workers who smoke have 4 to 5 times the risk for lung cancer as other smokers and over 50 times the risk of persons with neither expo- sure (18). Similarly, uranium miners who smoke have six to nine times the risk for lung cancer that other smokers have. Lung Cancer Lung cancer is the leading cause of cancer deaths among men and has already begun to surpass breast cancer as the leading cause of death from cancer among women. In 1981, there were 106,389 deaths from lung cancer in the United States. Approximately 91% of lung cancer in men and 77% in women is attributable to cigarette smoking. This translates to approximately 90,000 lung cancer deaths caused by smoking in 1981. Smokers have up to 25 times the risk of developing lung cancer as do nonsmokers, depending on the number of cigarettes they smoked, their age when they began smoking, and the degree of inhala— tion. Heavy smokers (more than 2 packs per day) have lung cancer mortality rates 15 to 25 times greater than non- smokers. Laryngeal Cancer Cancer of the larynx was responsible for about 1% of cancer deaths (3,529 deaths) in 1981. The major cause of this disease is smoking; cigarette, pipe, and cigar smokers have a similar risk. Approximately 75% of laryngeal cancer in men and 43% in women is caused by smoking. As with lung cancer, the epidemiologic evidence for the relation- ship is Strong and consistent across studies, with risks for heavy smokers 20 to 30 times higher than for nonsmokers. Oral Cancers Oral cancers accounted for about 2% of the cancer deaths (8,506 deaths) in the United States in 1981, with men having two to three times the risk of women. Pro— spective studies have found mortality rates for oral cancers in smokers ranging from 1.2 to greater than 13 times the risk for nonsmokers. Cigarette, pipe, and cigar smokers have a similar risk. Esophageal Cancer Smoking is a major cause of esophageal cancers, which accounted for about 2% of all cancer deaths (8,082 deaths) in 1981. Prospective studies show that risk of death from esophageal cancer is 1.2 to 12 times greater for smokers than for nonsmokers; this risk depends on the amount of tobacco smoked. NCl MONOGRAPHS, NUMBER 2. 1986 Pancreas, Bladder, and Kidney Cancers Cigarette smoking is a contributory factor in cancers of the pancreas, bladder, and kidney. Pancreatic cancer ac- counted for 21,790 deaths in 1981. Approximately 40% of this neoplasm in men and 25% in women is attributable to smoking. Most recent studies have found the risk for smokers to be double that of nonsmokers (I8). Bladder cancer accounted for 9,813 deaths in 1981. Approximately 56% of bladder cancer in men and 30% in women is attributable to smoking (18). Relative risks for smokers range from 1.3 to 7.3. Kidney cancer accounted for 7,702 deaths in 1981. The risk of death from kidney cancer is 1.1 to 5 times greater for smokers than for nonsmokers. Extent and Trends of Smoklng Adults The percentage of adults who smoke1 has declined steadily since 1965 as public awareness of the health hazards of smoking has increased. As shown in figure [-1, in 1965, 52.1% of men over 20 were smokers; in 1980, 37.9% were smokers; in 1983, 35.4%. For women in the same years, the rates were 34.2%, 29.8%, and 29.5%, respectively. Although smoking prevalence rates for wo- men have never been as high as those for men, their rate of decline is considerably less than that of men. The per- centage decrease in smoking for women from 1965 to 1983 is about two-fifths that for men, e.g., 12.6% (34.2% falling to 29.9%), compared with 32.1% for men. In 1980, as shown in figure 11—], more black men smoked than did white men (44.9% vs. 37.1%), whereas about the same per- centage of black and white women smoked (30.6% and 30.0%, respectively). On the other hand, although rates for black and white men declined similarly between 1965 and 1980 (24.6% and 27.7%, respectively), the decline in smok- ing for black women was less than one-half that of whites (6.4% vs. 13.0%). Among all gender and race groups, the prevalence of smoking is sharply lower among those over age 65. These trends have been accompanied by the adoption of cigarettes with a lower tar yield. Cigarettes yielding less than 15 mg of tar were smoked by only 6.8% of women and 1.8% of men smokers in 1970; by 1979, these propor- tions had increased to 23.4% for women and 17.0% for men. A less favorable trend is that the percentage of smokers who smoke 25 or more cigarettes a day has increased since 1965 for both men and women (table 11-2). In 1965. 26% of the white men who smoked were heavy smokers; in 1983, 36% were. In that same period, white women who were heavy smokers increased from 14% to 22%. When coupled with the decrease in the percentage of men who smoke at all, these figures show that the overall prevalence of men who are heavy smokers has remained at a constant 13%, with cessation gains largely among smokers who smoke fewer than 25 cigarettes a day. Similarly, the preva- ‘ Regular smokers are defined for the purposes of most surveys as persons who have smoked at least 100 cigarettes and who were smokers at the time of the interview. CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 Females Males 60 _ 55-1 50— 45- 35- 30— 25— 20— 15- 10— 5— ADULTS SMOKING (PERCENT) [)— l l 1965 1980 1965 1980 I 1965 1980 1965 1980 YEAR FIGURE II-1.—Trends in smoking prevalence by sex and race in the United States, 1965—80. lence for women who are heavy smokers nearly doubled from 4.4% in 1965 to 7.5% in 1980. This increase in the proportion of heavy smokers may be a statistical effect of higher cessation rates among lighter smokers or an actual increase in the number of heavy smokers. The fact that blacks have higher rates of lung cancer despite a smaller percentage of heavy smokers than whites may be due to earlier initiation of smoking, the type of cigarettes smoked, or perhaps even the typography of their smoking, e.g., depth of inhalation. Strong socioeconomic differentials in smoking rates exist. In 1980, 52% of the population 25 to 44 years old who had less than a high school education were smokers, compared with 31% of that age group with some college education. During the past 10 years, the prevalence of smoking among those with less than a high school educa— tion decreased by only 7% compared with a 31% drop among those with some college education. In general, fewer white-collar workers smoke than do blue-collar workers, although the trend is not as strong for women as for men. Low family income is also associated with higher smoking prevalence. Youth In 1975, among high school seniors, 19.5% of the boys and 15.5% of the girls smoked one-half pack or more of cigarettes per day. After 1977, this percentage began to decline for boys and reached approximately 13% in 1983. For the girls, the percentage rose from 15.5% to almost 19% between 1975 and 1977. Thereafter, the prevalence TABLE ILL—Smokers smoking more than 25 cigarettes per day, %" Smokers 1965 1976 1980 1983 White men 26.0 33.3 37.3 36.3 White women 13.9 20.9 25.2 21.7 Black men 8.6 10.8 13.8 11.6 Black women 4.6 5.6 8.6 5.3 " Source of data is the Health Interview Survey conducted by the NCHS. began to decline, although at a slower rate than for boys. By 1983, the prevalence for girls was similar to that of boys, about 13%. Daily smokers of less than one-half pack per day in 1983 were 29% for boys and 21% for girls. Although these statistics for high school seniors are informative, they are likely to be underestimates for that age group because they do not include those who have dropped out of school before the senior year, a group often at higher risk for smoking by virtue of race and lower SES. Also, a number of studies have indicated that when self-reported rates are compared with rates verified by biochemical tests, those in self-reports are often under- estimated. Finally, comprehensive survey data are lacking on smoking habits in the years immediately after high school. Potential Impact of Prevention All major studies in which the relationship between ces- sation of smoking and lung cancer was examined report a decrease in risk; smokers who have quit for 15 or more years have lung cancer mortality rates between one and two times that of nonsmokers. The reduction in risk and eventual residual risk are determined by the lifetime cumu- lative exposure before cessation. Studies show a similar risk reduction with smoking cessation for laryngeal and oral cancers after 15 years. For esophageal cancer, smok— ing cessation reduces risk to about one-third that of the continuing smoker. Because of the lag time required for reduction in risk after cessation, intensive cessation efforts must be made soon if we are to have a major change in lung cancer mortality by the year 2000. In addition to the emphasis on smoking cessation in this report, efforts must also be undertaken for prevention of smoking. Although the effect of these efforts for the young will not be realized until after the year 2000, the mortality reduction goal can only be achieved through the elimina- tion of tobacco use in each age group. Chapter I of this report and Appendixes A and B outline the potential impact of the objectives. Reductions will be achieved through the combined efforts of federal agencies (includ- ing the NIH); voluntary organizations such as the Ameri— can Cancer Society, the American Heart Association, and the American Lung Association; and state, local, and pri- vate agencies and organizations. The actions recommended by the working groups to achieve the reductions are based on interventions presently known to be effective in pro- ducing cessation and prevention, as well as on ongoing research funded by NCI that is expected to produce addi- tional methods suitable for broad-scale implementation within 5 years. Oblectlves Ior Smoking Reduction The Prevention Working Group and NCI staff devel— oped objectives as reasonable and feasible to achieve by the year 2000. The objectives are based on the assumption that the overall rate of decline in cigarette smoking of about 2% per year (in men) can be doubled and will apply to men and women in an intensive and coordinated 15- year campaign against smoking. The objectives incorpo- rate and extend those of the DHHS (1). To meet the year 2000 objectives, this Working Group and NCI recommend that the following objectives be achieved by 1990. 18 YEAR 1990 INTERMEDIATE SMOKING OBJECTIVES OF THE NATIONAL CANCER INSTITUTE Risk Factor Reduction The targeted smoking prevalence for 1990 for various groups and recent figures on smoking prevalence for each group are as follows: Recent smoking m 1990 targeted Population Year % smoking prevalence, % Adults 1980 33.8 <25 High school seniors 1979 11.7 $6 Children and youth 1980 11.7 <6 (12—18 yr) Black adult males 1980 44.9 327 Black adult females 1980 30.6 £19 Reduction of smoking should occur in equal propor- tions among all levels of smokers, i.e., heavy (25 or more cigarettes/day), moderate (11 to 24 cigarettes/day), and light (I to 10 cigarettes/day). The rate of decline of smoking should be 4% or more per year. (In 1984, the rate of decline of smoking was 2%.) Increased Public and Professional Awareness At least 90% of the adult population should be aware that smoking is a major cause of lung cancer and of mul- tiple other cancers including laryngeal, esophageal, and bladder. (In 1985, about 35% and 85% of the adult popu- lation knew that smoking increases their chances of bladder and esophageal cancer, respectively.) At least 65% of the 12-year-old children should be able to identify smoking cigarettes as a cause of increased risk of serious disease of the heart and lungs. (Baseline data are unavailable.) Improved Services and Protection At least 65% of all workers should be offered employer- or employee-sponsored or -supported smoking cessation programs, either at the worksite or in the community. (In 1979, 15% of the business firms in the country had pro- grams to encourage or assist their employees in smoking cessation.) Laws should exist in all 50 states and in all jurisdictions that would prohibit smoking in enclosed public places and establish separate smoking areas at work and in dining establishments. (In 1978, 31 states had some form of such restriction laws.) YEAR 2000 SMOKING OBJECTIVES OF THE NATIONAL CANCER INSTITUTE Risk Factor Reduction The proportion of all persons 21 years and older who smoke should be reduced to 15% or less. The percentage of youths who begin to smoke by age 20 should be reduced to 15% or less (12—18 yr, <3%). NCI MONOGRAPHS, NUMBER 2. I986 Recommended Actions for Reduction of Tobacco Use The following recommended actions to reduce smoking are based on future large-scale application of current and developing cessation and prevention intervention tech- niques. Many regional and national programs conducted by local, state, federal, and private sectors already exist and have contributed significantly to the decline in smok- ing prevalence during the past 20 years. To affect current smokers, all sectors must make a concerted effort to put in place a system of antismoking measures that is effec- tive, acceptable to the public, cost-efficient, and self-per- petuating. It is expected that a wide range of strategies will be necessary including educational, behavioral, social, and regulatory. These strategies must be implemented through a variety of agents and channels who should provide the sustained and persuasive influence required to reduce smoking prevalence to the goal levels stated in this report. The actions recommended address eight categories of agents and channels capable of conducting programs or developing policies to influence smoking reduction in the United States. Because of the significant impact they can have, NCI encourages these agents to accelerate and inten- sify their efforts for cancer control. For reduction of tobacco use, the chief roles of the NCI are to 1) guide and support research in cessation and prevention of tobacco use, 2) provide information and technical assistance to other agencies and organizations interested in tobacco— related prevention programs, and 3) conduct public educa- tion programs about the cancer risks of tobacco use. National Cancer Institute The following are examples of actions within the pur- view of the NCI that will be continued or undertaken: Continue to publicize the risks of smoking and to- bacco use through the Office of Cancer Communica— tions, its Cancer Prevention Awareness Program, and other NCI programs. Inform the public about available resources to help in the prevention and cessation of smoking and tobacco use through the CIS. Increase the role of the Comprehensive Cancer Cen- ters in the prevention and cessation of smoking and tobacco use. Promote close collaboration in research by develop- ing a network among smoking intervention researchers funded by the NCI Smoking, Tobacco, and Cancer Program. Involve NCI-sponsored researchers of smoking inter- vention programs in NCl’s year 2000 activities and, especially, in the application of findings that come from their research. State and Local Health Departments These departments are encouraged to consider the fol- lowing recommended actions: Collaborate at the state and local levels to integrate smoking prevention and control techniques into health care delivery, health promotion, or cancer control programs. CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 Develop the capability to conduct periodic surveys on the prevalence of smoking and to assess problems and progress. Develop statewide or local nonsmoking plans and coalitions. Develop and implement statewide health risk ap- praisal and risk reduction programs that include smoking as a risk factor. The Workplace Corporate and union leaders are encouraged to consider the following recommended actions: Develop and promote model nonsmoking standards as an integral part of worksite health promotion and fitness programs. Increase awareness of tobacco and smoking-related risks, especially in industries with high risk for cancer, Design smoking cessation programs for the under- served working population (recent immigrants; mi- grant, part-time, and shift workers). Physlclans and Health Care Providers Physicians and health care providers are encouraged to consider the following recommended actions: Educate medical students, physicians, and health care providers about not only the health consequences of smoking and tobacco use but also prevention and ces- sation methods. Encourage physicians and other health care providers to serve as role models by refraining from tobacco use. Encourage collaboration among medical associations, specialty societies, nursing, and allied health profes- sions in educating physicians and others about to- bacco and health and their roles in prevention. Include smoking cessation and education programs in all occupational medicine programs. Increase questions about tobacco and cancer on certi- fication and licensure examinations for medical and other health professionals. Promote articles about tobacco and cancer in the major medical and health profession journals. Expand the curriculum relative to smoking, tobacco, and health in schools for health professionals. Promote the adoption of smoking prevention and cessation intervention techniques as a standard prac- tice by all physicians and health care providers. For example, intervention can be directed especially to— ward heavy smokers; patients with lung disease, heart disease, cancer, or alcohol dependence; and patients in occupations with high risk for cancer. Assist patients in smoking cessation or refer them to programs that offer aid. Increase programs for members of all health profes- sional associations concerning smoking, tobacco, and cancer, and effective intervention techniques. 19 Health Care Institutions Health care institutions are encouraged to consider the following recommended actions: Restrict smoking in all health care institutions. Prohibit sale of cigarettes in hospitals, nursing homes, other health care facilities, and pharmacies Involve health maintenance organizations, the Ameri- can Hospital Association, and local hospitals in the collaborative development of standards requiring hospital—based cessation and prevention programs. Voluntary and Other Organizations Voluntary and other organizations are encouraged to consider the following recommendations: Develop widespread collaboration with the recently formed “National Coalition” of the American Cancer Society, the American Lung Association, and the American Heart Association to achieve their goals for a “Smoke—Free Society” in the year 2000. Establish and expand insurance company policies by offering lower premiums to nonsmokers for health, life, home, auto, and property insurance. Collaborate with the DHHS Office on Smoking and Health in the implementation of the Comprehensive Smoking Education Act. Schools Schools are encouraged to consider the following recom- mended actions: Promote the prohibition of smoking and tobacco use on school property. Promote comprehensive health education, emphasiz- ing avoidance of smoking and other high-risk behav— ior from kindergarten through grade 12. Advocate health education courses in all teacher training and certification programs. Involve parents in planning nonsmoking and health education programs. Make cessation programs available to students, fac- ulty, and staff. Promote the establishment of a “peer corps” of high school seniors, who might be certified and given incentives, such as scholarships, for their roles in prevention and cessation. Media The media are encouraged to consider the following recommended actions: Increase the coverage of smoking, tobacco use, and health issues. Establish communication between appropriate health— related groups and the media to increase the public’s knowledge of the hazards of tobacco. Improve communication between professionals in the media and those in science and health. 20 DIETARY FACTORS Contribution to Cancer Mortality and Risk Roughly 35% of cancers may be attributable to dietary factors (2). A large number of studies, conducted with vari- ous methodologies, indicate that excessive fat intake, inadequate dietary fiber, and inadequate consumption of certain micronutrients (vitamins and minerals) are asso- ciated with higher rates of certain cancers. At this time, the evidence that would fully characterize these associa- tions is incomplete, and precise quantification of the mag- nitude of those associations is not possible. Dietary factors are associated with cancers of the gastrointestinal tract (esophagus, stomach, colon, rectum, pancreas, and liver) and some sex— and hormone—specific sites (breast, prostate, ovaries, and endometrium). They also may be related to cancers of the respiratory system and the urinary bladder. Dietary Patterns Our national food statistics, measured from 1909 to the present, show trends in the diet according to per capita supplies of nutrients and foods in consumer channels. These statistics, termed “disappearance data,” measure quantities of food available for consumption but are not direct estimates of consumption. Nevertheless, disappear— ance data are the only available time—trend series for stud- ies of shifts in the dietary behavior of the population his- torically. The historical series Nutrient Content of the US. Food Supply, compiled by the Department of Agri- culture, contains estimates of the changes in per capita levels of nutrients and in the percentages of nutrients from major food groups since the beginning of the twentieth century (20). The disappearance data indicate that our diet patterns have changed dramatically in the course of the century. Overall, Americans have been eating more meat, poultry, fish, dairy products, refined sugars and sweeteners, fats and oils, and processed fruit and vegetables than at the start of the century. Americans have also been eating fewer grain products, potatoes, fresh fruits and vegetables, and eggs. It is clear from the trends in the disappearance data that American diets have changed and continue to do so; this change reflects alterations in food processing and distribu- tion as well as consumer demand. Thus it would appear that modifications in consumption are feasible and might be achieved by consumer education and changes in food production. In addition to experimental evidence, four types of epidemiologic studies show an association between dietary components and cancer: 1) international correlational stud- ies comparing dietary intake to cancer rates, 2) migrant population studies comparing cancer rates in persons who have moved from (or to) areas with low cancer rates to (or from) areas with high cancer rates, 3) comparisons of cer— tain low—risk populations (e.g., Seventh-Day Adventists, Mormons) to the general population in the United States, and 4) case—control and cohort studies in which dietary habits of cancer patients are compared with others in the study population. NCI MONOGRAPHS, NUMBER 2, I986 Dietary Fat and Fiber A high intake of fat has been shown in some studies to be associated with cancers of the breast, colon, rectum, prostate, and possibly pancreas, uterus, and ovaries. 0n the other hand, high intake of dietary fiber may be asso- ciated with lower risk for colon and rectal cancers. Al- though it is not possible for one to estimate precisely the relative contribution of fat or the protection of fiber in the risk for colon and rectal cancers, a few studies suggest a separate contribution from each. The protective effects from a reduction in fat consumption and an increase in fiber intake may be additive. Increases in dietary micronutrients, such as vitamins C and E, folate, and the carotenoids (chemicals related to vitamin A), have been linked to the prevention of cancer. Although the specific protective effects of micronutrients are under study, the National Research Council has rec- ommended an increase of micronutrient intake through daily consumption of those fruits and vegetables rich in these vitamins (3). Alcohol The analysis by Doll and Peto (2) links alcohol with about 2% to 4% of cancer mortality. The particular cancer sites that have been associated with high alcohol consump- tion include buccal cavity, pharynx, larynx, esophagus, liver, large bowel, and breast (21). The DHHS has identi— fied the relationship between head and neck cancer and alcohol as an important one to be conveyed to the public. Alcohol was not addressed as a specific risk factor in the process of setting NCI objectives, but the possibility for reducing the incidence of head and neck cancers rests partly on a reduction in excessive alcohol consumption. Potential Impact of Prevention International data indicate a strong direct relationship between the amount of dietary fat and increased risk of death from breast cancer. Although other epidemiologic data are not conclusive, they provide sufficient evidence to warrant prudent dietary modifications, as suggested by the NAS. If the international data are used for estimates of the magnitude of a potential effect, a warranted conclu- sion is that a reduction of dietary fat from 40% to 30% of calories could reduce the death rate from breast cancer by 25% in 10 years; however, it is also possible that no reduc— tion would occur. Today, 36,000 women die of breast cancer each year; a 25% reduction would save as many as 9,000 lives per year. As for other dietary changes, quantitative information is insufficient for a precise estimation of the impact of increased fiber intake. There are 58,000 deaths per year from colon cancer in the United States. With the assump- tion that the protective effects of increased fiber and decreased fat are equivalent to the differences found in studies on migrant populations, this death rate could be reduced by 50% in 10 years and could save 29,000 lives annually. Although one cannot quantify the precise contribution of diet to overall cancer risk nor estimate with certainty the reduction in cancer mortality to be expected from die- tary modifications, international and migrant population data suggest the following possible reductions in cancer CANCER CONTROL OBJECTIVES FOR THE NATION: I985 2000 incidence if the objectives stated in the next section were met: I) a 50% reduction in cancer of the colon and rectum; 2) a 25% reduction in cancer of the breast; 3) a 15% reduc- tion in cancers of the prostate, endometrium, and gall- bladder; and 4) a possible, but not precisely quantifiable, reduction in cancers of the stomach, esophagus, pancreas, ovaries, liver, lung, and bladder. Objectives for Dietary Factors The following objectives strike a balance between our acting now upon existing knowledge and waiting for the results of research under way or projected to begin. If the research results support the hypotheses about the relation- ships between dietary components and cancer, waiting to change dietary behavior until after those results are avail- able could mean the unnecessary loss of thousands of lives. To begin efforts to change diets now seems only prudent in light of our existing knowledge and the rec- ommendations of the National Research Council. These behaviors are in keeping with good dietary habits for general health, in particular the dietary fat recommenda- tions, and should reduce the mortality from heart disease as well as cancer. To meet the year 2000 objectives, NCI recommends that the following intermediate objectives be reached by 1990. YEAR 1990 INTERMEDIATE DIETARY OBJECTIVES OF THE NATIONAL CANCER INSTITUTE Risk Factor Reduction The per capita consumption of fiber from grains, fruits, and vegetables should be increased to 15 g or more per day. (In 1976—80, the per capita consump- tion of fiber from these sources was 8—12 g.)2 The per capita consumption of fat should be below 35% of total calories. (In 1976—80, the per capita con— sumption of fat was 37%-38% of total calories.) Increased Public and Professional Awareness The proportion of the population able to identify the principal dietary factors known or strongly suspected to be related to cancer should exceed 75%. At least 70% of the adult population should be able to identify foods that are low in fat and high in di- etary fiber. The proportion of adults who are aware of the added risk of head and neck cancer for people with excessive alcohol consumption should exceed 75%. Improved Services and Protection The labels of all packaged foods should contain use— ful calorie and nutrient information to enable con- sumers to select diets that promote and protect good health; similar information should be displayed where nonpackaged foods are obtained or purchased. 2 Per capita daily consumption of fruits and vegetables rich in carotenoids and other micronutrients should increase concomi- tantly. 2i All states should include nutrition courses as part of the required comprehensive school health education curricula at elementary and secondary levels. (In 1979, only 10 states mandated nutrition as a core content area in school health education.) Virtually all routine health contacts with health pro- fessionals should include some element of nutrition education and counseling. All managers of institutional food service operations should understand and actively promote dietary pat- terns that are in accord with current knowledge of the relationship between diet and good health. YEAR 2000 DIETARY OBJECTIVES OF THE NATIONAL CANCER INSTITUTE Risk Factor Reduction The per capita daily consumption of fat should de- crease from the 37%—38% of total calories that it was in [976780 to 30%. The per capita consumption of fiber from grains, fruits, and vegetables should increase to 20—30 g/day from the 8 to 12 g/day it was in 1976—80.2 Recommended Actions for Dietary Factors For dietary factors, the chief roles of the NCI are to 1) guide and support research on the cancer-related effects of dietary fat and fiber, chemoprevention, and dietary behavior; 2) provide information and technical assistance to other agencies and organizations interested in diet-re- lated prevention programs; and 3) conduct public edu- cation programs about the health advantages and cancer risks of relevant dietary components. Recommended Actions for Other Federal Agencies and Industries Federal agencies and industries are encouraged to con- sider the following recommended actions: Reflect the cancer prevention dietary recommenda- tions in federal food production, marketing, and dis- tribution policies to ensure that the entire population has access to and is encouraged to use appropriate foods. Modify regulations governing standards of identity to allow less fat in the formulation of foods. Modify federal meat and milk grading programs to encourage producers to expand production and mar- keting of leaner meat products and lower fat content of dairy products. For example, the red meat industry has proposed new grading standards that permit the choice and prime grades to be leaner, which would result in less fat in American diets. Intensify research on selective breeding to promote leaner livestock products. Promote research for the enhancement of quality and desirability of fiber-rich foods, including cereals, breads, fruits, and vegetables. 22 Through all federally funded food, nutrition, and health programs for the public, inform the people about the relationship between diet and cancer and motivate them to eat less fat and more foods rich in fiber and micronutrients. Give priority in federal funding to programs to edu- cate high-risk segments of the public about diet and cancer risks. Expand nutrition labeling to cover the full range of mass-marketed foods contributing fat to American diets (including fresh meats), so consumers can be better informed and make wise shopping decisions. Develop diet and cancer programs that make use of the mass media and other high-technology communi- cation approaches. Role ot State and Local Agencies State and local agencies are encouraged to consider the following recommended actions: Review and evaluate elementary and secondary school health and home economics curricula and training guides and draw upon appropriate local experts to upgrade these to reflect newer knowledge of diet and cancer risks and strategies for risk reduction. In addi- tion, school menus should be reviewed in relation to the cancer control objectives. Encourage, whenever possible, and assist voluntary and private-sector groups to modify existing health promotion programs to include information on cancer risk reduction. Encourage restaurants to provide sufficient informa- tion to allow patrons to choose nutritious foods. Coordinate program planning activities in state agen- cies for agriculture, environmental protection, health, and aging to ensure that attention is given to reducing dietary risk factors for cancer. Assign health and nutrition leaders to work with local mass media representatives to disseminate accurate, appealing information about proper food selection to protect against cancer risk. Include in state and county food, nutrition, and health programs information on eating for better health and for avoiding cancer risk with the use of particularly innovative approaches to reach high-risk groups. OCCUPATIONAL EXPOSURE TO CARCINOGENS Numerous substances that are or were present in the occupational environment have been established as car- cinogens for humans; others have been established as car— cinogens for animals only. A list of industrial processes and chemicals carcinogenic to humans presented in table 11-3 is derived from a larger table compiled by the IARC. Contribution of Occupational Exposures to Cancer Mortality and Risk Because of the complex nature of occupational expo- sures, the inadequate information on the number of peo- NCI MONOGRAPHS, NUMBER 2. I986 TABLE II-3.—lndustria1 processes and chemicals carcinogenic for humans as determined by the IA RC Industrial processes and occupations Auramine manufacture Isopropyl alcohol manufacture (strong-acid process) Nickel refining Underground hematite mining (with exposure to radon) Chemicals and groups of chemicals 4-Aminobiphenyl Arsenic and arsenic compounds Asbestos Benzene Benzidine N,N—bis(2—Chloroethyl)—2—naphthylamine (chlornaphazine) Bis(chloromethyl) ether and technical-grade chloromethyl methyl ether Chromium and certain chromium compounds Diethylstilbestrol Melphalan Mustard gas 2-Naphthylamine Soots, tars, and oils Vinyl chloride ple at risk, the changing character and magnitudes of exposures, and the interaction of occupational with other well-established risk factors (particularly smoking), neither the contribution of occupational exposures to cancer inci- dence and mortality nor the likely effects of any preventive action can be precisely quantified. Occupational exposure to carcinogens is generally thought to be responsible for 4% to 6% of cancer mortality in industrialized countries. However, this estimate is a generalization for the entire population, and those specific groups at high levels of exposure suffer a much greater risk. In addition, sub- stances encountered in the workplace in relatively high concentrations may also permeate the environment as pol- lutants or as inclusions in various consumer products. The risk for cancer caused in these ways has not been deter— mined because of the inadequacy of existing analytical methods and data. Just as for the other risk factors, this estimate is expressed as a proportion of total cancer cases and may change if other cancer causes increase or decrease their proportions, new factors are discovered, or if the estimates of the impact of new or old factors are revised. The majority of known occupational carcinogens were identifed from the mid-19605 to the early 1970s. The search for occupational carcinogens cannot be abandoned, but the most effective prevention efforts would likely come from a focus on safeguards to the known and strongly suspected human and animal carcinogens that have already been identified. Although additional agents may be identi- fied, their impact before the year 2000 is not likely to be great if they have only recently been introduced into the environment. However, their contribution to cancer inci- dence and mortality may ultimately be significant. Thus exposure to occupational carcinogens should be controlled to the extent that is technically feasible both in the work— place and in the general environment. Only limited data are available that show the effective- ness of knowledge and regulationsin changing occupa— tional exposure patterns. Such reductions in exposure as a CANCER CONTROL OBJECTIVES FOR THE NATION: I985 2000 result of the gradual tightening of health standards im- posed on industry should result in a significant reduction in risk for those exposed to carcinogens in the workplace. Slgnilicant Carcinogens in the Workplace Asbestos Asbestos, thought to be the greatest contributor to risk of cancer in the workplace, was introduced into the Ameri- can environment on a large-scale basis during World War II and through the early l970s. At that time, regulations were imposed in response to the increasing body of medi- cal evidence linking asbestos to cancer and other debilitat- ing diseases. Thereafter, asbestos use in production and construction in the United States diminished markedly. Estimates of the number of cancer cases attributed to asbestos vary widely. A generally accepted estimate that accounts for long latency between exposure and disease occurrence is that mortality caused by exposure to asbes- tos will increase approximately 50% by 1995—99 and then will gradually decline. Cigarette smoking in combination with asbestos exposure increases the risk of cancer many— fold, about 50 times the risk of the nonexposed non- smoker. Benzene During the last three decades, the production and use of many chemicals have increased primarily because of the expansion of petrochemical production. Benzene may be regarded as an “index” chemical in this regard because of its ubiquitous use; its production has increased at least six- fold since 1955. Although benzene is established as a human carcinogen (table 11-3), little is known about the risks. Because of lengthy latent periods, our knowledge today is based on exposures that took place decades ago. In the 19505, recommended exposure levels were 35 ppm. Today, exposure levels are required to be less than 10 ppm and are thought to be, at least in industrial production, below 1 ppm in most instances. It should be noted that control of exposure to carcino- gens in the workplace has great potential to reduce cancer mortality by a large percentage in the exposed groups. The analyses reported in chapter I and Appendixes A and B do not account for this reduction, which would be highly significant for relatively small population groups. The net effect is about a 4% mortality reduction (table II-l). Objectives for Prevention of Occupational Exposure to Carcinogens Specific objectives for the year 2000 related to occupa- tional exposure were not defined during this effort. Rather, the following set of objectives for the year 1990, outlined by the DHHS (I), were reiterated: All firms with more than 500 employees should have a plan of hazard control for those processes, equip- ment, and installations associated with established or suspected carcinogens. At least 25% of workers should be able, prior to employment, to state the nature of their occupational health and safety risks and the potential consequences; they should be informed of changes in these risks 23 while employed.3 (In 1979, an estimated 5% of workers were fully informed.) The majority of workers should be routinely informed of life-style behaviors and health factors that interact with factors in the work environment to increase risks for occupationally induced cancers.4 At least 70% of the primary health care providers should routinely query patients about hazardous ex— posures in the work environment as part of the medi- cal history and should know how to interpret the information for patients in an understandable manner. Recommended Actions for Prevention of Occupational Exposure to Carcinogens These recommendations are made with the acknowl- edgment that currently available cancer control technolo— gies are underutilized in practice, especially engineering controls, work practices, personal protective devices, and habits of hygiene. Effective engineering controls that elimi— nate or greatly reduce workplace exposures are the ideal goal for workplace safety. Much has been accomplished, e.g., in relationship to asbestos and benzene, despite tech— nical problems and costs to management. In addition, strategies have been developed and implemented to train workers to eliminate or reduce exposures through avail- able technologies, and new methods and approaches deal- ing with potential occupational exposure continue to be developed. An important prevention intervention would be identifi- cation of workers who have a high risk of cancer because of exposures to carcinogens in the workplace. In principle, these individuals are identifiable, and their chances of sur- viving cancer can be increased by earlier detection of the 3Statement is reworded from the DHHS objective. 4Objective target date was changed from 1985 to l990. 24 disease. In addition, cancer incidence may be reduced through control of interactive exposures, such as cigarette smoking by individuals who were exposed to asbestos. It is recognized that the practical applications of such pre~ vention interventions may be limited by difficulties in identification of individuals at risk, inadequacies in the financing and delivery systems for medical care or health education, and legal obstacles. Federal Agencies For control of occupation—related cancer, the role of the NC] is to l) conduct research to identify carcinogenic substances in the workplace and 2) estimate and monitor their impact. Appropriate activities are cancer prevention trials in humans and medical surveillance. In these activi- ties, NCI coordinates its research with the National Insti- tute of Occupational Safety and Health and the Occu- pational Safety and Health Administration, which are re— sponsible for research, information dissemination, and enforcement activities in the broad field of occupational safety and health. Numerous other agencies and organiza- tions in the public and private sectors also participate in this area. Analytic principles exist to help determine when preventive measures are needed. The National Cancer Advisory Board has concurred with the principles for risk assessment as set forth in a report by the NAS. Under these principles, data from nonhuman testing systems, when combined with other elements of quantitative risk assessment, can indicate when preventive actions are needed. A high level of coordination and cooperation among federal agencies should be continued with sufficient re- sources allocated for implementation of effective occupa- tional cancer prevention programs. State Agencies Improved surveillance and registration of occupational exposures are needed, and such registration should be linked to cancer surveillance systems. III. Screening Ill. Screening Screening is the application of tests or examinations to detect disease in persons who typically display no symp- toms of the disease in question. The screening procedure itself usually does not yield a definitive diagnosis; rather, the information it provides generally either excludes the likelihood of disease or identifies a basis for recommend- ing more sophisticated diagnostic procedures. “Mass screening” implies a systematic application of procedures to a large population, usually through a central mecha- nism established specifically to do the screening. Its pur- pose is to identify those for whom the probability of cur- rent disease is low and those for whom the probability of disease is high enough to justify further diagnostic exam- inations. This chapter addresses the technologies available to physicians to screen for particular cancers on a wide- spread or mass screening basis. Chapter IV, which deals with treatment, includes recommendations for early detec- tion procedures for the symptomatic or high-risk indi- vidual. POTENTIAL OF SCREENING TO REDUCE CANCER MORTALITY Because of the wide variety of cancer types and screen- ing procedures, the potential of screening to reduce mor- tality must be assessed for each cancer site. This section summarizes for certain cancer sites the evidence that screening techniques are effective in reducing mortality and the extent of the reduction achievable with screening and follow-up treatment. However, it should be borne in mind that the relatively low incidence rate or low death rate for many cancer sites, such as the prostate, oral cav- ity, endrometrium, or testis, makes the conduct of clinical trials (during which the effectiveness of screening is tested) extremely difficult, if not impossible. To judge the quality of evidence that screening reduces mortality, the Working Group used criteria consistent with those presented by the Canadian Task Force (22), the American Cancer Society (23), and the workshop spon— sored by the International Union Against Cancer (24). It is generally agreed that the strongest evidence is provided by randomized, controlled, clinical trials that use mortal— ity as an end point. Evidence from studies that are con- trolled but not randomized, such as those with historic controls and case—control studies, is generally weaker but still useful. The mere evidence that a screening procedure is capable of detecting a cancer is not by itself evidence that screening will reduce mortality. However, if there is evidence that therapy is more efficacious when delivered to cancers in early stages, then evidence that a screening procedure is able to detect a cancer early indicates, but does not prove conclusively, that screening reduces mor- tality. _ This discussion is divided into three categories of effec- tiveness: l) cancers and screening techniques for which there is general agreement that screening reduces mortal- ity; 2) cancers and screening techniques for which there is general agreement that screening does not appreciably reduce mortality; and 3) cancers and screening techniques for which there is no general agreement that mass screen- ing reduces mortality, either because of a lack of informa- tion or because of different interpretations of the existing information. Category 1: Screening Effective Breast Cancer Detection for Women Over the Age 0! 50 The effectiveness of screening women over the age of 50 with a combination of mammography and physical breast examination by a health care professional has been exam- ined in 5 major studies. The HIP of New York study, a controlled clinical trial in a defined population, showed a 30% reduction in breast cancer mortality after a 10-year follow-up (6), as has a recent randomized trial in Sweden (7). The BCDDP, although not a controlled trial, showed results consistent with the hypothesis that physical exami- nations and mammography reduce mortality from breast cancer (25, 26). Two recent case—control studies also sup- port the results of the HIP study (27, 28). Conical Cancer It is widely agreed that screening by Pap smears at least once every 3 years for women aged 20 to 70 reduces mor- tality from cervical cancer. Although no randomized, con- trolled, clinical trials have been conducted, a large body of evidence supports the value of the Pap smear for detecting preinvasive lesions of the cervix and early invasive cancers. The evidence includes comparisons of cervical cancer inci- dence and mortality in regions in a country (29—33) where women were unscreened and heavily screened and histori- cal comparisons of incidence and mortality rates before and after the introduction of screening (8, 34—38). These comparisons indicate that Pap smears can detect cervical cancers in early stages and thereby can reduce mortality (31). The 3—year interval between screenings is based on the studies referred to above, on knowledge of the natural his- tory of cervical cancer, and on mathematical modeling (39). There is wide agreement that the great majority of lesions progress through stages from dysplasia, through carcinoma in situ, to microinvasive and invasive cancer (40, 41). Given the risk of false-negative results on a single 27 Pap smear, or the possibility that some lesions progress rapidly, some have argued for a screening interval of 1 year (42). However, a minimum of I screening examina- tion every 3 years is deemed reasonable and highly cost- effective by most investigators and organizations (23, 43-47). The best available estimates are that screening every 3 years can lead to a 70%—95% reduction in cervical cancer mortality. Category 2: Screening Ineffective Several clinical studies conducted in the 1960s and 1970s, involving chest x-rays, sputum cytology, and question- naires delivered in various combinations and frequencies, did not indicate that lung cancer screening significantly affects mortality from that disease (48—52). Also, 3 recent randomized controlled trials, involving chest x-rays and sputum cytology given as often as every 4 months, have not shown that lung cancer screening has a significant effect on mortality (53—55). Although some individual practitioners and researchers have expressed dissent, vir- tually every organization that has reviewed the available data has concluded that lung cancer screening with these techniques is not effective (22, 23, 44, 47, 56, 57). Category 3: Screening Impact Uncertain Without extensive clinical trials or supporting studies, it is difficult for one to establish a causal link between a screening technique and changes in mortality. Indeed, for many low incidence cancer sites, amassing of definitive evidence for the impact of mass screening or early detec- tion techniques on cancer mortality is difficult because of the infeasibility of performing prospective clinical trials. For techniques in category 3, the existing evidence is insufficient and does not permit conclusions to be drawn as to their effectiveness in reducing mortality. Some of these techniques are highly recommended as part of peri- odic health examinations (by physicians) or as techniques used by individuals to detect early signs of cancer. Rec- ommendations for the use of the techniques for early detection by physicians or by the individual were not addressed by the Screening Working Group and are not affected by these categories; the Treatment Working Group considered the recommendations related to early detection that are outlined in chapter IV of this report. Within category 3, the techniques and cancers are ar— ranged in 3 groups. The first group includes those gener- ally recommended as part of either a periodic health examination or as part of an individual’s periodic self- examination (22, 23): breast self-examination; prostate, oral, and testicular cancers; and melanoma. They are an essential part of clinical practice but are included in cate- gory 3 because consensus estimates of their impact on mortality as screening techniques are not available. The second group of techniques and cancers includes breast mammography coupled with physical examination in women aged 40 to 49. Both techniques are currently under study and therefore fall into category 3. The third group includes bladder, endometrial, esopha- geal, and gastric cancers. The techniques used in the detection and diagnosis of these cancers also have been considered for mass screening; however, their potential to reduce mortality is not known. 28 Breast Self-Examination The evidence about the effectiveness of breast self— examination is weaker than that for mammography and physical breast examination for women under 50. Despite indirect evidence that breast self—examination is effective (58, 59), there is no general agreement on the extent to which breast self-examination reduces cancer mortality. Prostate Cancer The digital rectal examination is the primary method for detection of prostate cancer. Studies verify that by dig— ital rectal examination physicians can be expected to detect early cases of prostate cancer (60, 61). The exami- nation is included in the recommendations of the Ameri- can Cancer Society for a periodic annual health checkup for males aged 40 and above (23). No randomized clinical trial with mortality as an end point has been conducted so that precise calculations of the impact of the examination on cancer mortality are not possible. Oral Cancer Although no randomized trials documenting the value of oral examination have been reported, annual examina- tion by a health professional (as part of a medical or den- tal checkup) has been advocated by some observers. This plan of examination seems likely to result in earlier detec- tion of oral cancer, but the extent of mortality reduction is not known. Testicular Cancer Monthly self-examination for testicular cancer has been recommended by some investigators for males aged 15 to 45. The lack of clinical studies keeps others from recom- mending this type of screening (24). Melanoma Experience in Queensland, Australia, suggests that a campaign to promote early detection of melanoma may result in improved survival (62). Periodic self-examination and examination by others have been encouraged for people over 30 who have fair skin and are subject to heavy actinic exposures. Periodic medical examinations or self- examinations have been recommended for persons with dysplastic nevus syndrome (a potential precursor to malig- nant melanoma) and for those with a history of melanoma in a first-degree relative. Baseline photography of nevi also has been claimed to help document the appearance of lesions in high—risk individuals. At this time, the lack of controlled clinical trials prevents the estimation of impact for these screening activities (24). Breast Cancer Detection In Women Aged 40 to 49 The HIP study showed a statistically significant differ- ence in mortality only for women age 50 and above who were screened with annual physical examinations and mammograms. For women under 50, there was a small reduction in mortality, but the results were not statistically significant. The BCDDP results indicate that mammog- raphy has improved since the HIP study was conducted and that the procedure is almost as effective for younger as for older women. However, the BCDDP was not a con- trolled study, and firm conclusions cannot be drawn. In fact, the results of a case—control study (27) lead one to an opposite conclusion. In short, the evidence needed to make definitive statements about the efficacy of annual physical NCI MONOGRAPHS, NUMBER 2, I986 examinations and mammograms for women under 50 is not available. The value of breast cancer screening for women between the ages of 40 and 49 is being assessed in a randomized trial in Canada. Colorecfal Cancer A randomized, controlled, clinical trial during which investigators compared multiphasic health examinations incorporating digital rectal and sigmoidoscopy examina- tions with “usual care” provides some evidence that peri— odic screening for colorectal cancer reduces mortality from that disease (63). This study showed that the screened group had fewer overall colorectal cancer deaths (5 vs. 18, P <.05); fewer deaths among patients diagnosed during the study (4 vs. 13, P <.05); a more favorable stage distri- bution for cases found during the study (60% vs. 48% in situ or Dukes’ A lesions); and more favorable 5—year sur— vival rates for cases diagnosed during the study (80% vs. 48%). However, several factors limit the conclusions that can be drawn from this study. Because the study involved many diseases, a P-value of .05 for a particular end point (colorectal cancer deaths) cannot be interpreted to imply statistical significance in the usual sense. Secondly, the study group had an excess number of deaths from hemato- logic cancers and from suicides, which indicates to some observers a possible imbalance between the 2 groups with respect to some factor that might affect mortality from colorectal cancer. Evidence about the possible effectiveness of screening with sigmoidoscopy is also provided by 2 uncontrolled studies. One involved 18,158 persons who were given periodic sigmoidoscopy examinations, some of whom had polypoid lesions removed (64). The fact that only 11 colo- rectal cancers occurred among the screened subjects, com- pared with 75 cancers estimated from cancer registry data, provides evidence about the value of sigmoidoscopy in interrupting the possible polyp—cancer sequence. Further- more, the case-survival rates of colorectal cancers detected through screening were in the range of 90%, a rate much higher than would have been expected in an unscreened population. However, the interpretation of this study becomes con— fused when several facts are considered. First of all, the study only included subjects who had no signs or symp- toms, which biases any comparison with the population at large as captured by cancer registry data. Secondly, the prognosis of colorectal cancers diagnosed by sigmoidos- copy may be better than that for colorectal cancers diag- nosed between examinations (upon presentation of symp- toms) or than the prognosis for colorectal cancers that occur in unscreened people. Thirdly, biases relating to the selection of subjects (health conscious people may be more likely to attend screening programs and may have higher survival rates) and length-bias sampling (slow- growing cancers are more likely to be detected by screen- ing) also confuse interpretation of this study. In another study of a similar design, 28,126 persons were examined by digital rectal examination, sigmoidos— copy, and stool guaiac test for occult blood (65). Fifty- eight colorectal cancers were detected. The survival of 50 patients, who have been followed for more than 15 years, was nearly 90%, which is greater than the survival ex- pected in an unscreened population (66). This study, how- CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 ever, suffers from many of the biases affecting the pre- viously described study. Two controlled trials of the FOBT for colorectal cancer are currently in progress (67, 68). These studies indicate that the FOBT is capable of detecting adenomatous polyps and colorectal cancers in asymptomatic people and screen- ing can identify a higher proporttion of cancers in early stages. However, it is still too early for either of these studies to indicate the impact of screening on overall colo- rectal cancer mortality. A mathematical model of colorectal cancer screening has been developed that incorporates evidence of in— cidence and mortality rates, the anatomy and natural his- tory of colorectal cancers, the detection ability of various screening techniques, and the effectiveness of treatment. Application of the model suggests that an annual FOBT could reduce colorectal cancer mortality about 30%, a combination of annual FOBT and flexible sigmoidoscopy every 5 years could reduce mortality about 40%, and an annual FOBT plus barium enema every 5 years could reduce colorectal cancer mortality about 60% (69). Be- cause these estimates are based on indirect observations of colorectal cancer mortality, they are a weaker form of evi- dence about screening effectiveness. When reviewing all the available evidence, members of different organizations have come to different conclusions (22, 23). At this time, colorectal cancer must be classified as a type for which there is no general agreement that screening definitely reduces mortality. Bladder Cancer Although no controlled trial data are available, some researchers recommend periodic screening for people at high risk of developing bladder cancer because of occupa- tional exposure. However, questions have been raised about the relative effectiveness of early treatment with the use of currently available therapies for many of the lesions detected. At this time, there is no general agreement on the extent of impact of bladder cancer screening in reduc- ing mortality (24). Endometrlal Cancer Endometrial and ovarian cancers and other neoplasms of the corpus uteri are generally symptomatic at the time of diagnosis. Investigators still consider endometrial aspi- ration to detect endometrial cancer and pelvic examina- tion for ovarian and uterine tumors unproven as screening techniques. Esophageal and Gastric Cancers Reports from Japan and other countries indicate that early detection of esophageal and gastric cancers by endos- copy may reduce mortality from these diseases (70—73). However, because of their low incidence in the United States, no consensus has been reached that screening with endoscopy for these 2 cancers would be appropriate in this country. Serum Marker Assays Numerous assays have been developed and tested, and others continue to be developed. At present, for a variety of reasons, all assays have failed to be useful as screening techniques. The reasons relate to factors such as overall sensitivity for early stage disease, false-positive rates, and the rarity of some of the cancers being sought. 29 Earlier Dlagnosls by Earller Response to Symptoms The public’s increased awareness of signs and symptoms for cancer may lead to earlier diagnosis and improved survival. Efforts by health professionals to help people recognize and respond to signs and symptoms are also considered in the treatment section of this report for the following cancers: colorectal cancer (blood in stool), small cell lung cancer (multiple symptoms), cervical and uterine cancer (abnormal bleeding), and testicular cancer (testicu- lar pain or swelling). TRENDS IN SCREENING USE Level of Use Recent surveys (1980, 1983) document the current fre- quency of various examinations (table III-l). Other than Pap tests in younger women, the overall level of utiliza— tion of screening tests is low, and no trend toward in- creased utilization is apparent, which means that there is considerable room for improvement. Numerous impedi- ments keep the utilization of screening low; e.g., policy- holders of health care insurance are not generally reim- bursed for taking preventive measures. Nevertheless, even in situations when cost is not a barrier, utilization is incomplete. Lack of knowledge among the public and health care professionals about the value of screening procedures may limit utilization. Inconvenience and dis— comfort may also be factors. Any increase in the utiliza- tion of screening will require more knowledge about edu- cation and motivation of the public and health care professionals to use screening, as well as information about systems of incentives that would lead to wider utili— zation of screening procedures. Quality of Use Existing screening activities must be evaluated on an ongoing basis if quality of performance is to be ensured and impediments to wider utilization of screening are to be identified. The following paragraphs give some of the key points in technique that must be considered in any judging of training and effectiveness of the screening pro— gram (74). Breast Cancer Screening Attention to technical detail is important for optimal results in breast cancer screening. The physical examina- TABLE III-l .iObjectives for screening participation, 1990 and 2000 Recent par— ticipation, Objectives, . . %” % Screening Site —— —— and technique Age, yr 1980 I983 I990 2000 Breast Physical examination, every 50—70 43 45 70 80 year Mammogram, every year 50—70 14 15 45 80 Cervix 20439 76 79 85 90 40—70 65 57 70 80 Pap smear, every 3 yr " Data are from the “Public Awareness and Use of Cancer Detection Tests: 1980 and 1983 Surveys,“ The American Cancer Society. 30 tion should include inspection and palpation with the patient in three positions: supine, oblique, and sitting. Palpation should be done with the palmar surface of the digits rather than the fingertips. Mannequins for teaching are available. In mammography, one should aim for the lowest possible radiation dose while maintaining maximal resolution. Participation in the quality control programs of the American College of Radiology will help to achieve this balance between low dose and high resolution. Attain- able limits on radiation dose are 170 millirad for x-ray film mammography and 780 millirad for xeromammog- raphy (75). Some investigators prefer film screen mam- mography to xeromammography because of the lower radiation dose for film screen and anecdotal case reports in which abnormalities seen on a film screen examination could not be seen on xeromammograms. So-called phan- toms used to document maximal resolution are available, although none has been universally accepted. Cervlcal Cancer Screenlng The initial examination is visual inspection of the cer- vix. A spatula scrape 0f the cervical surface for one smear and an endocervical swab, or aspirate, for a second slide have been reported to increase detection by 26% to 40%, compared with a single slide method (76, 77). Quality con- trol considerations in the cytology laboratory include duplicate blinded reading of slides (ranging from all slides for newly trained readers to a statistically valid subsample for experienced readers) and retrospective review of pre— viously obtained slides on all subsequently diagnosed cases of carcinoma in situ or microinvasive or invasive cancer. Facilities for colposcopy should be available, particularly for evaluation of Pap smears interpreted as dysplasia. OBJECTIVES FOR SCREENING AND RECOMMENDED ACTIONS The objectives for increasing the levels of participation in screening by 1990 and 2000 are outlined in table III-l by cancer site. Each figure is the percentage of eligible per- sons screened according to the site, technique, and fre— quency. The upper limit on age for breast and cervical cancer screening was based on consideration of the in- creased risk of death from causes other than breast or cer- vical cancer after age 70. The participation in screening above that age depends on the individual and her overall health status. On the basis of the evidence available, the NCI recom- mends establishment of the following objectives for the year 2000: The percentage of women aged 50—70 who have annual physical breast examinations plus mammog- raphy should be increased to 80% from 45% for physical examination alone and to 80% from 15% for mammography. The percentages of women who have a cervical Pap smear every 3 years should be increased to 90% (ages 20—39) from 79% and to 80% from 57% (ages 40—70). Actions are recommended for eight categories of organ- izations, institutions, and other channels that can influ- ence the utilization of screening. These recommendations NCI MONOGRAPHS, NUMBER 2, I986 relate only to proven mass screening techniques for asymptomatic individuals. Recommendations related to the early detection of cancer for symptomatic or high- risk individuals are in the chapter on treatment. For screening, NCI’s role is to 1) guide and conduct research to develop and test the most efficacious screening tech- niques, 2) disseminate information to professionals and the public about the benefits of early detection through screening and the procedures for follow-up diagnosis when screening indicates likelihood of cancer, and 3) be a source of technical assistance and information for other agents interested in conducting screening programs. Government Agencies Government agencies at the federal, state, and local lev- els, as appropriate, are encouraged to consider the follow- ing recommended actions: Develop policies for wide application of efficacious cancer screening techniques. Establish a coordinating committee of government agencies which have the responsibility of facilitating planning and implementation of cancer screening. Agencies to be included are the NCI, the Centers for Disease Control, HCFA, and state and local health agencies. Establish policies for reimbursement for proven can— cer screening strategies, e.g., a prospective reimburse- ment system. Support research on cancer screening technology and on utilization of screening techniques. Facilitate and implement efficacious screening pro- grams for cancer. Assess present coverage of screening programs by health agencies and determine whether utilization of these services should be expanded. Ensure that state—of—the—art techniques and appro- priate quality control measures are used in screening programs conducted by health agencies. Industry Industry is encouraged to consider the following recom- mended actions: Manufacturers of efficacious screening technology systems should work with health planners and health financing agencies to implement their technology on a cost—effective basis. Manufacturers of technology useful for screening should advertise its value for prevention of cancer deaths and cite authoritative sources of background information. Employers should include screening for cancer ac- cording to specified protocols as a component of their health care packages. Professional Education Educators are encouraged to consider the following recommended actions: Emphasize in postgraduate programs of professional CANCER CONTROL OBJECTIVES FOR THE NATION: I98572000 organizations the benefits and techniques of screening for cancer. Target primary care physicians and other primary care health professionals (e.g., pharmacists, nurses work- ing in occupational health services) for educational campaigns about the importance of screening and early detection. Include in the curricula of medical schools and other health professional schools specific instructions about the scientific and technical aspects of cancer screening and have active cancer screening programs for dem- onstration to students. Include cancer screening in the training programs of medical schools and in community hospitals with residency programs. Include questions on both the scientific and technical aspects of cancer screening on National Board exam- inations in relevant medical specialties. Promote cancer screening continuing education pro- grams for health professionals. Make members of professional organizations aware of the latest information on cancer screening. Health Care Financing Third-party payers are encouraged to consider the fol- lowing recommended actions: Consider reimbursement of policyholders or provid- ers for screening techniques with proven efficacy through private policies, Medicare, and Medicaid. Remind clients about cancer screening recommenda- tions. Offer rate incentives to clients who participate in recommended cancer screening activities. Health Care Providers Health care providers are encouraged to consider the following recommended actions: Make available, throughout the health care system, appropriate facilities and expertise to meet the needs for screening the appropriate population segments. These services should 1) be readily available at reason- able cost, 2) be of high quality, and 3) have appro— priate quality control measures. Inform patients of the value of cancer screening and recommend utilization of efficacious screening pro— cedures. Voluntary Organizations Voluntary organizations are encouraged to consider the following recommended actions: Continue to expand efforts to increase the utilization of efficacious screening for cancer and develop pro- grams to overcome identified impediments to utili- zation. Develop and provide to state legislatures model legis— lation for state programs for cancer screening. 31 Inform and educate the public about the value of effi- cacious screening for cancer through contacts with organizations such as teachers’ associations and press clubs. Encourage health care organizations to provide low- cost screening programs and, through contacts with senior citizens’ associations, to alert their members to use them. Medla The full spectrum of the media is encouraged to con- sider the following recommended action: 32 Present information about the value of efficacious screening for cancer and actions to be taken by the public and by health professionals. The Publlc The public is encouraged to consider the following rec- ommended actions: Use appropriate cancer screening techniques. Work through unions, consumer groups, and coali- tions to increase availability and utilization of cancer screening techniques. IV. Treatment I'v— IV. Treatment TRENDS IN TREATMENT Recent data show that survival rates for cancer patients have been increasing. Data from the SEER Program of the NCI show a 49%, 5-year relative survival rate among all patients diagnosed between 1976 and 1981, compared with 48% for patients diagnosed between 1973 and 1975. Ear— lier NCI data (78) show a 42%, 5-year relative survival for patients diagnosed between 1970 and 1973 and a 38% sur- vival between 1960 and 1963. This general trend reflects better diagnostic capabilities and more effective treatments for cancer patients. The increase in survival is evident in a number of cancers, e.g., Hodgkin’s disease, melanoma, testicular cancer, and the childhood cancers (fig. [-2) Treatment has advanced since the early 1960s when multidisciplinary management emerged and radiation ther- apy and chemotherapy gained acceptance as potentially curative treatments. Cell kinetic studies first established dose—response relationships for radiation and chemother— apy, and disease staging was becoming the norm. In the late 19605, clinical studies showed that chemotherapy could be curative for certain cancers, even in patients with disseminated disease. In recent years the results from surgery, chemotherapy, and radiation therapy suggest that multidisciplinary treat- ment planning should result in even greater cancer patient survival and decreased morbidity. The analysis of mortal— ity changes outlined in chapter I is based on widespread application of current state-of-the-art therapies for pri— mary disease and micrometastases. The analysis indicates that by application of state—of-the-art therapies, cancer mortality could possibly be reduced by 10% to 20% and by another 10% through gains in treatment that may occur during the next 15 years, for a total of 30% by the year 2000. As stated in chapter I, it is estimated that aggressive widespread application of current state-of-the-art cancer treatment (i.e., without any additional advances in treat— ment) should result in about a 15% decrease in the cancer mortality rate within the next decade. Coupled with this reduction is the additional reduction which may be expected to result from continued advances in cancer treatment. Findings from ongoing clinical trials are at the core of these projections. For example, a recent review of the effects of adjuvant chemotherapy in stage 11 breast cancer estimates that a 32%—36% reduction in mor- tality can be achieved for premenopausal women; in addi- tion, 16%—18% mortality reduction for postmenopausal women can be achieved with hormonal therapy (79). The results of a clinical trial in pancreatic cancer suggest that 2-year survival is near tripled (from 15% to 42%) through treatment with combination chemotherapy and radiation therapy (80). For patients with limited-stage, small cell lung cancer, a combination of chemotherapy and radia— tion therapy resulted in a 33%, 2- to 4-year survival compared with a 10% survival for chemotherapy alone (81). With regard to colon cancer, further clinical trials are under way to confirm earlier positive results from an adjuvant chemotherapy study with 5-fluorouracil and levamisole (82). Advances in the treatment of ovarian cancer also have been recently reported, with the observa- tion that use of high-dose cisplatin and cyclophosphamide may mean that less radical surgery is necessary (83, 84). These recent, initial treatment advances are cited as a reminder that progress toward improved treatment and increased cancer survival is made gradually and, for most advances, on a site-by—site basis. If these positive initial findings are confirmed, these advances and continued expected gains in other sites hold promise for major reductions in cancer mortality beyond that now achiev- able through widespread application of current state—of— the-art treatment. However promising the results of advances in cancer therapy have been over the past few decades, significant changes in cancer mortality are not achievable unless widespread dissemination of treatment results and aggres- sive application of proven therapies occur. At present, the number of patients having access to state-of-the—art ther- apy is not optimal because of the great variability in expertise and resources in the health care delivery system. A primary concern of the NCI cancer control effort is to ensure that early detection and state-of-the-art therapies are available to as many cancer patients as is practical but especially to those with curable cancers. During the past decade, significant treatment gains have been demonstrated for various cancers, e.g., testicular, ovarian, breast, rectal, bladder, melanoma, small cell lung, adult leukemia and lymphoma, and childhood leukemia; this information should be accessible to the general population. On the other hand, some patients with unresponsive diseases con- tinue to be subjected to treatment proven to be ineffective; this practice should be stopped. POTENTIAL OF TREATMENT TO REDUCE MORTALITY Comparisons of SEER data with those from large phase III clinical trials show differences in treatment out— come. The SEER data describe average current treatment, whereas clinical trial data describe what may be consid- ered achievable state-of-the-art therapy. Although the pro- cess of patient selection for participation in many clin- ical trials often introduces a bias into the data (e.g., if the 35 subjects are younger and healthier than the general popu- lation), the analvsis of a series of trials in differing popula- tions corrects this distortion to some extent. The new PDQ system is an information system on cancer care through which NCI provides a useful consen- sus source for outlining state—of—the-art therapy and its outcomes. The treatment outcomes reported in PDQ are largely based on the results of large clinical trials which have been interpreted by a group of knowledgeable cancer specialists and summarized in consensus outcome figures. These outcome results represent a reliable, optimal stan- dard derived from protocol studies (specifying the patient workup and treatment) and conducted in institutions with adequate treatment capabilities (expertise and facilities). Differences between SEER results and clinical trial out- come data (representing state-of-the-art management) provide the goals for the suggested treatment objectives and actions for the specific cancer sites. The use of available state-of—the-art cancer therapies has been demonstrated to extend patient survival (85—87); however, no data exist for one to assume that widespread dissemination of treatment results and use of the most effective therapies have reached a desirable maximum level. References were made in the first chapter to differ- ences in mortality rates by geographic area and by race; in addition, survival differences by geographic area and can- cer site are evident in the SEER data. Different treatment practices may be a factor in these observed differences. Optimal treatment for all cancer patients requires a combination of appropriately trained professionals, coop- erating in planning and exercising multidisciplinary ther- apies, with adequate support services. The objectives of improvement in cancer control through efforts in the treatment area can best be achieved by actions which extend and improve 1) public awareness of the importance of early recognition of symptoms, medical attention, and the understanding that there are effective treatments for many cancers; 2) patient access to state—of—the-art cancer treatment, including early diagnosis of disease and prompt multidisciplinary treatment; and 3) application of state-of- the-art cancer patient therapy by health care profession- als, including adequate patient referral for multidisciplin- ary treatment and aggressive patient follow-up. Publlc Awareness Public perceptions about cancer are flawed or incom- plete. For example, the public has little awareness that cancer is a curable disease with an acceptable quality of life after treatment. Current statistics show that nearly 50% of all cancer patients will survive 5 years or more after diagnosis. Yet 65% of whites and 77% of blacks believe the percentage to be lower than 33%, and 9% of whites and 22% of blacks believe that even with early detection, there is “not much chance” of a cure. In addi- tion, only about 50% of whites and 25% of blacks can identify at least five of seven cancer warning signs (88). Delay in seeking medical attention is often longer for cancer symptoms than for those of other diseases; fear of cancer diagnosis and treatment, as well as a lack of knowledge about the importance of cancer symptoms, the benefits of early diagnosis, and the effectiveness of treat— ment are related to the delay. Although often construc- tive, the media may at times reinforce the public’s negative 36 perception of cancer by reporting unacceptable or exces- sive treatments for unresponsive cancers. The public should receive a balanced view and also be informed about the benefits of various types of proven, effective treatments for certain cancers. For example, patients with prostate cancer should be aware that combination treat- ment (surgery, radiotherapy, and hormonal therapy) is available and may be appropriate, depending on the stage of the tumor. Breast cancer patients should be aware of the two-stage procedure for treatment (biopsy with defini- tive diagnosis prior to definitive treatment) because of the procedure’s potential to reduce cancer morbidity (not mortality). When women are made aware of the need for definitive cancer treatment and the options available, they can be better prepared psychologically for the surgical procedure and may elect breast-saving treatment after dis— cussing the options. Women also should know about all aspects of treatment: available modalities (e.g., radical mastectomy, modified radical, lumpectomy with radiation therapy); the importance of nodal dissection; and the pos- sibility for reconstructive breast surgery. Breast cancer patients should understand the risks and benefits of sur- gery, primary radiotherapy, and postsurgical treatment options of adjuvant therapy (chemotherapy and hormonal therapy). Access to Treatment From comparisons of treatments and treatment out- comes among institutions, it can be inferred that many patients do not have access to state—of—the-art treatment (85, 86, 89, 90). Patient access to trained cancer specialists (medical, radiologic, pediatric, and surgical oncologists), adequate equipment, and other support services is not uniform. Many patients do not have financial resources to pay for available care, including consultations, drugs, treatment complications, and support services. Patient choice may enhance or retard the application of state-of- the-art treatment. Fear, lack of knowledge, religious be- liefs, and economic considerations are the most common reasons patients do not pursue treatment. These obstacles must be overcome, and cancer patients and their families must be made aware of treatment options and the availa- bility of specialized treatment facilities with trained pro- fessionals. They must be able to enter and negotiate the medical system to secure optimal treatment. Application The application of state—of-the-art cancer treatment is complex. At all levels of the health service delivery sys- tem, from the primary care physician who has initial con- tact with the patient to specialists directing the cancer treatment, physician knowledge of and training in state- of—the-art treatment are not yet optimal. Application should include not only knowledge of cur— rent information and techniques on the part of physicians but also an interest in ensuring early multidisciplinary decisionmaking, when appropriate to cancer treatment. For about 70% of the cancers, optimal therapy derives from multidisciplinary discussions. Treatment decisions made by physicians shortly after the diagnosis of cancer are critical in ensuring a cure for patients with responsive disease. NCI MONOGRAPHS, NUMBER 2, I986 Some of the most responsive tumors occur relatively infrequently, but proficient treatment for them can be maintained only where there are sufficient numbers of patients, i.e., at major cancer centers. Diseases such as the acute leukemias, DHL, and soft tissue sarcomas are rela- tively uncommon tumors that are often best treated in centers with specialized facilities and support services. For some physicians, economics may be a major barrier to their applying state-of-the—art care, especially when signif— icant disincentives block the referral of patients, e.g., the high costs of many tests, pressures to use local hospitals instead of referring the patient to a major cancer treat— ment center, the fee—for-service payment system, and some prospective payment systems that have the reimbursement for care fixed in advance. Malpractice considerations may result in physicians selecting “safe” therapy, which offers neither significant risk nor the chance of cure. A major determinant of outcome for most newly diag- nosed cancer patients with curable disease hinges on early multidisciplinary treatment planning and the availability of expertise and resources to conduct such a treatment plan. Obstacles in the health care system that impede the widespread use of multidisciplinary, state-of—the-art ap- proaches to patient care often are interrelated. Actions for any change in the current cancer care system for reduced cancer morbidity and mortality must address these fac- tors. Interventions in the areas of public and professional education, institutional controls, and regulatory and fiscal changes will be necessary components of cancer control efforts. TREATMENT OBJECTIVES The identification of differences between observed out- comes from SEER data and those of state-of-the-art ther- apy provides a basis for establishing treatment objectives and actions. Site-specific objectives, developed by the Treatment Working Group and NCI staff based on these data, are listed in table IV—l for the years 1990 (interme- diate objectives) and 2000. The two summary objectives for the year 2000 are: Increase adoption of state-of-the-art treatment. Continue to advance treatment, as reflected in in- creasing cancer survival. The specific year 2000 objectives are expressed as in- creases in 5-year observed survival rates' based on appli- cation of state-of—the-art technology and treatment for each cancer site. In general, the year 1990 intermediate objectives reflect the care decisions associated with opti— mum patient management, i.e., the diagnostic and thera- peutic decisions expected to give the best prognosis for increased survival. For example, an appropriate multidis- ciplinary workup plus prompt referral for appropriate surgical interventions for children with brain tumors and women with ovarian cancers would improve their survival and decrease morbidity. In addition, the diagnosis of most cancers at the earliest stage, such as rectal and pros- lObserved rates were used by the Treatment Working Group in its deliberations because many of the results are reported in the clinical trial literature in this manner. CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 tate cancers, is associated with much higher survival rates. Generally, clinical trial data that show gains in survival from mass screening practices are available only for two sites (cervix and breast, chapter 111). The use of the digital rectal examination, however, is considered an essential part of good patient management (23, 91). The early diagnosis of rectal and prostate cancers, in part through the appropriate use of rectal examinations in persons at high risk of one or both of the cancers, notably those over 50 years of age, could improve the outcomes for these patients. The relative survival from melanoma when de- tected in its earliest stages is above 90%. The difference in outcome achievable by state-of-the—art compared with cur— rent average therapy should be brought to the attention of practitioners, and the underlying factors contributing to nonoptimal treatment decisions should be identified. State-of—the-art survival has been estimated as of 1985; however, gains in treatment research are expected to lead to gains in survival. Various scenarios are outlined in con- cept in figure IV-I. Survival has been gradually increasing for all cancer since 1938. This increase is reflected in the lower curve in the figure. State-of—the-art survival pre- sumably leads (or exceeds) current survival in any year. The treatment goal is that current survival be brought as close as possible to the state-of-the-art survival by the year 2000. If state-of-the-art survival continues to increase and if survival for the average cancer patient can be brought closer to it through improved therapies and diffusion efforts, then outcome D occurs. Similarly, if the gap between current survival and state-of—the-art survival re- mains constant, and there is no improvement in the state— of-the-art treatment, then outcome A occurs. Various sce- narios for this increase in survival (ranging from no change to an increase of 1.5% per year) are considered in the analysis outlined in Appendix A. Table IV-2 shows the estimated reductions in mortality that will result if these objectives are achieved. The reductions are conser- vative in that table IV-2 figures were calculated with no assumption of any increase in the state-of-the-art survival figures beyond the current ones. RECOMMENDED ACTIONS FOR MEETING THE TREATMENT OBJECTIVES A common set of actions can be taken to improve treatment for some cancer sites, but more specific actions and those specific to high-risk target groups are needed if we are to ensure successful management for other cancer sites. Recommended actions follow for several areas: pub- lic education, professional education, intrainstitutional strategies, standard setting, and fiscal practices. The rec- ommendations specify a particular cancer site if a unique intervention, target group, or specific plan is proposed. (Recommendations related to breast and cervical screen- ing programs are made in chapter 111 of this report.) For treatment, NCI’s roles are to 1) guide and support re- search, especially clinical trials to determine the efficacy of various treatments; 2) analyze the impact of treatment; 3) work with government agencies and professional organi- zations on setting standards for treatment; and 4) ensure that effective treatments are made known to the medical community promptly and effectively. 37 TABLE lV-lt—Treatment objectives for 1990 and 2000 Cancer site/type Breast Colon Bladder Lung, small cell Cervix Corpus uteri Ovary Rectum Testis, nonseminoma Prostate Adult leukemia Adult non-Hodgkin’s lymphoma Melanoma Childhood brain tumors Childhood leukemia Year l990 intermediate objectives: Desired increase or decrease” in specified action, %” Patients staged from 50 to 100 No. of two-step procedures (biopsy with definitive pathology prior to surgery) from 45 to 80 Patients diagnosed in early stage from 40 to 70 Morbidity in stages I and [I through knowledge of conservative treatment options“ Use of adjuvant combined chemotherapy in stage ll (positive axillary nodes) premenopausal women from 65 to 90 Use of adjuvant hormonal therapy in receptor positive stage II (positive axillary nodes) postmenopausal women from 50 to 90 Multidisciplinary management with systemic therapy followed by RT, surgery, or both, for stages Ill and IV Patients adequately staged through the application of state-of-the-art surgical and pathologic techniques No intermediate objective Complete response in patients with limited disease to 80 through state-of—the-art multidisciplinary management Use of intracavity radiation to 90 in stage II and 80 in stage III Screening to detect disease at an earlier stage Five-yr observed survival for black women to the rate for white women through state-of—the-art multidisciplinary management Complete response through aggressive surgery, chemotherapy, and radiation Disease-free survival in stage I to 90 by state-of—the-art management Local—regional recurrence at 3 yr with combined modality treatment (RT and surgery): stage II from 35 to 5, stage II] from 60 to l5” Proportion of patients diagnosed at stages I and II by provision of routine rectal examination; for black patients diagnosed at stage ll from 16 to 21 Cure for all stages, stage 111 from 8] to 90 through aggressive multidisciplinary patient management Local control at 3 yr through state-of-the-art multidisciplinary management: stage B to 90, stage C to 75 Proportion of patients diagnosed at stage B from 4| to 65 by provision of routine rectal examinations Complete response to 60 at 2 yr and cure rate to l5 through state-of-the-art management Cure of patients with DHL through use of state-of—the-art treatment to 60 Patients diagnosed at stage I from 44 to 60 Proficiency in multidisciplinary management and in timely referral to centers Cure of patients to 50 through state-of—the-art management and provision of proficient management and support services to patients Year 2000 objectives: Desired increase in S—yr survival, %‘ From 66 to 72 for post- menopausal women From 41 to 51 From 55 to 58 From 9 to I54 From 61 to 67 From 46 to 76 for blacks From 33 to 37 From 38 to 59 From 78 to 94 (all stages) From 48 to 59 For AML, from 5 to 25; for ALL, from 15 to 20 From 33 to 60 for DHL From 75 to 82 For cerebellar astrocytoma, from 90 to 95; for astrocy- toma stages II and 1V, from l0 to 40; for medul- loblastoma, from 40 to 80 For AML, from 2! to 40; for ALL, from 65 to 75 " Desired objective is a decrease in morbidity and recurrence, respectively. b Intermediate objectives apply to all patients, unless otherwise stated. ‘ All survival figures refer to observed survival. d Desired increase in survival is for 2 yr. Public Education Channels Public education for treatment objectives refers to strate- gies recommended by the Treatment Working Group to inform the public about the symptoms of cancer and actions people can take to improve their prognoses when cancer develops. Public education programs should em- phasize the importance of people being able to recognize 38 early warning symptoms of cancer, the seeking of a prompt diagnosis, the curability or improved survival rate when certain cancers are detected and treated early, the appro- priateness of seeking information about treatment options, and the appropriateness of seeking referrals to cancer spe- cialists and subspecialists. Public education is not limited to particular communication channels; rather, the media, schools, organizations, and health care professionals (phy- NCl MONOGRAPHS, NUMBER 2, I986 SURVIVAL IMPROVEMENT AND INCREASED DIFFUSION ‘ N0 IMPROVEMENT AND 0 CONSTANT DIFFUSION NO IMPROVEMENT OR DIFFUSION FIGURE IV-1.—Trends in survival: Possible scenarios for state-of- the-art treatment and SEER-re- ported survival. /” / ll - x ,9 ’— x Z 60 '— I/ // uJ ’ I U z’ z’ a: , I 3.” STATE~0F—THEEART e . TREATMENT a ........ 2 so — "“" > i: D m E REPORTED END RESULTS GROUP E] w 40 .— SURVIVAL >,' SEER o In ESTIMATED STATE—OFETHE ART A 4L 1 1 l J 1 60-63 70-73 73-76 77-80 1990 2000 YEAR sicians, pharmacists, and public and occupational health nurses) are all possible channels and should be involved as appropriate. In regard to early detection and diagnosis, public edu- cation channels are encouraged to consider the following recommendations: In public education efforts, target high-risk persons, such as those with a strong family history of cancer of the breast, ovaries, and colon, and melanoma; ado- lescent boys and young men (testicular cancer); and those exposed to known environmental or occupa— tional carcinogens. In educational efforts regarding cervical cancer, in- clude information for women of all ages about risk TABLE IV-2.—Estimates of lives saved in the year 2000 by provision factors for the disease, the importance of follow-up for abnormal bleeding, and the availability and use- fulness of the Pap smear test. Include as a target group high-risk women, i.e., those having frequent intercourse with multiple sexual partners (beginning in their teenage years), who have a history of dyspla- sia, or who are in the lower socioeconomic level. In public education campaigns about uterine cancers, emphasize the importance of pelvic examinations and follow-up for abnormal bleeding, particularly for postmenopausal women, especially those with lower SES and those who are obese. In public information programs about colon, rectal, and prostate cancers, emphasize the high risk of these 0f stale—of-the-art therapy for selected cancer sites” No. of deaths No. of deaths under in year 2000 state-of—the—art Site/type under 1985 survival treatment No. of lives saved Reduction, % Prostate 36,685 21,266 15,419 42.0 Breast 53,023 45,465 7,558 14.3 Colon 51,826 41,450 10,376 20.0 Rectum 22,466 13,641 8,825 39.3 Lung, small cell 40,376 37,407 2,969 7.4 Bladder 13,618 11,625 1,993 14.6 Melanoma 19,778 14,891 4,887 24.7 Cervix 2,872 2,089 783 27.3 Corpus uteri 2,604 1,907 697 26.8 DHL 3,093 1,377 1,716 55.5 Ovary 12,030 11,114 916 7.6 Testis, nonseminoma 477 320 157 32.9 Childhood ALL 852 565 287 33.7 AML 195 149 46 23.6 Medulloblastoma 601 301 300 49.9 Adult ALL 6,223 4,745 1,478 23.8 Subtotal 266,719 208,312 58,407 21.9 All other sites 374,906 374,906 0 0.0 Total all sites 641,625 583,218 58,407 9.1 " Prevention and screening activities are held constant at 1980 levels. CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 39 cancers for adults aged 50 and above, the importance of early recognition of symptoms, the value of ag- gressive follow-up for occult blood (colon and rectal cancers), and the importance of rectal examinations as part of a routine physical examination. Consider innovative channels of communication, e. g., mailing of information on colon, rectal, and prostate cancers with Social Security checks so as to reach retired elderly persons; and distribution of literature at retail stores and pharmacies. In regard to treatment choices, public education chan- nels are encouraged to consider the following recommen- dations: In education programs, emphasize that patients should become aware of options available for treatment of certain cancers, including the possible availability of clinical trials for evaluation of potential advances in treatment. Physicians should be available to help diagnosed cancer patients and their families identify available treatment options. In public information programs, emphasize the po- tential curability of childhood leukemia and other childhood cancers when patients receive appropriate treatment at specialized centers. Professional Education Channels Professional education regarding cancer includes actions to inform physicians and other primary health care pro— fessionals of advances in the state of the art as well as information on risk factors, risk factor reduction, and other cancer issues. Channels for professional education are encouraged to consider the following recommenda- tions: Direct programs through cooperative groups, cancer centers, and professional organizations at increasing knowledge of treatment alternatives and combina- tions through collegial decisionmaking and multidis- ciplinary management for the following cancers: Site / type Treatment Rectum Surgery, radiotherapy Breast Surgery, radiotherapy, chemotherapy, including adjuvant and hormonal Prostate Surgery, radiotherapy Bladder Surgery, radiotherapy Lung, small cell Chemotherapy, radiotherapy Uterus Surgery, radiotherapy Ovaries Surgery, chemotherapy Adult leukemia and Chemotherapy, support services lymphoma Childhood leukemia Chemotherapy, support services In educational programs for specialty physicians (surgeons, urologists, gynecologists, gastroenterolo- gists), emphasize the importance of proper multidis- ciplinary management of cancer patients; aggressive workups (radiology and endoscopy) for occult blood; and disease staging before treatment for all cancers, 40 especially rectal, colon, breast, prostate, small cell lung, cervical, uterine, ovarian, and skin (melanoma) cancers. Encourage these specialty physicians and oncologists to participate in clinical trial activities where avail- able. In strategies for disseminating information to profes- sionals, include all the traditional methods, such as professional meetings, journals, special continuing medical education symposia and consensus confer- ences, revised medical school curricula, and residency training. Use the PDQ system as a vehicle for dissemination of the most up-to-date cancer treatment information to physicians. lntralnstitutlonal Strategies lntrainstitutional strategies are to be initiated at the hospital, health plan, health planning, or other institu- tional levels; they are meant to ensure that patients receive optimal care. Treatment decisions by physicians made shortly after the diagnosis of cancer are critical to ensur- ing 1) the best opportunity for those patients with curable disease and 2) decreased morbidity for those in whom long disease-free intervals are possible with appropriate treatment choices. Because many of the most responsive diseases are those in which the morbidity may be mini- mized, a multimodality decisionmaking process invoked shortly after diagnosis is optimal. For approximately 70% of cancers, some multidisciplinary discussions before the first therapy is administered would be an optimal ap- proach. Because some of the diseases with the greatest chance of cure are relatively rare or require extraordinary management, referral should be considered if the expertise and resources are insufficient for the patient to receive state-of—the—art therapy. Strategies at the institutional level that will guarantee adequate reimbursement and promote early consultations and referrals, if appropriate, are essen- tial to achieve the best cancer treatment results. Institu- tions are encouraged to consider the following recom- mendations: In collaboration with the American College of Sur- geons’ Commission on Cancer and the Joint Com- mission on Accreditation of Hospitals, hospital offi- cials should develop strategies for the medical staff and cancer committees to evaluate critically their available resources and develop a realistic assessment of the ability of that hospital’s staff to provide optimal care for cancer patients, site by site. This hospital cancer plan should become a requirement for a cancer program or hospital accreditation, or both. The relative rarity of some of the most responsive cancers, such as the acute leukemias, DHL, and soft tissue sarcomas, means that proficient, state—of—the- art treatment is best maintained in major cancer cen- ters and that patients be referred to those centers. Hospital management should consider limiting hospi- tal privileges in regard to cancer treatment for physi- NCl MONOGRAPHS. NUMBER 2, I986 cians who do not maintain proficiency in cancer treatment. Enhance professional oncology education through the PDQ system, oncology nursing support, treatment guidelines, protocol participation, and quality assur- ance requirements by the Joint Commission on Ac- creditation of Hospitals. Upon admission, provide a routine breast examina— tion and a Pap smear (if the patient is of the appro- priate age and has not had one in the preceding 3 yr) as well as, if appropriate, a rectal examination. Re- quirements for staging prior to initiation of definitive therapy should be established. Require discussion of treatment options with patients. These requirements can be a major force for encour- aging questions from patients and their seeking knowledge of the most current treatment options from physicians. Standard Setting and Fiscal Policies Standards and fiscal policies include actions for estab- lishment of guidelines, professional standards, or legal requirements relating to medical practice. Fiscal policy refers to economic actions related to either reimbursement by the patient, insurance, or other third—party payer, i.e., generally, the financing of health care. Appropriate chan— nels are encouraged to consider the following recommen- dations: Improve and enforce professional standards to the extent possible through peer review and professional organizations, including technical proficiency guide- lines, continuing education requirements, and recerti- fication and accreditation requirements. Develop guidelines for referral of patients with cancer at specific sites to appropriate institutions for multi— disciplinary care and for the minimal resources neces- sary to treat cancer patients. Provide adequate health care insurance to all Ameri- cans, especially coverage for preventive services, comprehensive diagnostic workups, two-stage surgi- cal procedures for relevant sites (definitive biopsy, then surgery), catastrophic illness costs, and rehabil— itative services. Offer reimbursement incentives to health care provid- ers, including actions that would ensure reimburse- ment for comprehensive diagnostic workups, adequate clinical staging prior to treatment, appropriate multi- disciplinary referrals, and application of state-of-the- art treatment. A final strategy that deserves further exploration is that of state standards. For example, it has been suggested that if states require discussion of treatment options prior to breast cancer surgery, there will be greater involvement of patients in treatment decisions and more multidisciplinary involvement in treatment. For the recommended actions, table IV-3 specifies the agent for initiating the action and a target group to implement the recommended intervention. For example, the HCFA and other third-party payers are the appro— TABLE lV-3.—Specific approaches for optimal treatment Intervention Agent Action Target groups Improvement of standards for medical oncology American Society of Clinical Oncology and specialty boards Improvement of standards for surgical proficiency Professional associations, cancer centers, hospitals, and specialty boards Improvement of standards for RT Professional associations, specialty boards Requirement for staging for reimbursement with incentives for referral Provision for universal cata- strophic coverage Third-party payers, HCFA, and self-insurance plans Preferred provider lists and programs Second-opinion programs Third-party payers, HCFA, and self-insurance plans Reimbursement for proper Private health insurers, HCFA workups (e.g., occult bloods, pathology for melanoma, biopsy, mammography) Reimbursement for multi- disciplinary consultations Requirement for medical oncology consultation with surgical procedures Accreditation, quality assurance requirements Hospitals and physicians 1) Professional recognition 1) Physicians 2) Accreditation, hospital privi- leges, quality assurance requirements 2) Hospitals and physicians Accreditation, quality assurance requirements Hospitals and physicians Reimbursement and regulation Hospitals and physicians Reimbursement and regulation Patients, hospitals, and physicians Reimbursement and regulation Physicians CANCER CONTROL OBJECTIVES FOR THE NATION: l985 2000 41 priate agents to facilitate regulatory and reimbursement policy changes to provide greater incentives to physicians for patient referral. Hospitals and private physicians would be the appropriate target groups for efforts by HCFA and third-party payers. However, it should be noted that, although quite effective, reimbursement restrictions aimed at assurance of optimal state-of—the-art therapy are diffi- cult for the agents to apply in a consistent manner and are unlikely to be received readily by most health profession- als. Interventions for improvement and enforcement of 42 professional standards might be undertaken by profes- sional associations, hospital associations (e.g., the Ameri- can Hospital Association), and cancer centers. Actions to be taken by these groups are those which would affect professional recognition, accreditation, and the granting of hospital privileges to private physicians. It should be emphasized that 1) cancer is a complex disease requiring collegial decisionmaking, and 2) obtaining consultation and referrals on the part of the physician is a positive rather than a negative behavior. V. Surveillance V. Surveillance The purpose of a surveillance system for cancer control is to monitor progress toward achieving the objectives and goal of a 50% reduction in the age-adjusted cancer mortal- ity rate by the year 2000. Measurement of progress re- quires a series of indicators and corresponding data collection efforts. For some indicators, such as cancer mortality, data are being collected on a systematic, on- going basis. For others, researchers and planners can only obtain data by modifying or augmenting existing systems. For still others, however, new data collection mechanisms must be established. In this chapter, the Working Group explains a basis for decisionmaking by the NCI and other federal agencies about data collection systems and surveil- lance activities; the information presented here is also a useful framework for the development of regional cancer control indicators. The basic surveillance strategy is not only monitoring of mortality from specific cancers but also changes in cancer incidence; cancer patient survival; the prevalence of var- ious risk factors; medical practices affecting cancer out- come; participation in cancer screening; access to medical care; behavior related to persons seeking medical care; knowledge, attitudes, and beliefs of the public and profes— sionals that are related to cancer; measures related to facilities for the treatment of cancer; and financing of the associated medical care. Some of the indicators needed for this effort are already clear, but if others are to be formu- lated, the objectives must be made more specific. The fol- lowing sections discuss these indices and the issues related to the data systems necessary for their construction. SURVEILLANCE INDICATORS The following is an outline of the types of indicators needed for the overall surveillance effort. The list is not intended to be exhaustive but rather to be illustrative of objectives within the major areas of cancer control. This outline is intended to serve as background for the identifi- cation, planning, and integration of the data systems needed to construct these indicators. I) Indicators related to cancer control outcomes A) Age—, sex-, race-specific and age-adjusted mor- tality rates for specific cancers B) Age-, sex-, race-specific and age-adjusted inci- dence rates for specific cancers C) Observed and relative survival rates among patients with specific cancers, for specified lengths of time after diagnosis D) Distribution by stage at initial diagnosis 11) Indicators of cancer screening activities' A) Cervical cancer: proportion of eligible women aged 20 to 70 years receiving a Pap test for cancer of the cervix within the past 3 years B) Breast cancer 1) Proportion of asymptomatic women aged 50 to 70 years receiving mammography and / or physical examinations for unsuspected breast cancer within the past year 2) Proportion of high-risk women aged 40 to 49 years receiving mammography for breast cancer within the past year 3) Proportion of newly detected cases of breast cancer diagnosed as carcinoma in situ and the proportion diagnosed as stage I 4) Proportion of women conducting breast self-examination C) Rectal cancer: proportion of population aged 50 to 70 years receiving digital rectal examina- tions, FOBT, and sigmoidoscopy within the past 3 years D) Prostate cancer: proportion of men over 50 years of age receiving digital rectal examina- tions within the past 3 years E) Other cancers 1) Possible screening for neoplasms of the oral cavity, esophagus, stomach, testis, and blad— der, and melanoma 2) Development of indices following new deci- sions on screening efforts F) All cancers: decrease in proportion of late- stage diagnoses as an indicator of successful screening III) Indicators of cancer prevention activities A) Smoking 1) Division of percentage of population age 20 and over who are smokers by sex, geo— graphic region, race, and by socioeconomic category into heavy smokers (>25 cigarettes / day) and all others lSome of the following indicators refer to screening activities not recommended in chapter [11; however, the Working Group believed that these indicators should be included to identify the full range of screening activities under way in order to assess the direction of, and resources devoted to, screening activities. 45 2) Percentage of population under 20 years of age who are smokers (categorized as above) B) Diet 1) Proportion of caloric intake attributed to fat for a cross-section of the population 2) Average number of grams of fiber consumed per day for a cross-section of the population 3) Average number of servings per day of fruits and vegetables rich in carotenoids and as- corbic acid overall and by age, race, sex, and SES 4) Proportion of the population that is over- weight by age, race, and sex categories C) Occupation: Measures taken (categorized by industry-related factors) of occupational expo- sures in specific geographic areas IV) Indicators of cancer treatment2 A) Proportion of cancer patients treated by the various medical specialties B) Number of patients treated by type of hospital and facility, including those involved in clinical trials C) Breast cancer 1) Proportion of premenopausal women with stage 11 breast cancer receiving adjuvant che- motherapy 2) Proportion of postmenopausal women with stage II, hormone receptor—positive breast cancer receiving adjuvant hormonal therapy D) Cancer of the cervix uteri 1) Five-year survival rate 2) Proportion of women with stage II cervical cancer receiving intracavity radiation 3) Proportion of women with stage III cervical cancer receiving intracavity radiation V) Measures related to knowledge, attitudes, beliefs, and behavior concerning cancer3 A) Proportion of the population (by sex, race, SES) with knowledge of curable cancers B) Proportion of women knowing the importance of Pap smears C) Proportion of women knowing the importance of breast screening (physical examination and mammography) VI) Resources, financing, and cancer costs A) Resources 1) Proportion of population served by clinics and/or programs for Pap smears (i.e., treat- ment other than in private physicians’ of- fices) 2) Number and distribution of oncologists per unit population 2Indicators must relate to treatment adequacy and patient and physician access to and use of state—of-the-art therapies. 3A number of measures are possible and should be tailored to the objectives and programs implemented to improve public and professional awareness concerning cancer. 46 B) Financing 1) Proportion of women covered by insurance plans providing reimbursement for mammog- raphy and Pap smears 2) Extent of reimbursement for pathology with biopsy for melanoma 3) Extent of diagnosis-related group coverage for patients in community hospitals in clini- cal trials C) Cancer costs 1) Estimated direct cost of cancer by site 2) Estimated indirect cost (value of productiv- ity losses for people ill and disabled from cancer and those who die prematurely) by site DATA SYSTEMS A number of data systems exist within the NC], NCHS, and elsewhere within the federal government; state and local governments; and in the health care industry that could be coordinated to produce the indicators needed for monitoring progress toward the objectives. This section briefly describes these systems, their current status, and the limitations of each data system for this purpose. Cancer Mortality Each year, data tapes are available from the NCHS on all deaths in the United States. Statisticians use these data to monitor trends in mortality from specific cancers over time. Thus the data form the centerpiece for the monitor— ing of progress toward the goal of a 50% reduction in cancer mortality rates by the year 2000. One of the limitations of this system is the basic prob- lem of timeliness; the data on deaths occurring in 1981 became available in mid-1984; for 1983, data became available in the fall of 1985. Although an improvement, further efforts are needed to make the 100% mortality data available on a much more current basis. The 10% monthly mortality sample could provide more current data, but the categories of cancers used in the tabulations produced from the sample are too broad to be useful. Personnel of other agencies and the NCHS are now dis- cussing ways of modifying the categories of cancer in the monthly reports and making the final data on mortality more timely. Another limitation of mortality data is the degree of accuracy of cause of death information now under study by the NCHS and other agencies. The NDI, established by the NCHS, allows researchers to determine whether a given person has died and to fol— low large groups of persons to determine their survival. The NDI continues to improve the algorithms applied to match name and other personal and demographic data on record with corresponding information on decedents as reported by the 50 states. In such a system, matches will not be perfect, but accuracy continues to improve. The NDI may be a highly cost-effective means for cancer patient follow-up. Cancer Incidence and Patient Survival Surveillance, Epldemlology, and End Results Program Data on every case of cancer diagnosed among resi— dents of 11 geographic areas, accounting for over 12% of NC] MONOGRAPHS, NUMBER 2, 1986 the United States population, are obtained continuously through the SEER Program. Data are accumulated from six states, i.e., Connecticut, New Jersey, Iowa, New Mex- ico, Utah, and Hawaii; the four large metropolitan areas of Atlanta, Detroit, San Francisco, and Seattle; and the Commonwealth of Pucrto Rico. The Program provides detailed systematic data on cancer incidence and cancer patient survival. As of late 1985, incidence data are avail- able on patients diagnosed through 1984, and survival rates are being calculated from follow—up data through 1983. Inasmuch as the SEER data are accumulated only for the areas mentioned above, they cannot provide direct estimates for other geographic areas. Because the medical record is the basic source of the data, the system does not lend itself to the collection of good, systematic data on such factors as smoking status and occupation. Some of the data now included have limitations, e.g., the amount of chemotherapy given. Some treatments given in doctors’ offices also are not included. Data on Hispanics are not fully representative of the United States, and rural blacks are underrepresented. Other Population-based Cancer Registries In addition to SEER, 12 population-based cancer regis- tries in the United States attempt to access data of entire states. Los Angeles County also has a population-based cancer registry covering a population of over 7 million people. These areas account for a total of 82 million peo- ple, or about 36% of the total population. Only 6 of these registries, serving a population of 4.5 million, also obtain data on cancer patient survival. The data compiled by these registries represent various degrees of completeness and quality. For the most part, the data are not comparable across the various registries, nor are they compatible with those of the SEER Program. Hospital Cancer Registries As of April 1984, the American College of Surgeons had certified 1,055 hospitals in the United States as having “approved” cancer programs. Among the requirements for approval is that the hospital must have a cancer registry that meets specific standards. Although there is no provi- sion for these hospitals and registries to report data rou- tinely for all forms of cancer to the American College of Surgeons, they do participate on a voluntary basis in patient care evaluation surveys in which they provide abstracts of cases of specific cancers through the Ameri- can College of Surgeons’ Patterns of Care Studies. Thus these hospital cancer registries provide researchers and planners with an opportunity to obtain data on cancer patient survival from groups of hospitals in areas outside those covered by the SEER Program. These data cannot be related to specific populations, nor are they uniformly and readily available. Some of these registries are computerized; others are not. Defini— tions, procedures, and quality vary. Therefore, a data col- lection scheme is essential if they are to be used system- atically. Other Patient-oriented Systems Medicare Statistical System Medicare covers 95% of the population aged 65 years and older and also provides coverage for some disabled CANCER CONTROL OBJECTIVES FOR THE NATION: I985 2000 persons under age 65. It covers certain services provided by hospitals, physicians, hospital outpatient facilities, skilled nursing facilities, and home health agencies. Be— cause this statistical system is a by-product of the opera- tion of the Medicare program itself, its basic files contain three elements: 1) an enrollment file containing basic demographic information on every enrollee, 2) a file of service providers with information on numbers of beds by type and geographic location, and 3) hospital bills based on the claims system. Data from all three of these sources are linked to create records of hospital stays. Based on the Medicare files, the statistical system now consists of two main files: 1) hospital stays which include information on diagnoses, procedures, and dates of admission and dis- charge, as well as information on charges and characteris- tics of both the enrollees and providers; and 2) the contin- uous Medicare history file, which combines all information on an individual in one record for a 5% sample of Medi- care enrollees. Beginning in 1985, data files were estab- lished on Medicare, Part B, that include a file on a 1% sample of physicians, a file on a 5% sample of enrollees, and a 100% file of procedures. This system is limited in value because it pertains only to persons who are 65 years old or over and some disabled persons under 65. National Hospital Discharge Survey This survey, conducted by the NCHS, is a probability sample of patients discharged from hospitals in the United States. It provides annual data on up to seven diagnoses and up to four procedures for each patient discharged. Because the survey is based on hospital episodes, it does not provide estimates of cancer incidence but it helps estimate the volume of medical care for cancer patients. It cannot be used to provide estimates for small areas unless the sample is augmented in a given area. National Ambulatory Medical Survey This NCHS survey is a probability sample of visits to doctors’ offices by individuals; it provides information on diagnoses, associated procedures, and prescriptions related to each person. These data provide a basis for estimates of the volume of outpatient care for cancer and outpa- tient administration of chemotherapeutic drugs. The survey is based on visits and therefore cannot be used for estimates of total care provided to individuals, nor can it provide data for small areas. Furthermore, the survey is conducted only every 3 years and thus cannot provide annual data. Surveys of the Centers tor Disease Control At present, personnel at the Centers for Disease Con- trol are implementing an automated record-based report- ing system in which information will be reported on-line from computers in the field to a central computer. Thus information can be obtained on specific diseases or condi— tions on a highly current basis. Also sampled weekly are 121 cities for reports on all deaths (and on pneumonia and influenza deaths in particular) for 25% of the population and 33% of the deaths. It is possible that each of these mechanisms could be applied to at least rare cancers. These surveys have been traditionally oriented to infec- tious diseases. Whether the reporting of a chronic disease such as cancer would be adequate remains to be tested. 47 SURVEYS OF THE GENERAL POPULATION Many of the cancer control indicators involve mea- surement of levels of risk factors, behavior, knowledge, etc., in the general population. A number of surveys of the general population already exist that obtain information on some of these factors and also provide a basis for obtaining additional information that will produce useful indicators. One limitation which cuts across all these sur- veys, however, is the fact that these are probability sam- ples of the population and therefore are limited in the extent to which they can provide estimates for specific geographic areas or subpopulation groups. The following are the major surveys: National Health Interview Survey The NHIS is conducted continuously during the year and annually includes about 40,000 households represent- ing between 100,000 and 110,000 individuals. Current plans call for an increase in the sample to 50,000 house- holds and 140,000 individuals. Detailed information on the health, use of medical care services, and disability sta- tus of the population, as well as some aspects of health behavior, are obtained. In addition, data on the cigarette smoking status of the population have been obtained every 2 or 3 years as part of a supplement to the main survey. Because questions on smoking status and on other vari- ables of interest for the cancer effort are not included as part of the core questionnaire, arrangements must be made to have such questions included in supplements every year or 2 years, depending on how often the data would be needed. National Health and Nutrition Examination Survey The NHANES, conducted by the NCHS periodically on a probability sample of the population, obtains informa- tion on health and dietary intake. As part of a systematic physical examination on this sample, body levels of var- ious nutrients are measured through analyses of serum and urine specimens. This is the only large national survey in which dietary, health, and medical information is simul- taneously collected, as well as other risk factor data on the same individuals. The NHANES will provide valuable information for monitoring changes in the dietary habits of the popula— tion. It will consist of two consecutive national samples of 15,000 persons each and will be conducted in 1988—91 and 1992—94, with oversampling of blacks and Hispanics. Fu- ture plans call for the NHANES to be an almost continu— ous national health and nutrition survey. Current Population Survey The Bureau of the Census administers this ongoing sur— vey of a large probability sample of the population. Through a multiagency effort, this survey has been matched against the NDI to relate population characteris— tics to mortality. Current plans call for the Bureau to conduct this match each year, using the March sample. Inquiries on smoking and other risk factors could be added to the survey questionnaire. This sample provides estimates for the United States, but breakdowns for specific geographic areas are not read- ily available. 48 Survey 01 Income and Program Participation This is an ongoing longitudinal survey of a large proba- bility sample of the population, also conducted by the Bureau of the Census. A panel of 20,000 households, approximately 55,000 persons, is followed for a 2-year, 8- month period. A new panel of respondents is begun each January. Detailed and repeated measures of demographic and socioeconomic characteristics are collected. Current plans call for supplements on health care and disability. Additional or alternative questions in these and other areas could be negotiated. Also collected are detailed employee identification information that allows linking of establishment data to work-related risks. This survey only provides estimates for the United States but not for specific geographic areas. National Medical Care Utilization and Expenditure Survey This panel survey was conducted by the NCHS in 1980. It covered about 10,000 households and provided a prob— ability sample of the noninstitutionalized population. Five interviews at 3-month intervals were conducted for each household in the sample for detailed information on health care expenditures and utilization of health services. This first survey also included a separate study of the Medi— caid-eligible populations of New York, California, Texas, and Michigan. The items in the questionnaire provide information on insurance coverage; episodes of illness; number of bed days, restricted activity days, hospital admissions, and physician visits; other medical care en- counters; and purchases of prescription medicines. Because information on specific medical conditions is obtained, this survey can provide expenditure data for those receiv- ing care for cancer. A new survey, entitled the National Medical Expendi— tures Survey, is planned for 1987 by the National Center for Health Services Research. The National Medical Care Utilization and Expenditure Survey provides estimates for the entire United States and broad geographic regions; however, it is repeated at infre- quent intervals. Although similar in scope, the new Na- tional Medical Expenditures Survey will also be repeated infrequently. National Morbidity and Mortality Follow-back Survey The NCHS plans to conduct this survey in 1986. A probability sample of about 1% of all deaths will be drawn, and questionnaires will be sent to next of kin regarding l) socioeconomic differentials in mortality, 2) the association between certain risk factors and cause of death, 3) health care services provided in the last year of life, and 4) the reliability of items reported on the death certificate. Previous surveys have been conducted by NCHS in the 19505 and 19605, and the last was in 1968. The surveys have provided data available to the general epidemiologic community and have resulted in publica- tions that include the first large-scale quantification of urban~rural and male—female differences in lung cancer mortality according to smoking category. The relative infrequency of the survey may limit the use of the information as an aid to follow-up of changes in health care services for cancer. NCI MONOGRAPHS. NUMBER 2, I986 Nationwide Food Consumption Survey The United States Department of Agriculture is a major source of national dietary and food consumption data through its Nationwide Food Consumption Survey. It has been conducted approximately every 10 years and pro- vides assessments of dietary status at national and regional levels. This survey supplies measures of household food consumption and individual food intakes from which the nutritional quality of diets can be appraised for the entire population. Data are provided on all foods by quantity, source, and monetary value, as well as related information regarding the consumers and their food acquisition, prep- aration, and consumption patterns. The 1977—78 survey included a probability sample of the 48 contiguous states; samples of Alaska, Hawaii, and Puerto Rico; and special samples of the elderly and low—income households. The Administration on Aging, Social Security Administration, and Food and Drug Administration subsidize some aspects of the current survey, including oversampling of special groups (e.g., the elderly). Noncomparable methodologies may lead to confusing data and erroneous conclusions. It is important that the Department coordinates future nutrition surveys to ensure comparable methodologies, particularly for purposes of tracking food consumption patterns over time. Other Data Sources Some of the data resources of private organizations are useful for characterizing areas according to the availabil- ity of resources for cancer treatment. American Medical Association The personnel data maintained by the Association will be helpful in providing the distribution of medical oncol- ogists in the United States, which can be related to cancer outcome data. American Hospital Association The Association publishes a Hospital Guide to the Health Care Field each year listing every hospital in the United States by state and local community that provides information on facilities available within each hospital, the kinds of approvals or accreditations acquired, special- ized services accessible, etc. This information has already been useful, particularly when related to income and other demographic factors. RECOMMENDED ACTIONS FOR SURVEILLANCE Recommendations for enhancing the capability of the NC] to obtain the data necessary for cancer control sur- veillance are presented below in two categories: 1) uses and modifications of existing systems and 2) development of new data collection mechanisms. The appropriate agen- cies are encouraged to consider the following recommen- dations: Moditlcation of Existing Systems The NCl should: Pursue analyses of differences in treatment and out— come between patients included in the SEER Pro- gram and those treated by the cancer centers. CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 Take the lead in helping population-based cancer reg- istries achieve compatibility with SEER and help them to improve data quality. Evaluate the usefulness of data in the Medicare Sys- tem, primarily in the area of treatment. Seek to augment existing national surveys (for na- tional data) rather than conduct new ones. Explore the possibility of including additional items in the SEER data, such as smoking, occupation, and detailed treatment information. Install a data collection mechanism in SEER Pro- gram areas to obtain additional information from medical records or through slide review as specific questions arise. Consider recommending that occupational data be recorded in patient records in regions served by the SEER Program. The NCHS is encouraged to: Decrease the lag time in availability of mortality data to 1 year. Modify the cancer categories in the Monthly Vital Statistics Report for current mortality estimates. Obtain data on smoking habits on an annual basis and on knowledge and beliefs about cancer on a biennial basis through the NHIS. Continue the longitudinal study of the Current Popu- lation Survey sample matched to the NDI. Add questions on cancer screening in the population to ongoing population surveys, preferably the NHIS. Estimate the direct and indirect costs of cancer regu- larly by exploring existing systems or new surveys for analyzing costs of medical care. The Centers for Disease Control should explore the possibility of using the Morbidity and Mortality Weekly Report or the 121 City Mortality System for monitoring cancers that are amenable to rapid intervention. The Bureau of the Census should include one or more questions about smoking on the 1990 census. Development of New Systems The NC] should: Conduct special surveys to obtain data on the general population in specific geographic areas. Conduct population surveys to obtain detailed data for various surveillance indicators in SEER areas or in other areas with population-based cancer registries for correlations with measures of outcome. SURVEILLANCE MEASURES Only a part of the mechanisms needed for a complete surveillance system to monitor progress toward the goal is in place. The recommended actions cover large numbers of the remaining pieces of such a system. Table V—l out— lines the data sources that would be needed for one to 49 TABLE V- l . —Surveillance indicators Indicator Measure Source Mortality Deaths per 100,000 persons by age, sex, race, SES, NCHS mortality data geographic region; and total deaths Incidence Cases per l00,000 persons by age, sex, race, SES, SEER Program, other population-based registries geographic region; and estimated total cases Survival Relative survival by cancer site SEER Program and other registries Smoking Percentage of adults and children who smoke, NHIS, Current Population Survey percentage who stopped, and time since quitting Diet Percent obesity, and percent fat, fiber, NHANES, US. Department of Agriculture fruit, and vegetable consumption National Food Consumption Survey Occupation Percentage of workers exposed; percentage of NHIS, Current Population Survey, other workers screened in the workplace population surveys Screening Percentage of eligible persons screened NHIS, Current Population Survey, other population surveys Treatment Distribution by cancer stage of diagnosis, SEER Program; population-, hospital-, or cancer Knowledge, attitudes, and beliefs percentage of cancer patients treated by multi— disciplinary approaches, cancer patient survival Percentage of total population and ethnic groups with particular knowledge, beliefs, etc., about cancer center-based registry; Medicare Data System; Directory Medical Specialists Supplement to NHIS, independent population surveys Costs Indirect and direct Medicare Data System; National Medical Expenditures Survey; surveys of medical insurance obtain surveillance measures associated with each of the objectives. These will be refined further as the surveillance system is completed. The data mechanisms listed, either existing or proposed, will be sources of measures specifically related to the objectives in many instances but not in others. Researchers must exercise caution in the use of measures of dietary intake from population surveys to monitor changes in fat and fiber consumption because the technology is not suf- ficiently precise to yield reliable measures. Therefore, in these and certain other instances, the statement of objec— tives and the capability of the NC] to monitor progress toward them will have to be made compatible. EVALUATION STRATEGY For many years, the NCI has been monitoring the age- specific and age-adjusted cancer mortality rates for all sites combined and for individual sites. This will continue, of course, because it is basic to the assessment of progress in the cancer control program. Special attention will be given to those cancer sites for which specific interventions have been recommended. In addition, monitoring of age- specific and age-adjusted incidence rates will continue, both to help explain changes in the mortality rates and to assess the impact of some of the prevention efforts. Moni- toring of survival rates will also continue on an ongoing, systematic basis. Now that data on the extent of disease at diagnosis are being collected by the SEER Program in a way that per- 50 mits comparisons over time since 1977, these will be sys- tematically monitored. Monitoring will help us assess the effect of screening and of access to treatment and will explain part of the changes in survival rates. This will necessitate the monitoring of changes in survival within each stage and the calculation of survival rates adjusted for changes in the stage distribution. Except for the mortality rates, all of this can be done with data generated by the SEER Program, and monitor- ing of this sort is aimed at assessment of cancer outcome. Also important is that the impact of the intervention pro— cess be assessed. Thus if intervention efforts are aimed at a reduction in the proportion of heavy smokers in the population, it is of immediate interest that we determine whether their proportion is actually decreasing and by what amount. Therefore, from a refinement of the outline above, a set of specific indicators will be constructed, baseline values will be obtained, and progress will then be monitored over time. Other indicators will be constructed on an ad hoc basis that will help assess the impact of spe— cific intervention efforts, first by obtaining baseline mea- sures and later by obtaining the same measures at subse- quent periods. Some attention will be given to identification of the smallest number of key indicators that will succinctly reveal progress of the overall mortality reduction effort. It is likely that this can be done in such a way that major areas could be identified where interventions do not seem to be having an effect, so that corrections in the cancer control process can be made. VI. References REFERENCES (1) PUBLIC HEALTH SERVICE: Public Health Service implemen- tation plans for attaining the objectives for the Nation. Public Health Rep 98(suppl):l—178, 1983. (2) DOLL R, PETO R: The causes of cancer: Quantitative esti— mates of avoidable risks of cancer in the United States today. JNCI 66:1191—1308, 1981. (3) COMMITTEE ON DIET, NUTRITION, AND CANCER, NATIONAL RESEARCH COUNCIL: Diet, Nutrition, and Cancer. Wash- ington, DC: Natl Acad Press, 1983. w: ATIONAL CANCER INSTITUTE: Cancer Prevention: Good News; Better News; Best News. DHHS (NIH) Pub] No. 84-2671. Bethesda, MD: Natl Cancer Inst, 1984. (5) NATIONAL INSTITUTES OF HEALTH: Panel Report: Consen- sus Development Conference on Lowering Blood Choles- terol to Prevent Heart Disease. Bethesda, MD: Natl Inst Health, December, 1984. (6) SHAPIRO S, VENET W, STRAx P, et a1: Ten- to fourteen- year effect of screening on breast cancer mortality. JNCI 69:349—355, 1982. (7) TABAR L, FAGERBERG CJ, GAD A, et a1: Reduction in mor— tality from breast cancer after mass screening with mam- mography. Lancet 1:829—832, 1985. (8) STENKVIST B, BERGSTROM R, EKLUND G, et a1: Papanico- laou smear screening and cervical cancer: What you can expect. JAMA 252:1423-1426, 1984. k @SURVEILLANCE AND OPERATIONS RESEARCH BRANCH, DI- VlSION OF CANCER PREVENTION AND CONTROL, NATIONAL CANCER INSTITUTE: Annual Cancer Statistics Review. Presented to the National Cancer Advisory Board, Be- thesda, MD, November 26, 1984. (10) NATIONAL INSTITUTES OF HEALTH CONSENSUS DEVELOP- MENT PANEL ON ADJUVANT CHEMOTHERAPY AND ENDO- CRINE THERAPY FOR BREAST CANCER: Introduction and Conclusions. NCI Monogr 1:1-4, 1986. ¥< (11) POLLACK ES: Prognosis among cancer patients in relation to socioeconomic level of census tract of residence. Pre- sented at the Data Use Conference on Small Area Statis- tics, Snowbird, UT, August 29—31, 1984. ([2) LYON JL, GARDNER JW, WEST DW: Cancer in Mormons and non—Mormons in Utah during 1967—75. JNCI 65: 10554061, 1980. (13) ENSTROM JE, KAMIN LE: Populations at low risk. In Cancer Prevention in Clinical Medicine (Newell G, ed). New York: Raven Press, 1983. 4 MICHAELS D: Occupation and cancer in the black popula- tion: The health effects ofjob discrimination. J Natl Med . Assoc 75:1014—1018, 1983. (15) LIPWORTH L, ABELIN T, CONNELLY RR: Socio-economic factors in the prognosis of cancer patients. J Chronic Dis 232105—116, 1970. (16) DIVISION OF CANCER PREVENTION AND CONTROL, NA- TIONAL CANCER INSTITUTE: Cancer Among Blacks and Other Minorities: Statistical Profiles. DHHS (NIH) Publ No. 86-2785. Bethesda, MD: Natl Cancer Inst, 1986. (I7) LUBIN JH, BLOT WJ, BERRINO F, et a1: Modifying risk of developing lung cancer by changing habits of cigarette smoking. Br Med J 288:1953—1956, 1984. (18) OFFICE OF SMOKING AND HEALTH, PUBLIC HEALTH SER- VICE: The Health Consequences of Smoking: Cancer. A Report of the Surgeon General. DHHS (PHS) 82-50179. Washington, DC: US. Govt Print Office, 1982. (19) WINN DM, BLOT WJ, SHY CM, et a1: Snuff dipping and oral cancer among women in the southern United States. N Engl J Med 304:745—749, 1981. (20) WELCH S, MARSTON RM: Review of trends in food use in the United States: 1909-1980. J Am Diet Assoc 81:120— 125, 1982. (21) SCHOTTENFELD D, FRAUMENI JF JR (eds): Cancer Epide— miology and Prevention. Philadelphia: Saunders, 1982. (22) CANADIAN TASK FORCE ON THE PERIODIC HEALTH EXAM- INATION: The periodic health examination. Can Med Assoc J 12121—46, 1979. Q3) AMERICAN CANCER SOCIETY: Guidelines for the Cancer- Related Checkup: Recommendations and Rationale, Publ No. 3347—PE. New York: Am Cancer Soc, 1980. (24) PROROK PC, CHAMBERLAIN J, DAY NE, et a1: UICC work- shop on the evaluation of screening programmes for cancer. Int J Cancer 3421—4, 1984. (25) BAKER LH: Breast Cancer Detection Demonstration Proj- ect: Five-year summary report. CA 32:194—225, 1982. (26) BAKER LH, CHIN DY, WAGNER KV: Progress in screening for early breast cancer. J Surg Oncol 30:96—102, 1985. (27) VERBEEK AL, HENDRIKS J H, HOLLAND R, et a1: Reduction of breast cancer mortality through mass screening with modern mammography. First results of the Nijmegen project, 1975—1981. Lancet 1:1222—1224, 1984. (28) COLLETTE HJ, DAY NE, ROMBACH JJ, et a1: Evaluation of screening for breast cancer in a non-randomised study (the DOM project) by means of a case—control study. Lancet 1:1224—1226, 1984. (29) HAKAMA M, RASANEN-VIRTANEN U: Effect of a mass screening program on the risk of cervical cancer. Am J Epidemiol 103:512—517, 1976. (30) FIDLER HK, BOYES DA, WORTH AJ: Cervical cancer detec- tion in British Columbia. J Obstet Gynecol Br Common- wealth 75:392—404, 1968. (31) MILLER AB, LINDSAY J, HILL GB: Mortality from cancer of the uterus in Canada and its relationship to screening from cancer of the cervix. Int J Cancer 17:602—612, 1976. (32) MACGREGOR J E: Evaluation of mass screening for cervical cancer in Northwest Scotland. Tumori 62:287—295, 1976. (33) MACGREGOR JE, TEPER S: Mortality from carcinoma of cervix uteri in Britain. Lancet 2:774—776, 1978. (34) JOHANNESSON G, GEIRSSON G, DAY NE: The effect of mass screening in Iceland, 1965—74, on the incidence and mor- tality of cervical carcinoma. Int J Cancer 21:418—425, 1978. (35) THORARISSON A, JENNSON O, BJARNASON 0: Screening for uterine cancer in Iceland. Acta Cytol 13:304—308, 1969. (36) TIMONEN S, NIEMINEN U, KAURANIEMI T: Mass screening for cervical carcinoma in Finland: Organization and effect on morbidity and mortality. Ann Chir Gynaecol Fenn 63:104—112, 1974. (37) WALTON RJ: Cervical cancer screening programs: Errata in full 1982 report (letter). Can Med Assoc J 127:953, 1982. (38) CHRISTOPHERSON WM, LUNDIN FE JR, MENDEZ WM, et a1: Cervical cancer control: A study of morbidity and trends over a twenty-one year period. Cancer 38:1357v - 1366, 1976. (39 EDDY DM: Guidelines for the cancer-related checkup. Rec- ommendations and rationale. CA 30:194—240, 1980. (40) PETERSON O: Spontaneous course of cervical precancerous conditions. Am J Obstet Gynecol 72:1063, 1956. (41) Ross LG, STEWARD FW, FOOTE FW, et a1: Some histolog- 53 ical aspects of behavior of epidermoid carcinoma in situ and related lesions of the uterine cervix: A long-term prospective study. Cancer 16:1160—1211, 1963. (42) INTERNATIONAL ACADEMY OF CYTOLOGY: The International Academy of Cytology’s statement on the frequency of gynecologic screening. Acta Cytol 24:371—372, 1980. (43) WIED GL: Frequency of cervical cancer screening. JAMA 2412294, 1979. (44) INTERNATIONAL UNION AGAINST CANCER: Screening in cancer: A report of a UICC International Workshop, Toronto, Canada, April 24—27, 1978. (45) PREVENTIVE MEDICINE TASK FORCE: Reports on Preven- tive Medicine USA. New York: Prodist, 1976. 4 BRESLOW L, SOMERS AR: The lifetime health-monitoring program: A practical approach to preventive medicine. N Engl J Med 296:601—608, 1977. 47 OFFICE OF THE SURGEON GENERAL, PUBLIC HEALTH SER- VICE: Healthy People: A Report of the Surgeon General on Health Promotion and Disease Prevention. DHEW (PHS) Publ No. 79-55071. Washington, DC: US. Govt Print Office, 1979. (48) BOUCOT KR, WEISS W: Is curable lung cancer detected by semiannual screening? JAMA 224:1361—1365, 1973. (49) WEISS W, BOUCOT KR, SEIDMAN H: The Philadelphia Pulmonary Neoplasm Research Project. Clin Chest Med 3:243—256, 1982. (50) LILIENFELD A, ARCHER PG, BURNETT CH, et al: An eval- uation of radiologic and cytologic screening for the detec- tion of lung cancer: A cooperative pilot study of the American Cancer Society and the Veterans Administra- tion. Cancer Res 26:2083—2121, 1966. (51) BRETT GZ: Earlier diagnosis and survival in lung cancer: Br Med J 4:260—262, 1969. (52) NASH RA, MORGAN JM, TOMKINS JG: South London cancer study. Br Med J 2:715~721, 1968. (53) FONTANA RS, TAYLOR WF: Screening for lung cancer: The Mayo Lung Project. In Lung Cancer: Causes and Preven- tion (Mizel M, Correa P, eds). Deerfield Beach, FL: Ver- lag Chemie Int, 1984, pp 161—174. (54) MELAMED MR, FLEHINGER BJ, ZAMAN MB, et al: Screen- ing for early lung cancer: Results of the Memorial Sloan- Kettering study in New York. Chest 86:44—53, 1984. (55) MELAMED MR, FLEHINGER BJ: Screening for lung cancer. Chest 8622—3, 1984. (56) NATIONAL INSTITUTES OF HEALTH: Cervical cancer screen- ing: The Pap smear. Consensus Development Conference Summary, vol 3, No. 4. Bethesda, MD: Natl Inst Health, 1980. (57) FOOD AND DRUG ADMINISTRATION: Chest x-ray screening statements. FDA Drug Bull 13:13—14, 1983. (58) FOSTER RS JR, COSTANZA MC: Breast self-examination practices and breast cancer survival. Cancer 532999—1005, 1984. (59) COSTANZA MC, FOSTER RS JR: Relationship between breast self-examination and death from breast cancer by age groups. Cancer Detect Prev 7:1037108, 1984. (60) THOMPSON lM, ERNST JJ, GANGAI MP, et al, Adenocarci- noma of the prostate: Results of routine urological screen- ing. J. Urol 132:690—692, 1984. (61) CHODAK GW, SCHOENBERG HW: Early detection of pros— tate cancer by routine screening. JAMA 252:3261—3264, 1984. (62) SMITH T: The Queensland Melanoma Project: An exercise in health education. Br Med J 1:253—254, 1979. DALES LG, FRIEDMAN GD, COLLEN MF: Evaluating peri- odic multiphasic checkups: A controlled trial. J Chronic Dis 321385404, 1979. (64) GILBERTSEN VA: Proctosigmoidoscopy and polypectomy in 54 reducing the incidence of rectal cancer. Cancer 34:936— 939, 1974. (65) HERTZ RE, DEDDISH MR, DAY E: Value of periodic exami— nations in detecting cancer of the rectum and colon. Post- grad Med 27:290—294, 1960. (66) WINAWER SJ, MILLER DG: Screening approaches based on risk. In Colorectal Cancer: Prevention, Epidemiology, and Screening (Winawer SJ, Schottenfeld D, Sherlock P, eds). New York: Raven Press, 1980, pp 159—165. (67) WINAWER SJ : Detection and diagnosis of colorectal cancer. Cancer 51:2519—2524, 1983. (68) GILBERTSEN VA: Prevention of hereditary large bowel cancer: Status of screening for large bowel cancers. Prog Clin Biol Res 115:79—80, 1983. (69) EDDY DM: Cost-effectiveness of colorectal cancer screen— ing. In Frontiers in Gastrointestinal Cancer (Levin B, Riddell RH, eds). New York: Elsevier, 1984, pp 203—219. (70) COORDINATING GROUPS FOR RESEARCH ON ESOPHAGEAL CANCER: Early diagnosis and surgical treatment of esopha— geal cancer under rural conditions. Chin Med J 2:113—1 16, 1976. (71) NABEYA K: Markers of cancer risk in the esophagus and surveillance of high-risk groups. In Precancerous Lesions of the Gastrointestinal Track (Sherlock P, Morson BC, Barbara L, et al, eds). New York: Raven Press, 1983, pp 71—86. (72) OKUI K, TEJIMA H: Evaluation of gastric mass survey. Acta Chir Scand 146:185—187, 1980. (73) OSHIMA A, HANAI A, FUJIMOTO 1: Evaluation of a mass screening program for stomach cancer. Natl Cancer Inst Monogr 53:181-186, 1979. (74) GREENWALD P, COSTLOW R, PROROK P, et al: Screening in the context of cancer control. In Cancer Screening (Miller AB, ed). New York: Academic Press, 1984, pp 25—48. (75) NATIONAL COUNCIL ON RADIATION PROTECTION AND MEA- SUREMENT: Mammography Report No. 66. Bethesda, MD: Natl Council Radiat Protection, 1980. (76) SHULMAN JJ, HONTZ A, SEDLIS A, et al: The Pap smear: Take two. Am J Obstet Gynecol 121:1024—1028, 1975. (77) LUTHY DA, BRIGGS RM, BUYCO A, et al: Cervical cytol- ogy. Increased sensitivity with a second cervical smear. Obstet Gynecol 51:713—717, 1978. (78) AXTELL LM, ASIRE AJ, MYERS MH (eds): Cancer Patient Survival, Report No. 5. DHEW Publ No. (NIH) 77—992. Washington, DC: US. Govt Print Off, 1976. (79) EDITORIAL: ’Review of mortality results in randomized trials in early breast cancer. Lancet 221205, 1984. (80) MOERTEL CG, FRYTAK A, HAHN RG, et al: Therapy of locally unresectable pancreatic carcinoma: A randomized comparison of high dose (6000 rads) radiation alone, moderate dose radiation (4000 rads + 5—fluorouracil), and high dose radiation + S-fluorouracil. Cancer 48: 1705—1710, 1981. (81) IHDE DC, BUNN PA, LICHTER AS, et al: Randomized trial of chemotherapy with or without adjuvant chest irradia- tion in limited stage small-cell lung cancer. In Adjuvant Therapy of Cancer IV (Jones SE, Salmon SE, eds). Orlando, FL: Grune & Stratton, 1984. (82) DEVITA VT JR, LIPPMAN M, HUBBARD S, et al: The effect of combined modality therapy on local control and sur- vival. Presented at the Third Rome International Sympo- sium of the Associazione Italiane per la Promozione della Studia delle Malattie Oncologiche, Rome, May 17, 1985. (83) WERNz JC, SPEYER JL, NOUMOFF J, et al: Cisplatin (DDP/Cytoxan): A high-dose DDP regimen for ad- vanced stage ovarian carcinoma. Proc ASCO 1:112, 1982. (84) YOUNG RC, MEYERS CE, OSTCHEGA Y, et al: CPR: Cyclo- phosphamide, a high—dose cisplatin and radiation for advanced ovarian carcinoma. Proc ASCO 4:119, 1985. NCI MONOGRAPHS, NUMBER 2. I986 (85) HANKS GE, KRAMER S: The patterns of care study: A spe- cialty effort to improve care. Appl Radiol 12:69—75, 1983. (86) DUFFNER PK, COHEN ME, FLANNERY JT: Referral pat- terns of childhood brain tumors in the State of Connecti- cut. Cancer 5011636—1640, 1982. (87) LEIBEL SA, HANKS GE, KRAMER S: Patterns of care out- come studies: Results of the national practice in adeno- carcinoma of the prostate. Int J Radiat Oncol Biol Phys ,_,_\ 102401409, 1984. (8:?) AMERICAN CANCER SOCIETY: A Study of Black America’s CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 Attitude Toward Cancer and Cancer Tests. New York: Am Cancer Soc, 1980. (89) GRIFFEL M: Wilms’ tumor in New York State: Epidemiol- ogy and survivorship. Cancer 40:3140—3145, 1977. (90) KlNZlE JJ, HANKS GE, MACLEAN CJ, et a1: Patterns of care study: Hodgkin’s disease relapse rates and adequacy of portals. Cancer 52:39—42, 1983. (91) DEVITA VT JR, HELLMAN S, ROSENBERG SA (eds): Cancer: Principles and Practices of Oncology, 2nd ed. Philadel- phia: Lippincott, 1985. 55 Appendixes A: Use of Models to Project Cancer Mortality in the Year 2000 B: Analysis of the Impact of the Cancer Control Objectives Appendix A: Use of Models to Project Cancer Mortality in the Year 2000 For the purpose of estimating the impact of the cancer control objectives specified by the Working Groups, a population model of cancer incidence, survival, and mor- tality was created to fit the basic outlines of the Working Groups’ recommendations. The conceptual framework of this model, developed by Dr. David Eddy of Duke Uni- versity, and the mathematical details of a computer pro- gram written to implement the model are given elsewhere in this monograph (1,2). This Appendix describes the demographic input, data on the cancer sites, and the crea— tion of modeling scenarios blending the recommendations to the model’s requirements. Appendix B presents the results of our running the computer program with the var- ious scenarios given here. The modeling efforts described in this Appendix are only the first in a series of estimates of the effect of the programs envisioned by the Working Groups. Additional analyses of the impact of many of their recommendations, as well as analyses regarding the assumptions necessary for the modeling efforts, will continue as cancer preven- tion and control efforts progress. Figure A-l shows a general overview of the model. The starting population is followed, and records are kept of the three critical events (cancer incidence, detection, and treatment) and specific outcomes. The recommendations of the Working Groups affect the critical events and the like— lihood of death through prevention, screening, and treat- ment interventions. Each intervention affects one or more parameters in the overall model and eventually affects cancer incidence and mortality. DEMOGRAPHIC DATA An age—sex distribution of the United States population is part of the basic input for the model. The age distribu- tion in 5-year groups is shown in table A-l. The age~sex breakdown is modeled in single years starting with live births and ending with a 99-and-over age group. Single- year data are necessitated by the time—dependent Weibull survival mode]. The source for the age—sex distribution of the population on July 1, 1980, was the United States Census Bureau and its recent publications from the 1980 decennial census (3). The model ages this population, sim- ulating the progression of time, and accrues cases of cancer and cancer deaths; therefore, the resultant popula- tion in the year 2000 is not necessarily the same as an estimate would be from the Census Bureau. This poses no problem in conclusions being drawn because the model is used in comparisons of scenarios for different cancer planning efforts and always starts with the same popula- tion distribution. For each year beginning with 1980, estimates of the number of births from the Census middle series and the gender ratio estimated by NCHS (4.5) were used (table A-2). These were provided at 5-year intervals, i.e., 1980, 1985. To obtain accurate estimates of the number of births each year, we used linear interpolation between the 5-year estimates. Mortality rates from all causes and deaths from cardio- vascular disease (table A-3) were collected from the Social Security Administration’s Trustee Report (6). These data were given by age and sex in 5-year age groups with the exception of death rates for live births and for the l- to 4-year-old children separately. Here and in virtually all other instances when only 5-year, age-grouped data were available, the rate for a specified S-year age group was generalized to each of the 5 years. No smoothing or other adjustment was obtained. To compute the mortality rate for diseases other than cardiovascular disease, as given by the Trustees’ Report, we subtracted the mortality rate for heart and vascular diseases from the total mortality rate for all causes. Overall, the total percentage of deaths from heart disease in the United States in 1982 was 49%. Since the 1950s, this rate has been steadily dropping at a rate of about 1.4% per year (7). Although the model allowed for use of continuing trends in heart disease mor- tality as an explicit input, this was ignored for preliminary estimates contained in the report. A consistent decline in heart disease mortality over the remaining years in this century will be used in future runs of the model, so that the sensitivity of cancer mortality to one of the major competing risks, heart disease, can be examined. BASELINE DATA FOR ALL CANCERS Age-specific mortality rates for the entire population for each sex and cancer site (tables A-4, 5) were necessary for the model. This need was similar to that for death rates from cardiovascular disease by age and sex. The site- specific mortality rates were taken from the 1978 to 1981 incidence and mortality data of the SEER Program (8) as derived from computer tapes supplied by the NCHS. The Working Groups made several assumptions in using the site-specific mortality rates because some of the cancer sites they used involved histologic breakdowns not avail- able from the NCHS data. For example, the Treatment 59 Population Outcomes Incidence thortion Treatment Trca1mcnt Model Prevention Mudt‘l Screening Mndul FIGURE A-1.—Cancer control model: Schematic View. Working Group focused only on small cell lung cancer as a neoplasm for which treatment improvements can be expected in the near future. However, the site-specific mortality rates published by the SEER Program are for all lung cancers; in this case, a separate analysis of the mor- tality data was necessary so that only the small cell tumors were counted. For testicular cancer, approximately 50% of the total rate was used for the nonseminoma cases only. For childhood brain tumors, the Working Group focused on only three subsets of this cancer type. Because these were the most prevalent childhood brain tumors, the site- specific mortality rate for all childhood brain tumors was used in the calculations. (The use of this rate is not likely to have a large impact on performance evaluation of new treatment methods with this modeling effort.) Data were available on all persons 85 years of age and over, and these were generalized for each of the last three age categories (85—89, 90—94, and 95+). The childhood disorders are defined as those occurring before the age of 15 years. Another key data element was incidence rates for each cancer site by age and sex (tables A—6, 7). Again, the TABLE A—1.7The 1980 Census populalion, by age and sex” TABLE A-2.—Projected annual number of births, by year and sex“ Age, yr Male Female <5 8,362,009 7,986,245 59 8,539,080 8,160,876 10714 9,316,221 8,925,908 15-19 10,755,409 10,412,715 2077 24 10,663,231 10,655,473 2529 9,705,107 9,815,812 3034 8,676,796 8,884,124 35,739 6,861,509 7,103,793 407 44 5,708,210 5,961,198 4549 5,388,249 5,701,506 507 54 5,620,670 6,089,362 5559 5,481,863 6,133,391 6064 4,669,892 5,417,729 65-69 3,902,955 4,879,526 70 74 2,853,547 3,944,577 757 79 1,847,661 2,946,061 8084 1,019,227 1,915,806 8589 477,185 1,043,017 90 94 159,077 397,515 95+ 45,263 118,010 Total 110,053,161 116,492,644 Year Male Female 1980 1,865,506 1,773,294 1981 1,884,721 1,791,559 1982 1,903,935 1,809,825 1983 1,923,150 1,828,090 1984 1,942,365 1,846,355 1985 1,961,580 1,864,620 1986 1,963,928 1,866,852 1987 1,966,276 1,869,084 1988 1,968,624 1,871,316 1989 1,970,972 1,873,548 1990 1,973,320 1,875,780 1991 1,950,640 1,854,220 1992 1,927,959 1,832,661 1993 1,905,279 1,811,101 1994 1,882,598 1,789,542 1995 1,859,918 1,767,982 1996 1,846,270 1,755,010 1997 1,832,623 1,742,037 1998 1,818,976 1,729,064 1999 1,805,328 1,716,092 2000 1,791,681 1,703,119 ” Sources of data are the NCHS (1981) and (5). SEER Program data were used to provide age-specific rates by sex and site for first primary cancers. Incidence data for first primaries only were used to correspond with the available survival statistics. Except for breast and cer- vical cancers, rates for invasive carcinoma only were used. In breast and cervical cancers, rates for in situ cancer were used because the intervention recommended by the Screen- ing Working Group included a projected stage shift for persons to be screened by the program that would increase TABLE A-3.7Morrality rates per 100,000 persunsfrom all causes of death” Noncardiovascular Cardiovascular Age, yr Male Female Male Female <1 1,387.7 1,110.5 33.0 25.8 174 68.8 51.3 3.4 3.1 59 34.0 24.4 1.2 1.3 10714 36.9 21.5 1.5 1.4 15* 19 138.1 50.7 3.9 2.7 20724 195.9 57.6 6.6 4.1 25729 184.4 61.1 10.3 6.7 30734 173.2 72.6 20.5 11.0 35739 192.3 102.0 50.7 21.4 40—44 246.1 155.2 116.0 44.6 45 49 357.1 238.5 229.5 84.3 50754 532.9 356.6 411.7 143.4 55759 782.8 497.0 678.4 251.6 6064 1,149.3 689.5 1,072.5 449.2 6569 1,731.1 931.3 1,734.3 801.7 70774 2,425.8 1,241.9 2,705.6 1,408.6 75 79 3,366.2 1,678.3 4,224.1 2,563.7 80 84 4,559.1 2,401.8 6,614.2 4,696.6 8589 6,225.5 3,531.9 10,310.] 8,221.4 9094 8,149.8 4,969.2 15,593.9 13,714.5 95+ 9,527.3 6,529.7 20,087.8 19,387.5 " Source of data is the Bureau of the Census, 1983. 60 a Source of data is (6). NC] MONOGRAPHS, NUMBER 2, 1986 .v.< 0.02 .0 20500. 000 0 .v.< v.02 .0 80500.. 00.0. a 0~0.~.0 00.00 3.0. 00.00 :00. 000.0 ~06. 0~.0 00.0"~ 00.0w .0.00~ 0v.00 ~000 00.0. .0000 000.0 «.0... 3.0. +00 v0~.v~v 00.~0 2.0. 00.0.. 00.00. 02.0 0.0. 00.0 .~.0~ ~000 00.00. V000 00.00. 00.0. 000.00 «00.0 3.0 00.0 0000 .~0.0~0 3.0~ 00.0. 0%.». 00.0.. 0.0.0 V0.0. 0.0 00.2 ~...~ 0~.00. 00.00 00.0 00.0 03.00 20.2 00.0 00.0 00:00 .~0.0v~ 00.v~ 00.2 00.00 00.00. 00~.0 00.0 v~.0 000. 00.3 .000 ~0.~v 00.- 00.0 0.0.00 000.2 00.0 3.0 00.00 00002 00.0~ ~0... -.00 00.00 .000 00.0 v~.v ...0 0~.0. 00.00 V0.0 .00. .0.0 «.0000 000.0. ~00 v0.v 00-00 ~00.00. 00.0. 00.2 00.0~ 0~.00 000.~ 0~.v 00.0 0.... v.0 00.2w 2 ..~ 000 00.0 000.00 000.0. 000 00.0 v0-00 .0000 0.0 00.0 00..~ ~0.00 0~..~ 00.~ 00.0 00.~ ~04q 0~.0~ 002 00.0 v0... 000.0». 0.02 00..“ 00.~ 00-00 00~.00 000 00.0 .~.0. 00200 Ev. ~v.~ 00.~ v0. 00.~ 00.0. 00.0 00.0 00.~ 000.00 0.0.0 00.~ .0. 0000 ~00.~0 0..~ ...0 0~.0 .000 0000 00.. 00.~ .00 ~v.. .00 00.0 0~.~ 0~.. 000.0. -0.v .0.. 00.0 0070.. ~30. .0. 0.0 V00 00.0~ 02.0 0.... 00.. v~0 00.0 .~.v 00.. 00.. .00 000.0 000.. 000 $0 $10». 00.0. 0V0 00.0 00.. 002 00~0 v... 0~.. 20 ~00 .0. 00.0 00.0 20 000.0 0.0.0 2 .0 2.0 00-00 00~.0 0~.0 00.. 00.0 2.0 02 0 000 00.0 00.0 0. .0 00.0 -.0 000 00.0 0000 0000 00.0 00.0 00-00 000... 00.0 ~00 00.0 .0. 00.0 0.0 0 ~00 .00 00.0 00.0 00.0 0.0 .00 00.0 20.0 00.0 3.0 0~-0~ ~«0.~ 000 0~0 000 0.0 000.0 .00 0V0 ~00 .00 0~0 .00 30 0.0 3.0 00.0 00.0 000.0 000.0 00.0 ~00 v~u0~ 00~.~ .00 00.0 -0 ~00 000.0 00.0 3.0 00.0 00.0 000 00.0 .00 00.0 .00 ~00 000 0.0.0.0 000.0 00.0 00.0 0.0. 00.0.. .00 .00 00.0 .00 .~00 0~.0 00.0 00.0 .00 00.0 00.0 .00 .00 000 00.0 000.0 000.0 ~00 .00 v.10. 000.. .00 00.0 ~00 00.0 :00 0~.0 00.. 00.0 00.0 00.0 00.0 .00 .00 00.0 00.0 000.0 000.0 .00 ~00 0‘0 00.,~ .00 00.0 .00 ~00 0~00 A. ~00 00.0 00.0 .00 00.0 00.0 .00 00.0 00.0 000 000.0 000.0 .00 .00 0V 8..” :2: 0.5.00 3:0 .030». .....n. 22¢. ..S.< ..1.< Ea... 2:95.02 .5000... 5:83. co.oU 035:5. 0.22020 30.2.0on :3 :8 080.5. 058 .0. .550 0:900 + ..~Em.:oZ =mEm + .825 dw< ..< :...< 00000.50 0.5.3.7.. ..:0< 000.. =05. 3 .8.0£m\.8\ 000.00. km 0.5 5.02er .6230 wSuan-0mVl.0-< 0.5.... 5.00 .126 0.5 0. mean 5. 3:2 5.3 030.05 33 ....E<+....< ..:0<: 0.0.5 5. 0~ 50:: news .5. 3:2 .o 8: 5.3 03002.3 203 3:0 ..0oo..0.EQ: 0 .mmuz 2: w. 28 0. 85am .. 0.~.000 000.0 ...~00 30.0 ~0.0~ .~.0_ 00.2... ~0.00 2.000 ~0.~2 00.0.. 00.00 .00000 0~0.v~ 0.00 00.0~ +00 00. .~00 000.0 3.2.~ 0.0.. . 00.v~ 00.2 ~0..0. 00.00 .0.0v~ 00.00 2.0 0~.00 .0003. 000.00 «.020 .00~ V0-00 «0~.000 0000 00.00~ 0.0.0 00.2 000. .~.00 00.00 00.2 ~0.00 00.00 0000 000.2. 000.00 v..0~ 00.0. 00-00 0.0.000 00~0 00.02 ~0~.0 2.! 00.0 00.0.. 00.0~ 0..~0. 00.00 00.0w 00.0~ 000.000 v2.4.0 0~.0~ 00.0. .000 0~0.~3 00.0 .000 000.0 00.0 0.0 00.0~ 00.0. 00.00 00.00 N000 00.0~ .2000 000.00 0~.0~ ~02 00-00 000.0v~ 00. .0 00.00 ....v 0Y0 00.0 00.0. 00.2 .060 00.00 0v.0~ 00..~ 0. ..0.~ 30.0w 00.0~ 0. .0. 00.00 30.00. 00.0 .02 000.0 0~... ~00 00.0 0.0 v0.~0 00.0~ 2.2 0~.0. 000.00. 000.0~ 00.1 00.0 00:00 00.00. 00~0 0~.v 0v..~ 00.~ 00.0 .00 0..»V .00. 00.2 00.0 00.0 000.00 .000. 00.0 0~.0 00.00 30.00 0.~.0 0... 0~.. 00.~ 00.0 00.. ~..~ 00.0 3.0 00..V 00.0V 00.2. «00.0 00.0q 00.~ 070.. 00.00 00~0 2.0 0000 00.. ...0 9.0 000 00.0 00.~ ~0.~ ~0.. ~0v.0. 0~0.~ .0. 0~.. $70.. 30.0. 0000 30 000.0 ~0.. 00.~ v.0 0V0 v~.~ 00.. 00.0 000 0-.0 0000 V0.0 000 0000 N000 00v0 00.0 20.0 0... ~0.. 00.0 0.0 v0. 000 000 20 00. .. 020 0. .0 . .0 00-00 00~.0 00v.0 .00 00~0 00.. 0 00.0 .00 2.0 00.0 2.0 0.0 00.0 00~0 .000 00.0 00.0 0~‘0~ 000... 030 .00 00~0 0~.. 00.0 000 000 00.0 .00 v0.0 -.0 00.0 00.0 ~00 .000 0.0.0 000 00.0 v~I0~ 0~0.0 00.0 00.0 ~0.0 ~0.. 00.0 00.0 000 00.0 00.0 .00 00.0 .00 .00 00.0 030 000.0 00.0 00.0 0.0. 0. ..~ 0.0.0 .00 000.0 000 0... 00.0 .00 .00 .00 ~00 00.0 .00 00.0 08.0 000.0 ~00 00.0 0.0. ov~.~ 000.0 .00 000.0 ~00 00.. V00 .00 00.0 00.0 .00 00.0 .00 .00 0.0.0 000.0 .00 .00 0-0 000.~ 0.0.0 00.0 ~30 0 000 00.0 00.0 .00 .00 00.0 .00 .00 00.0 00.0 000.0 000.0 ~00 00.0 0V v.2.“ chEEomao: 328;. 1....n. .32 1.3;. ....< 0020. «Ea—3.3: .5000... EEqu 00.00 320:9. 0808.0 0:02.080. :8 :8 5.0.3. .338 b. .550 .003... + 02.5002 .EEw + .803. dw< ..< 1.1.... 002.0000 0:005... ..:0< 0.5.. “2.: 00 60.05 5.0 000.00. ha 0.5 3:029: #0230 0%003.0MV|.Y< 2.2... 6] CANCER CONTROL OBJECTIVES FOR THE NATION: I985—2000 .6.< 0.62 .6 3058.. 8.6 a 6.4... 0.68 6 30502 8.6 a 66.33 3.63 66.63 3.63 3563 33 66.2 3.2 36.66 66.3 3.33 .565 2.36 666 3.3 v3... 666 63.63 +2. 3.62. 36.66 63.53 2.; 66.33 v.2 :2 3.2 3.63 3.2. 35.33 36.3 3.3 3.2 66.3 56.6 66.6 66.2 3.66 6663 66.3 66.3 6...»... 3.63 3.6. 33.6 3... m. .2. 2.65 3.23 36.5.. 2.3 3.6 55.3 3.2 2.6. 3.2 35-35 32m 3.3 3.3 36.3. 66.63 3.6 66.6 63.2 36.3 3.3 53.32 63.3 3.3 66.5 63.3 3.3 63.6 36.5. V565 3.33 63.3. 6. .3 6. ..v 3.33 3.6 56.3 36.2 66.3 66.3 3.2. 3.3 3.2 3.5 66. 3 3.3 3.2 5. .2 3636 33.36. 63.36. 3.3 35.3 3.33 2.6 ~66 63.2 3.2 3.3 3.55 35.2 2.2 69.3 3.2. 3.2 36.2 Av6.2 $166 3.3. 2.65 v. .3 3.3 3.3. 3.3 33.3 66.2 2.6. 66.3 65.2. 3.2 3.3 3.». 3.3 3.2 33... 3... 3'3 366. 66.3 2.3 2.3 36.55. 2.3 3.3 63.2 66.6 2.2 35.3 3.x 55.3 36.3 3.3 3.6. 65.5 3.5 363 m5.56 3.3 .62. 36.5. 3.32 63.. 66.3 2.2 3.v 3.5 2.6. 3.3 33 3.. 3.3 2.3 33 336 373 36.3 3.2 3.6.5 36.2 3.66. 3.6 3.. 3.2 63.3 3.3 36 3.. 3.3 366 3.2 66.~ 2.3 3.3 $16». 66.3 2.6 3.33 3.3 3.33 3.6 3.. 2... 3.. 6... 66.». 66.6 66.. .36 .66 3.6 N66 3.. 333 3.3 3.3 56.32 3.6. 333 3.6 3.. 56.3 3.6 66.6 3.3 3.6 36 66.6 3.. 66.6 3.6 36 363 3.3 2.6 362 2.3 636 3.6 56.6 3.5 3.6 36 66.. 6.6 2.6 366 36 36.6 63.6 65.6 3.3 3.2 63.6 2.36 3.3 63.. 3.6 3.6 566 2.6 2.6 3.6 26 6.6 66.6 2.6 66.6 2.6 3.6 363 2.2 26 53.5 2.. 26 3.6 3.6 a 6»... 36.6 .56 6.6 36.6 36.6 66.6 66.6 66.6 3.6 26 2'2 36 m6 6 36.6 5.6 66.6 66.6 3.6 66.. 55.. 3.6 66.6 66.6 36 366 66.6 66.6 66.6 66.6 36.6 26 216. 63.3 66.6 36.6 2 .6 666 2 .6 63.6 36.3 53.. 2 .6 66.6 666 86 66.6 66.6 66.6 66.6 66.6 66.6 36.6 3.6 66.6. 66.6 66.6 66.6 66.6 56.6 6 36.6 2.3 3.3 2 .6 36.6 66.6 66.6 66.6 66.6 66.6 66.6 66.6 36.6 36.6 3V 8% :8: x350 250 335 4:0 AE< 1.2.4. 4.3. ESQ «83202 SEEM. E303. E200 320:9. 62353 mama—.68.". :8 :3 5:05. .333 .3 .568 25.50 + .EEméoZ ..nEm + .825. dw< =< 44¢. coo—.220 5:927. :53. 2:... ~82 A6 .825:ng 666.62 tn 2? 8539:. .3953 “Suwmu.mmwl.5-< 592. 2:0 .26 65 2 flag .5. 8:: 5.3 60235. 33 42.6 + 1...< ..=v< 5:0 2 2 36:: 86.». .86 8:: 5.3 602.65“ 9.53 3.7. 202.6660: 6 .33. “SF—moi 22.3 2.. m. «36 .«0 3.60m a 66.36 3.6 3.33 2... 3.3 2 .3 2.. .3 3.2.. 3.33 6343 «.32 65.3 66.33 35.63 3.63 3.63 +36 2235 3.6 35.62. 36.2 V6.3 363 2.. 3 3.32 363 66. .5 35.6.. 6.33 223 3.3. 3.3 2 .66 «.66.. 2.666 .36 55.635 65.2 V6.2 2.63 3.3. 3.3. 3.363 2.65 3.3 513 2.33 3.3 36v 3.2.. 35.35 36.33 36 3.63 36.6. .62 6Y3 3.3.. 35.56. 5.363 2....6 33.66 35.3 5663 2.3 3.3 63.2. 65.65 66.656. 3.6 35.33 3.5 3.6. 3.63 3.3. 36.. 63.3.. 3.6». 3.3 2.3 3.23 V6.36 35.36 3.3 36-66 36.53 36 66.2.. 2.3 36.3 3.3 366 36.3 3.63 56.3 3.3 63.63 63.33 3.69 3.3 w56v «6.66 6663 9.6 3...: «VJ. 3... 2.2 2.2. 56.3 66.63 3.63 3.3 3.2 2.32 5.3 3.3. 36m 333 65.62 366 3.3 66.3 53.3 3.2 63.3 66.3 2.3 3.2 ...v. .63 35.53 66.63 3.3 3563 3-63 3.2. 36.. 3.3 66.. 2.3 3.2 2.2 3.6. 65.5. 65.6 2.3 66.3 26v 3.x 63.2 3.2 372. 63.66 2.. 5v.. 2... 3.. 3.2 63.6 3... 36.6 3.3 35.3 66.. 36.2 3.3 3.6 3.5 312. 63.3 2.3 :6 8.. 36.. 3.... 3.3 36.. 2.3 3.6 36.. 36 3.5 3.6 5.... 63.“ 323 36.63 6m.v 36.6 3.6 3.. 65.5 2... 666 2.3 6~.6 35.6 26 3.. 2.6 36.6 63.. 3-63 65.3 63.6 36.6 3.6 63.. 65.6q 65.6 2.6 3.. 36.6 3.6 66.6 636 366 63.6 66.6 3.3 3.6. 66.6 66.6 2.6 2... 3.3 3.6 56.6 36 2.6 2.6 366 2.6 36.6 5.6 36 363 66.2 3.3 2.6 3.6 3.3 a 3.6 63.6 36.6 2.6 36.6 36.6 66.6 26 66.6 2.6 63.6 23. 3.6 36.6 66.6 66.6 3.6 65.. 33.3 . . .6 36.6 66.6 66.6 36.6 66.6 66.6 36.6 66.6 5. .6 v. .6 1.2 V5... 2.6 366 2.6 3.6 3.3 35.3 66.6 66.6 66.6 66.6 m66 66.6 66.6 366 66.6 66.6 66.6 3:3 mm... 5V6 66.6 66.6 6 3.6 36.3 33 366 36.6 66.6 66.6 66.6 66.6 66.6 66.6 66.6 2.6 66.6 3V 3.: «ES—.8028: 888.... 2.3. :..2< 4.).< ..1.< =35 «Sousa: .56qu 8582 .580 3223.“. sum—=03 awn—.602 =8 :8 5:63.. 2.23 b. .550 .33... + .EEméoZ .EEm + .325 dw< ..< .3" 32.220 5.22... ..=6< 2:... aw...“ .3 .335 3\ 666.62 3Q 23 8:26.35 S23... uEuan.~MVl.6.< 66.2... NC] MONOGRAPHS, NUMBER 2. I986 62 the percentage currently detected in situ. Inasmuch as this substantial difference was likely to have a dramatic impact on cancer mortality in the year 2000, in situ cancers of the breast and cervix were included as disease stages and as part of the overall incidence rate. Because the death rate of in situ carcinomas is negligible, no adjustment was made to the site—specific mortality rates in these sites. The prob- lem apportioning the lung cancer incidence rate to small and non-small cell lung cancer was handled in the same way the site-specific mortality rates were, i.e., with sepa- rate cell type-specific runs by the SEER group. The testicu- lar nonseminoma cancer incidence rates represent 51.6% of the total age-specific testicular incidence rate. As before, incidence rates for the 85 and over group from the SEER Program were used for the last three age categories. Recent data from a variety of sources, including the SEER Program, suggest important secular trends in age- adjusted incidence rates. These trends were incorporated into the projections. The stage at which the cancer is detected is addressed as follows: For most of the cancer sites, the SEER Program data are complete, and the incident cases can be appor- tioned into appropriate stages, usually stages I through IV. There are several notable exceptions to use of this coding scheme, in particular, prostate, bladder, and breast can- cers. For breast cancer, five disease stages were used, sim- ilar to the system constructed by Young et a1. (9), including in situ cancer and stage I, II, and IV as separate stages, and stages 111a and IIIb combined into a single stage. An important concern was that a significant proportion of the SEER Program cases are unstaged because of limitations in the patient records. An examination of the survival data of the unstaged, in contrast to the staged, cases shows that, for most cancers, the unstaged cases are evidently not a random sample of the remaining cases. In particular, for a cancer such as stomach cancer, the survival data from the SEER Program show these unstaged cases to be uni— formly on the low side; thus most are likely to be late stage. Therefore, the unstaged cases should not be allo- cated in the current SEER program stage proportions. Allocation of the unstaged proportion so that the over- all survival would remain the same after the allocation as before was considered the most desirable approach. How— ever, in instances when there are more than two stages, this allocation is not unique. Another desirable feature of the proportional allocation of unstaged patients would be that the distribution by stage of the staged and unstaged patients with cancer at a given site would look as similar as possible to the original stage distribution, but with the original constraint retained that the weighted survival distribution of all patients, staged and unstaged, would remain (close to) the same. A quadratic mathematical programming problem can be constructed. The objective function, minimizing the sum of squared deviations be- tween the new and original stage proportions minus the unstaged patients, subject to the constraints that all the unstaged patients are allocated to a stage and that the weighted survival distribution of the unstaged patients after staging is equivalent to the preallocation survival, can be solved with the technique of Lagrange multipliers, letting the proportional assignments be unconstrained. The results as input to the modeling efforts here are shown in tables A-8 and 9. CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 TABLE A-8.—Five-year relative survival and stage distribution, male sites” Buccal cavity All stages Key: 0.5918 Stage Pharynx + larynx All stages Percent in stage (where 0.52l0 applicable) Lung 5—yr survival Small cell All stages 0.0323 Non—small cell All stages 0.1294 Esophagus I II III IV 0.04 0.32 0.23 0.41 0.1272 0.0556 0.0515 0.0264 Stomach I II III IV 0.04 0.13 0.295 0.535 0.6976 0.4853 0.1704 0.0253 Pancreas All stages 0.0255 Colon I II III IV 0.19 0.32 0.25 0.24 0.8816 0.7562 0.4607 0.0539 Rectum I II III IV 0.31 0.23 0.25 0.21 0.7802 0.6336 0.3605 0.0390 Bladder LOC NOS I II III IV 0.36 0.37 0.03 0.13 0.11 0.8349 0.8708 0.7025 0.4773 0.2317 Melanoma I II III IV 0.46 0.43 0.07 0.04 0.9400 0.7170 0.4692 0.1856 Childhood Brain All stages 0.4922 ALL All stages 0.5557 AML All stages 0.2051 Adult ALL+AML All stages 0.0714 DHL All stages 0.3810 Prostate B C D 0.59 0.16 0.25 0.8442 0.7727 0.3054 Testis, nonseminoma I II III IV 0.66 0.05 0.16 0.13 0.8940 0.9012 0.8009 0.4203 All other male sites All stages 0.3958 " Source of data is the SEER Program, I984. LOC NOS = localized, not otherwise specified. Estimates of relative survival, a measure of mortality among a specific group compared with that of the general population, were also required for the model. The SEER Program is an excellent source for these data by cancer site and stage (tables A-8, 9). As noted above, the availa- bility of the survival distribution coupled with the staging information, even though in some cases incomplete, allows considerable flexibility with the input data and with the modeling effort in general. SCENARIO CREATION Each Working Group identified cancer control objectives that, if met, were believed to lead to reductions in cancer 63 TABLE A-9.—Five-year relative survival and stage distribution, female sites" Buccal cavity All stages Key: 0.6121 Stage Pharynx + larynx All stages Percent in stage (where 0.4897 applicable) Lung 5-yr survival Small cell A11 stages 0.0656 Non-small cell All stages 0.1736 Esophagus 1 11 11] 1V 0.04 0.38 0.22 0.36 0.1022 0.1029 0.0510 0.0110 Stomach 1 11 111 IV 0.04 0.15 0.30 0.51 0.7834 0.5081 0.1978 0.0288 Pancreas A11 stages 0.0285 Colon l 11 111 IV 0.17 0.32 0.27 0.24 0.8917 0.7921 0.4686 0.0666 Rectum 1 11 111 IV 0.30 0.22 0.26 0.22 0.8095 0.6590 0.3942 0.0600 Bladder LOC NOS 1 11 111 IV 0.33 0.36 0.02 0.17 0.12 0.8383 0.8530 0.6526 0.4227 0.1332 Melanoma 1 11 111 IV 0.53 0.40 0.04 0.03 0.9587 0.8175 0.4876 0.2337 Childhood Brain A11 stages 0.5382 ALL All stages 0.6416 AML All stages 0.1961 Adult ALL+AML A11 stages 0.0774 DHL A11 stages 0.4256 Breast In situ I 11 111 IV 0.0464 0.1277 0.5490 0.2014 0.0755 1.0 0.9597 0.8258 0.5986 0.1665 Ovary 1 11 111+1V 0.25 0.03 0.72 0.8214 0.6326 0.1904 Cervix In situ 1 11 111 IV 0.79 0.11 0.04 0.04 0.02 1.0 0.8501 0.5475 0.3684 0.1363 Corpus uteri 1 11 111 IV 0.80 0.08 0.03 0.09 0.9346 0.7129 0.5177 0.2339 All other female sites All stages 0.4268 " See footnote a, table A-8. incidence and mortality. The objectives were aimed at re- ducing the incidence of certain cancers, bringing people into the medical care system for early diagnosis and treat- ment and prolonging survival by the diffusion of improved treatment practices. To estimate the impact of achieving these objectives on cancer incidence and mortality, the Working Groups had to translate each objective into a set 64 of numerical parameters to be used by the computer pro- gram. (For example, smoking fewer cigarettes changes a former smoker’s relative risk of developing lung cancer. The amount of change in the person’s risk and the time to full effect of cessation are just two of many parameters that must be considered in any projection of the effects on lung cancer incidence of a smoking reduction program.) A de— scription of this computer-based model is included else- where in this monograph (1,2). Each cancer site may be affected by any or all of the three categories of cancer control objectives: prevention, screening, and treatment. The set of parameters used in projecting the impact of a specific cancer control activity on the incidence and mortality of a specific cancer site, TABLE A-10.vlndex to cancer control scenarios" Cancer control program activity Site/ type Prevention Screening Treatment Males Prostate + — Bladder Lung Small cell Non-small cell Colon Rectum Melanoma Testis, nonseminoma DHL Adult ALL + AML Childhood Brain ALL AML Buccal cavity Pharynx + larynx Pancreas Esophagus Females Breast Cervix Bladder Lung Small cell Non-small cell Colon Rectum Melanoma Ovary Corpus uteri DHL Adult ALL + AML — — Childhood Brain ALL — — AML Buccal cavity Pharynx + larynx Pancreas Esophagus |++++ + I 1 ++ | I I ++++++|+ | | |+++ ++++l I I 1 | I++++ +++ I ++ +++ + l | I I I+++ +++++++I+ 1 | ++++| I I “ Programs all started in 1985, continue through year 2000. and admit new cohorts each year as starting age is attained. Plus sign indicates which cancer control program activity is included in the model for the cancer site. NCl MONOGRAPHS. NUMBER 2, 1986 e.g., prevention of lung cancer, is termed a scenario. The various scenarios are outlined in tables A-10—13. Table A-10 is an index to the following tables and indicates which cancer sites are affected by particular recommenda- tions. The following sections describe these scenarios ac- cording to whether the recommendations refer to preven- tion, screening, or treatment. Prevention Scenarios The Prevention Working Group set explicit objectives for three areas: smoking, diet, and occupational risk fac- tors. However, only the flrst two were modeled. Objectives related to improvement in the exposure profiles of those in certain high-risk occupations were not easily dealt with in a modeling scenario because the data for such estimations are limited and do not convert into scenarios conforming to the Working Group’s recommendations. The smoking scenario was taken directly from the Work- ing Group’s recommendations. Initially, lung cancer inci— dence and mortality trends were predicted for the remain- der of the century from a combination of past trend data on lung cancer mortality; limited information on the peak prevalence of smoking among men and women, estimated by Harris (10) from retrospective examination of the Health Interview Survey; and data reported by Doll and Peto (II). For each age—sex cohort, the year in which the peak lung cancer incidence and mortality would occur was estimated; we used time—trend data from 1969 through 1982 to estimate how rapidly current mortality would rise to that peak, followed by a relatively modest turnaround in subsequent years. Detailed descriptions of the modeling procedures used for the lung cancer incidence and mortal- ity projections are in preparation. These projections implic- itly assume that smoking behavior in the United States will continue its slow decline evident in recent surveys, particularly the Health Interview Survey, but will not drop dramatically. The proportion of quitters was estimated to increase from approximately 2% to 4% per year. We ran additional scenarios on the model to examine the effects of even greater smoking cessation, up to 10 times the cur- rent quit rate, on cancer mortality in the year 2000. Using available epidemiologic evidence, primarily cross- cultural and animal model studies, the Prevention Work— ing Group recommended changes in the amount of fat and fiber in the diet of the population that would reduce cancer incidence rates for several organ sites. The incidence rate of prostate cancer may be reduced by as much as 25% by changes in fat and fiber consumption. The diet scenario (table A-l 1) for elderly men from 65 years of age and over illustrates that a reduction of 15% of the incidence rate in prostate cancer would have a significant impact on subse- quent mortality from this disease. These same dietary objectives would also affect the incidence of colon and rec- tal cancer. In both cases, the estimated potential reduction TABLE A-l 1.—Prevention scenarios: Parameters used for estimates of the impact of the cancer control objectives on mortality“ Specification of program effect ‘ Proportion Proportion Reduction in Relative risk ' Age of Proportion of total of served mcrdence rate Delay until target of total population population among served Before At maximum maximum lnter- population, population with risk with risk population at inter- program program Site/type vention yr served, % factor. % factor, % maximum effect, % vention effect effect, yr Males Colon Diet 250 100.0 100.0 100.0 50.0 5 Rectum ” " " ” ” " ” Prostate ” Z65 " ” " 15.0 " Lung Small cell Smoking 220 9.8 34.8 ” 14.42 2.0 15 Non-small cell " ” " ” ” " ” " Bladder " ” " " ” 2.44 1.0 ” Buccal cavity " ” " ” ” 4.81 " " Pharynx+1arynx ” ” " ” " 12.73 2.0 " Pancreas ” " ” ’ ” 1.90 1.0 ” Esophagus ” " ” ” ” 7.40 ” " Females Colon Diet 250 100.0 100.0 100.0 50.0 5 Rectum ” " ” " ” ” " Breast ” >45 " ” " 25.0 " Corpus uteri ” " ” " ” 15.0 " Cervix ” All 12.0 24 0 ” 1.40 1.0 " Lung Small cell Smoking 220 7.4 29.5 ” 12.87 2.0 15 Non-small cell ” ” ” ” ” " ” " Bladder ” ” ” ” ” 2.29 1.0 " Buccal cavity ” " ” ” ” 4.31 ” " Pharynx+1arynx ” ” ” " ” 11.77 2.0 " Pancreas ” ” ” ” ” 1.83 1.0 ” Esophagus " ” ” " ” 6.72 " " " Programs all started in 1985, continue through year 2000, and admit new cohorts each year as starting age is attained. CANCER CONTROL OBJECTIVES FOR THE NATION: 1985—2000 65 TABLE A-l2.—Screening scenarios: Parameters used for estimates of the impact of the cancer control objectives on mortality“ _ Cancer site Screening program parameters Breast Cervix Screening test Mammography and Pap smear physical exam- ination Age of target population, yr 50—70 20—70 Proportion screened under 0.633 0.702 program Relative risk of screened 1 1 population Stage distribution before screening program begins In situ 0.046 0.790 1 0.128 0.1 10 11 0.549 0.040 111 0.201 0.040 IV 0.076 0.020 Stage distribution for screened population at maximum program effect In situ 0.150 0.916 I 0.600 0.049 11 0.150 0.030 111 0.100 0.005 [V 0.000 0.000 " Programs all started in 1985, continue through year 2000, with 5-yr delay until maximum program effect is reached. in the incidence rate of colon and rectal cancer could be as high as 50% if the objectives could be achieved. Two other sites that would be affected by the fat~fiber scenario are cancers of the breast and corpus uteri. For breast cancer, the potential reduction among women 45 years and older is estimated at 25% of the current incidence rate; the reduction in the same age group of women for cancer of the corpus uteri is 15%. The prevention scenarios are based on the Working Group recommendations being converted into a hypothet— ical national cancer control program that is aimed at re— ducing the target risk factor. The computer program works with the user in an interactive manner to develop the pa- rameters of the control program, i.e., the parameters appearing in table A—ll. In the following example, the creation of the scenario for smoking reduction and its impact on lung cancer is described in detail. The primary group targeted is all adults (aged 20 yr and over). The current adult male smoking prevalence is 34.8% (see chap— ter 11), and the Working Group objective targets this proportion to decline to 25% by the year 2000. To reduce the smoking prevalence from 34.8% to 25%, one can con— ceive of a hypothetical program serving 9.8% of all males (i.e., 34.8% — 25.0%). Thus the parameter “proportion of served population with risk factor” (i.e., the population specifically targeted for smoking reduction) is 100%. The model also requires an estimate of relative risk or, alternatively, the percent of all cancer patients with the risk factor. The former is available from past prospective studies of the American Cancer Society (12) and other research. On the basis of these studies, a weighted average across smoking doses produces a relative risk of 14.4. A 66 direct estimate of the percent reduction in incidence rate in the served population can be used but is rarely available in risk factor studies. One can obtain an alternative by esti- mating the reduction in relative risk over a specified pe— riod. For lung cancer, a reduction of 86% (from 14.4% to 2.0%) over 15 years is used and is based on evidence from prospective studies. Finally, the program requires specifi- cation of starting and stopping years (1985 and 2000, re- spectively) for the hypothetical program and whether each new cohort attaining the beginning target age is subjected to the program. Screening Scenarlos The Screening Working Group recommended cancer control objectives for cervical cancer and breast cancer screening, both shown in table A-12. The cervical cancer screening scenario is as follows: Cervical cancer screening is TABLE A-13.~Treatment scenarios: Parameters used for estimates of the impact of the cancer control objectives on mortality" Stage” Site/type l 11 111 IV Males Colon 1.000 0.848 0.527 0.135 Rectum 1.000 0.905 0.541 0.098 Bladder 0.975 0.703 0.477 0.232 Melanoma 1.000 0.749 0.546 0.206 Testis, nonseminoma 0.950 0.950 0.900 0.428 Prostate" 1.000 0.902 0.337 DHL 0.693 Adult ALL+AML 0.298 Childhood Brain 0.746 ALL 0.689 AML 0.394 Lung, small cell 0.1334 Females Colon 1.000 0.888 0.536 0.167 Rectum 1.000 0.941 0.591 0.150 Bladder 0.955 0.653 0.423 0.133 Melanoma 1.000 0.854 0.567 0260 Breast 0.964 0.879 0.693 0.175 Cervix 0.903 0.767 0.491 0.157 Corpus uteri 0.969 0.768 0.594 0.324 Ovary“ 0.865 0.690 0.238 DHL 0.774 Adult ALL+AML 0.289 Childhood Brain 0.769 ALL 0.795 AML 0.377 Lung. small cell 0.214” " Treatment objective is expressed as state—of-the-art. 5-yr relative survival (in percent) by stage at diagnosis as identified by the Working Group. Estimates of current 5-yr relative survival are shown in tables A-8 and A-9. Programs all started in 1985, continue through year 2000, with 5-yr delay until maximum program effect is reached. h For sites with table entries only in the stage I column, data are for all stages combined. " Stages for cancer of the prostate are B, C, and D. d Figure for small cell lung cancer is 2-yr relative survival. Current 2-yr survival for males and females is 0.0842 and 0.1355, respectively. " Stages for cancer of the ovary are I, II, and 111 + 1V combined. NCl MONOGRAPHS. NUMBER 2, 1986 recommended for women from 20 through 70 years of age, with an estimated proportion of the population reached at approximately 78% per year. Furthermore, a complete workup by physicians to determine if some treatment is appropriate is recommended for 90% of the women who obtain a positive result on screening. The stage distribu- tion for cervical cancer will shift to the following propor- tions: in situ, 0.92; stage I, 0.05; stage II, 0.03; stage 111, less than 0.01; stage IV, 0. The potential effect that cervi— cal cancer screening will reduce future incidence by the detection and removal of precursors to cervical cancer has not been included in this analysis. We found the breast cancer screening scenario slightly more difficult to create. The objective includes screening 100% of the women between the ages of 50 and 70 by mammography and physical examination. It is estimated that two-thirds of them would be reached and that 95% of the women who test positive would receive a full workup. Estimating the stage shift for this screening program was difficult because the data were from the landmark study of the HIP of New York (13). These data have not been eval- uated for accurate stage distribution but show a 30% de- cline in mortality from breast cancer, presumably in large part due to the screening project. Thus we can estimate a stage distribution that would mimic a 30% decline in mor- tality. The estimated stage distribution of the screened women is as follows: in situ, 0.15; stage I, 0.60; stage II, 0.15; stage III, 0.10; stage IV, 0. In this projection, it is as- sumed that the stage—specific mortality rates for screened women do not change. Direct estimates from the HIP study on the stage distribution of screened women should be available within the year, and the model can be reeval- uated with this information. Treatment Scenarios The basis for the objectives set by the Treatment Work- ing Group is examination of observed 5-year survival rates versus data from recent clinical trials or the PDQ system. For lung cancer, 2-year survival data were used. An esti- mation of the degree of improvement that could be made toward state-of—the-art therapy was determined by group consensus. The estimated survival change was based on 5-year observed survival, whereas the model that has been used requires relative survival as basic data input to its treatment modification projections. Inasmuch as the ratio of specific observed survivals for two populations is equiv— alent to the ratio of relative survival from the same groups, the Working Group projected change from current ther— apy to state-of—the-art therapy by creating a ratio of the two observed survivals and multiplying by currently avail- able relative survival data in the model. The relative sur- vival rates from the most recently available SEER data are presented in tables A-8 and A-9 for each site and for each stage within site. Table A—l3 includes the Treatment Work- ing Group’s objectives as adapted for the model. Careful examination of the scenarios used in the model- ing exercise reveal that some of the recommendations had to be transformed to be compatible with the model. In par- ticular, there are several sites for which the Working Group used stage data different from the available input data to the model. The clearest example of this is breast cancer, which was categorized by tumor size, nodal status, CANCER CONTROL OBJECTIVES FOR THE NATION: 1985 2000 and metastasis. The resulting breakdown included nine categories, to which the current modeling efforts are not adapted. The modeling-based breast cancer stages include in situ; stage I, which includes small tumor size and no nodes or metastasis; stage II, which includes small tumor size with nodal involvement, no metastasis, and medium tumor size; stage III, which includes large tumors and all stage IIIb cases; and stage IV, which includes all cases of metastasized cancer. This breakdown is adequate for esti- mates of impact of the cancer control objectives in this site. Several cancers were combined so that the model was not expanded too greatly, e.g., adult ALL and AML. Al- though both the survival distributions and the Working Group’s recommendations are different for these two leu- kemias, adaptation of the recommendations was made pos- sible by the combination of SEER data for them. Because of the relatively small numbers of cases of ALL and AML, even considerable inaccuracy will not reflect substantially on the overall impact of the cancer control objectives. REFERENCES (I) EDDY DM: A computer-based model for designing cancer control strategies. NCI Monogr 2:75—82, 1986. (2) LEVIN DL, GAIL MH, KESSLER LG, et al: A model for projecting cancer incidence and mortality in the presence of prevention, screening, and treatment programs. NCI Monogr 2:83—93, 1986. (3) BUREAU OF THE CENSUS: General population characteris- tics. In 1980 Census of Population: Characteristics of the Population, United States Summary, chap B, vol 1, Part 1 (PC80—l-Bl). Washington, DC: U.S. Govt Print Off, 1983. (4) NATIONAL CENTER FOR HEALTH STATISTICS: Advance Report of Final Natality Statistics, 1979. Monthly Vital Statistics Rep 30. Hyattsville, MD: Natl Center Health Stat, October 20, 1981. (5) SPENCER G: Projections of the population of the U.S. by age, sex, and race: 1983 to 2080. In Current Population Reports: Population Estimates and Projections, Ser P-25, No. 952. Washington, DC: U.S. Govt Print Off, 1984. (6) FABER JF: Life Tables for the United States: 1900—2050. Actuarial Study No. 87. Baltimore: Social Security Ad— ministration Actuary, September, 1982. (7) FEINLEIB M: The magnitude and nature of the decrease in coronary heart disease mortality rate. Am J Cardiol 54:2C—6C, 1984. (8) HORM JW, ASIRE AJ, YOUNG JL JR, et al: SEER Program: Cancer Incidence and Mortality in the United States, 1973—1981. DHHS (NIH) Pub] No. 85-1837. Bethesda, MD: Natl Cancer Inst, 1984. (9) YOUNG .IL JR, RIES LG, POLLACK ES: Cancer patient sur- vival among ethnic groups in the United States. JNCI 73:341—352, 1984. (10) HARRIS JE: Cigarette smoking among successive birth co- horts of men and women in the United States during 1900780. .INCI 7124737479, 1983. (11) DOLL R, PETO R: The causes of cancer: Quantitative esti- mates of avoidable risks of cancer in the United States today. JNCI 66:1191—1308, 1981. (12) HAMMOND EC: Smoking in relation to the death rates of one million men and women. Natl Cancer Inst Monogr 19:127—204, 1966. (I3) SHAPIRO S, VENET W, STRAX P, et al: Ten- to fourteen- year effect of screening on breast cancer mortality. JNCl 69349355, 1982. 67 Appendix B: Analysis of the Impact of the Cancer Control Objectives ANALYSIS OF THE IMPACT OF THE CANCER CONTROL OBJECTIVES The general mathematical model and the cancer control programs derived from the recommendations of the Work- ing Groups of the Board of Scientific Counselors, described in detail in Appendix A (the scenarios, tables A-10—13), form the basis of the analysis of the cancer control objec— tives. This Appendix outlines the potential impact of achieving the cancer control objectives, measured in changes in overall cancer mortality for the United States. The analysis includes the effects of two factors which will have a major impact on the cancer burden in the United States in the next 15 years: the “quit rate” of cur- rent smokers (the rate at which current smokers become ex-smokers) and general improvements in state-of—the—art treatment for cancer, i.e., changes in the technology and practice of cancer treatment, as manifested in changes in cancer patient survival rates. The hypothesized increase in the state-of—the-art curve in figure IV-l represents an increase in the overall cancer survival rate. If this increase were to continue as it has over the past two decades, it would represent an additional increase in survival to that outlined in the recommendations of the Treatment Work- ing Group. Those recommendations and objectives (table IV-l) represent “closing the gap” between both current treatment and state of the art. The analysis which follows considers several potential levels of reduced smoking prev— alence to be achieved by the year 1990 and various changes in the state of the art of cancer treatment as reflected by changes in cancer survival. The impact of these changes is then reported as changes in the cancer mortality rate in the year 2000. CURRENT CANCER MORTALITY IN THE UNITED STATES Table 8-] presents recent figures on cancer mortality for all cancer sites in men, women, and the total population. The mortality rate for men has been increasing at an aver- age annual rate of 0.7% per year, from 220.71 per 100,000 population in 1970 to 238.75 per 100,000 in 1982 (age- adjusted to the 1980 population). Most of the change occurred early in this period, with the rate of increase slowing down in recent years. For women, the trend was negligible in the early 19705 but has been increasing more rapidly recently. The average annual rate of change between 1970 and 1982 was 0.3%, rising from 142.65 per 100,000 to 147.03 per 100,000 population. PROJECTIONS FOR THE YEAR 2000 Projecting cancer mortality for 20 to 25 years requires consideration of the interactions between population dy- namics and changes in the health care system. The analysis incorporates these interactions through three factors: inci- dence trends (influenced by prevention programs), changes in relative survival of cancer patients (influenced by screen- ing and treatment), and the aging of the population. Important trends in incidence were discussed in Appen- dix A. Only for a few sites were the age-adjusted trends large enough to make inclusion of explicit trends neces- sary. The sites with increasing incidence include colon and melanoma for both sexes and prostate cancer. The two decreasing sites are those of the cervix and uterine corpus. Although the number of cases of stomach cancer is still decreasing, the number is not sufficiently large within the “all other” category to warrant separate treatment. Lung cancer trends pose a special modeling problem. For men, the most recent data show a consistent increase in the ages 65 and above and in ages 45 and above for women. However, the age-adjusted trends show a decrease in the year-to-year percent change for white men (fig. B-lA). The trend indicated (without consideration given to any enhancement of the trend to reduce the prevalence of smokers) that age-adjusted lung cancer mortality among males would begin to decrease after 1985 or earlier. The figures for white women (fig. B—IB) show an earlier increase in year-to-year lung cancer mortality with, more recently, a decrease in the year-to-year changes. Without increased intervention, the age-adjusted rate appears to increase more slowly than in the past through 1990, with the rate not rising more than 50% above current levels. With lung cancer among females representing about 15% of lung cancer mortality and male rates projected to remain con- stant or to decrease, it is reasonable for one to estimate that the impact of the projected future increases in lung cancer on the overall age-adjusted mortality rate in the year 2000 will be about 5%. The NCI staff, in collabora- tion with outside investigators, has been studying several methods for projecting lung cancer for the next 15 years. In the near future, such detailed projections will be avail- able for inclusion in the ongoing modeling and surveillance efforts for assessment of the impact of the cancer control objectives. Changes in survival pose several modeling problems. There has been a significant trend toward improved sur- vival in recent decades. Analyses of survival trends in the 69 TABLE B-l.—Age-adjusted (1980 United States Standard) cancer mortality rates, all sites, all races. I970—82 Year Total Males Females 1970 175.65 220.71 142.65 1971 175.68 222.43 141.99 1972 177.27 225.72 142.80 1973 175.78 224.93 141.33 1974 176.90 227.56 141.65 1975 176.13 227.77 140.60 1976 176.78 230.89 142.94 1977 179.49 232.49 143.29 1978 180.61 234.60 144.00 1979 180.74 235.86 143.47 1980 183.12 238.30 146.05 1981 182.02 236.33 145.72 1982 183.57 238.75 147.03 SEER data, collected from 1973 to the present, show a 0.5% increase in survival on an annual basis, based on an analysis that adjusts for changes in incidence. This change is likely caused by a combination of factors, including bet- ter treatment for certain cancers, improved supportive care for those with cancer, statistical artifacts of earlier rec- ognition of this disease, improvements in the treatment of diseases other than cancer (particularly heart disease) that disproportionately prolong the survival of cancer patients, and increased access to medical care in the United States generally. It is important that this trend, which may in— crease in future years because of better technology, be in- cluded in the analysis. Three alternative figures for future gain are included, 0.5%, 1.0%, and 1.5% annual gain in the average survival rate. SITE-SPECIFIC IMPACT On a site-specific basis, the reduction in mortality for men and women is shown in tables B-2 and 3. This reduc- tion of approximately 21% is based on achievement of the prevention, screening, and treatment objectives, which include a reduction in smoking prevalence to 25% in the year 2000. Tables B-4 through B-8 show the reduction in incidence and mortality achieved through prevention, screening, and treatment, each acting alone. Each table shows 1) projections of the incidence and mortality rates for the year 2000 based on current trends without any new intervention (reference projection); 2) projections of inci- dence and mortality rates for the year 2000 based on achievement of the objectives outlined in this report; and 3) the percent reduction in the rates that would occur if the objectives are met. One should note that these tables do not account for changes in survival due to future advances in the state of the art nor for the substantial reduction in smoking-related mortality that would be achieved if the smoking objectives are reached well before the year 2000. That analysis for various estimates of the annual percentage increase in the 5-year relative survival rate and for various levels of smoking reduction is reported later. The mortality reduction in percentages varies consider- ably from site to site, with substantial mortality reduction in major sites including colon, rectum, breast, prostate, and lung. The projections of total lives saved in the year 2000 that would stem from the cancer control objectives having been met indicate that prevention efforts contrib- ute slightly less than one-half of the reduction; screen- ing contributes about 3% for women; and treatment, with reference to closing the gap between the current and the estimated state-of—the-art figures, contributes slightly more than one-half to the total reduction. These estimates should be considered preliminary and are subject to further refine- ment. The projections do not include certain multiple pre- vention programs acting on the same site, such as the im- pact on breast cancer by an increase in dietary fiber and a decrease in fat, although such an impact is likely to result in only second-order effects on further reductions. FURTHER REDUCTIONS IN CANCER MORTALITY To increase the reduction in mortality beyond the 25% level, we need a further decrease in smoking prevalence to below 25% of adults by the year 1990 and maintenance, if 50 A 50 -8 T 40 _ 40 _ 30 _ 30 _ m 20 __ Lu 0 o z z < < :1: z: u u +— I: a a u 10- u M 1!. 1.1.1 Ll-l r— .20 _ v— 720 _ an _ so _ -4(1 _ 40 _ 511lllllllllllllllllllllllllllllll .5011||||||||||1||11111||||||||||_] 1950 19611 1971) 1981) 1950 1961) 1970 1980 YEAR 0'" DEM” YEAR or DEATH FIGURE B—l.iYear-t0-year percent changes in lung cancer mortality rates from 1950 to 1981 for white males (A) and females (B), all ages. 70 NCl MONOGRAPHS, NUMBER 2. 1986 TABLE B-2.—Estimated impact of the cancer control objectives in the year 2000: Males Incidence Mortality Projection . Projection . Reduction, Reduction, Site/ type Reference Objectives % Reference Objectives % Buccal cavity 11.12 9.35 15.9 5.48 4.88 10.9 Lung Small cell 14.58 11.45 21.5 13.69 10.40 24.0 Non-small cell 85.58 67.25 21.4 70.33 56.68 19.4 Colon 49.43 26.68 46.0 21.66 9.62 55.6 Rectum 18.29 9.90 45.9 10.44 3.70 64.5 Bladder 24.37 22.09 9.4 8.40 6.72 20.0 Melanoma 18.54 18.54 " 8.32 6.39 23.2 Childhood Brain 0.57 0.57 a 0.30 0.15 49.9 ALL 0.67 0.67 “ 0.39 0.28 28.7 AML 0.11 0.11 " 0.09 0.06 23.8 Adult ALL+ AML 3.23 3.23 " 2.86 2.16 24.4 DHL 2.18 2.18 " 1.28 0.64 50.4 Prostate 74.48 66.20 11.1 31.57 16.71 47.1 Testis, nonseminoma 2.05 2.05 " 0.41 0.27 32.5 Esophagus 5.31 4.28 19.4 4.88 4.01 17.8 Stomach 12.59 12.59 " 10.08 10.08 " Pharynx+1arynx 12.89 10.24 20.6 7.57 6.48 14.4 Pancreas 10.42 9.73 6.7 9.74 9.14 6.2 All other sites 117.15 117.15 " 69.92 69.92 " Total 463.56 394.26 14.9 277.41 218.29 21.3 " No interventions were related to these cancer sites. TABLE B-3.7Estimated impact of the cancer control objectives in the year 2000: Females Incidence Mortality Projection , Projection . Reduction, Reduction, Site/type Reference Objectives % Reference Objectives % Buccal cavity 4.27 3.75 12.1 2.06 1.89 8.2 Lung Small cell 13.67 11.23 17.9 11.66 8.70 25.3 Non-small cell 64.59 53.07 17.8 47.33 39.69 16.1 Colon 34.00 18.63 45.2 14.47 6.39 55.8 Rectum 11.05 6.10 44.8 5.85 1.99 66.0 Bladder 6.29 5.86 6.9 2.03 1.63 19.8 Melanoma 14.09 14.09 a 5.66 4.16 26.5 Childhood Brain 0.46 0.46 " 0.21 0.11 50.0 ALL 0.65 0.65 " 0.35 0.22 38.6 AML 0.11 0.11 " 0.08 0.06 22.5 Adult ALL+ AML 1.93 1.93 " 1.68 1.30 22.9 DHL 1.70 1.70 " 0.92 0.36 60.7 Breast 80.07 64.00 20.1 32.11 20.39 36.5 Ovary 12.20 12.20 " 7.65 7.06 7.6 Cervix 19.08 18.24 4.4 1.85 0.71 61.6 Corpus uteri 10.78 9.29 13.8 1.87 1.18 37.1 Esophagus 1.64 1.39 15.7 1.47 1.26 14.3 Stomach 5.50 5.50 " 4.18 4.18 ” Pharynx+1arynx 3.14 2.61 17.0 1.99 1.75 11.8 Pancreas 7.18 6.82 4.9 6.59 6.29 4.5 All other sites 69.39 69.39 a 37.09 37.09 a Total 361.79 307.02 15.1 187.10 146.57 21.7 " See footnote (1, table B-2. CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 71 TABLE B—4.——Ertimated impact of the cancer control objectives in the year 2000 from prevention only: Males lncidence Mortality P ' t' P ' t' rejec ion Reduction, rOjec ion Reduction, Site/ type Reference Objectives % Reference Objectives % Buccal cavity 11.12 9.35 15.9 5.48 4.88 10.9 Lung Small cell 14.58 11.45 21.5 13.69 10.99 19.7 Non-small cell 85.58 67.25 21.4 70.33 56.68 19.4 Colon 49.43 26.68 46.0 21.66 11.93 44.9 Rectum 18.29 9.90 45.9 10.44 5.91 43.3 Bladder 24.37 22.09 9.4 8.40 7.87 6.3 Esophagus 5.31 4.28 19.4 4.88 4.01 17.8 Pancreas 10.42 9.73 6.7 9.74 9.14 6.2 Pharynx+larynx 12.89‘ 10.24 20.6 7.57 6.48 14.4 Prostate 74.48 66.20 1 1.1 31.57 28.97 8.2 All other sites 157.09 157.09 " 93.65 93.65 " Total 463.56 394.26 14.9 277.41 240.51 13.3 a See footnote a, table B-2. TABLE B-5.—Estimated impact of the cancer control objectives in the year 2000 from prevention only: Females Incidence Mortality Pro’ect'on Pro'ect'on J I Reduction, J I Reduction, Site/ type Reference Objectives % Reference Objectives % Buccal cavity 4.27 3.75 12._1 2.06 1.89 8.2 Lung Small cell 13.67 11.23 17.8 11.66 9.75 16.4 Non—small cell 64.59 53.07 17.8 47.33 39.69 16.1 Colon 34.00 18.63 45.2 14.47 8.06 44.3 Rectum 11.05 6.10 44.8 5.85 3.35 42.8 Bladder 6.29 5.86 6.9 2.03 1.92 5.7 Esophagus 1.64 1.39 15.7 1.47 1.26 14.3 Pancreas 7.18 6.82 4.9 6.59 6.29 4.5 Pharynx+1arynx 3.14 2.61 17.0 1.99 1.75 11.8 Breast 80.07 64.00 20.1 32.11 27.35 14.8 Cervix 19.08 18.24 4.4 1.85 1.78 4.2 Corpus uteri 10.78 9.29 13.8 1.87 1.62 13.6 All other sites 106.03 106.03 " 57.82 57.82 " Total 361.79 307.00 15.1 187.10 162.53 13.1 " See footnote a, table B-2. not enhancement, of the trend in cancer patient survival observed over the past two decades. The ability to decrease smoking prevalence below 25%, e.g., to achieve a 15% an- nual smoking prevalence, depends on our adopting a na- tional resolve to reduce smoking prevalence and on effec- tive application of today’s knowledge on how to change smoking behavior. As noted in chapter 11 of this report, the trend in smoking prevalence is down and, if maintained at the current rates, should reach a 25% smoking prevalence by the year 1990. Cancer survival, represented by 5—year relative survival rates, has improved signficantly over the last two decades. Despite a change in the mix of incident cases toward sites 72 TABLE B-6.—Estimated impact of the cancer objectives in the year 2000 from screening only: Females Mortality P 'e to to] Cl n Reduction, Site / type Reference Objectives % Breast 32.11 26.98 16.0 Cervix 1.85 1.07 42.1 All other sites 153.14 153.14 " Total 187.10 181.19 3.2 " See footnote a, table B-2. NCl MONOGRAPHS, NUMBER 2. 1986 TABLE B-7.—Estimated impact of the cancer control objectives in the year 2000 from treatment only: Males Mortality Projection . Reduction, Site/ type Reference Objectives % Lung, small cell 13.69 12.95 5.4 Colon 21.66 17.44 19.5 Rectum 10.44 6.46 38.1 Bladder 8.40 7.17 14.6 Melanoma 8.32 6.39 23.2 Childhood Brain 0.30 0.15 49.9 ALL 0.39 0.28 28.7 AML 0.09 0.06 23.8 Adult ALL + AML 2.86 2.16 24.4 DHL 1.28 0.64 50.4 Prostate 31.57 18.20 42.4 Testis, nonseminoma 0.41 0.27 32.5 All other sites 178.00 178.00 " Total 277.41 250.17 9.8 “ See footnote a, table B-2. that show poor survival, most particularly lung cancer, the 5-year relative rate has climbed from 38% among patients diagnosed in the period 1960 to 1963 (based on the End Results Study) to 49% among patients diagnosed between 1976 and 1981 (based on data from the SEER Program) (fig. B-2). The End Results Study and the data from the SEER Program are not strictly comparable because of dif- ferences in the geographic reporting sites, but they do show the general trend in survival. The improvements have been most notable among children. The trend is particularly sharp in diseases for which spe- cific treatment advances have been documented, e.g., TABLE B-8.7Estimated impact of the cancer control objectives in the year 2000 from treatment only: Females Mortality Projection , Reduction, Site / type Reference Objectives % Lung, small cell 11.66 10.60 9.1 Colon 14.47 11.46 20.8 Rectum 5.85 3.46 40.9 Bladder 2.03 1.73 14.8 Melanoma 5.66 4.16 26.5 Childhood Brain 0.21 0.11 50.0 ALL 0.35 0.22 38.6 AML 0.08 0.06 22.5 Adult ALL + AML 1.68 1.30 22.9 DHL 0.92 0.36 60.7 Breast 32.11 27.45 14.5 Ovary 7.65 7.06 7.6 Cervix 1.85 1.36 26.6 Corpus uteri 1.87 1.36 27.3 All other sites 100.71 100.71 “ Total 187.10 171.40 8.4 " See footnote a, table B-2. CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 100 _. I White - Children 90_ O Whites - All Ages A All Races - All Ages 80 _ Cl Blacks - All Ages p Z 70 _ Lu 2 m 60 ._ B E 50 __ q < 40 ._ 2 E 30 _ :J m 20 _. 10 ._ 0 1 | I J 1960-63 1970-73 1973—75 1978-81 FIGURE B-2.~Five-yr relative survival rates, all cancer sites combined. Data are from the End Results Study and the SEER Program. Hodgkin’s disease, the childhood tumors, testicular cancer, and others, but is also noticeable in other cancers as well. Both new technologies and additional vigilance on the part of the physician and lay public (i.e., toward early detection) may be one cause for the survival increase. However, other factors may also enter into the increase. For example, some of the statistics may be biased because of earlier diagnosis with no concomitant change in benefit for the patient. Al- though some of this bias is likely in the data, the data also reflect real medical advances as well. On the other hand, the increased access to appropriate medical care through- out the period from 1960 to 1980, resulting principally from the increase in public (Medicare and Medicaid) and private health insurance, may also be a factor in the sur- vival increase. The model has been adopted to account for various trends in 5-year relative survival with the use of the modi— fied exponential survival model. The model assumes an improvement in survival per year that is then translated into a percent decrease in the proportion of the cancer in- cident cases exposed to the hazard component of the sur- vival equation. The analysis shows that an increase in cancer patient sur- vival is an extremely important cancer control component in the efforts of health professionals to reduce mortality. Prevention and screening alone may account for a 20% to 30% reduction in the mortality rate; further reductions must come from improved prognosis. The last several years have given rise to many new possibilities in cancer treat- ment (e.g., through molecular biology), and continued progress has been made in conventional therapy. The extent to which this research, resulting in improved sur- vival, might affect mortality can be determined from the analysis outlined here. The projections of the impact of meeting the cancer con- trol objectives (fig. B—3) show the decrease in the age-ad- justed cancer mortality rate in the year 2000 for various 73 300 - Smoking: Survival: Percent Percent of Annual - Adults Gain (1990) (All Sites) Z 9 32 0 E 21% Mortality Reduction __, 200 - E - - _ _ _ 2 HF”— \ \3: ~‘:\\~ \ \ ‘ 23 O D \\:\:\ \:\\ \‘\ o _ x \ \\ \\ \‘ g; ‘ :t‘:\\:\\\/ 15 0 y—1 \ \ \‘ m 50% Mortality Reduction / \\\\:\\:\/ LL] \ \ \ “— \\ ‘~ "‘" 25 1 0 § 100 — \ \ \ l—t \ é \ 15 1.0 15 1.5 () l l l l l l 1970 1980 1990 2000 YEAR OF DEATH FIGURE B—3.—Projections of the cancer mortality rate (age—adjusted to 1980) based on adult smoking prevalence in 1990 and annual changes in survival. The achievement of the diet, screening, and treatment objectives (table H) is assumed. assumptions about the percentage of adults smoking in the year 1990 and for several alternative projections of the annual overall increase in cancer survival. Meeting the smoking reduction goal before the year 2000 will result in a greater decrease in mortality by the year 2000. (The analy— ses in which the 15% smoking goal is met by 1990 also in- clude the assumption that the percentage of adults who smoke continues to decline gradually to 10% by the year 2000.) Under the assumption of a 50% decrease in smoking prevalence in adults between 1983 and 1990 (Le, 15% of adults smoking in 1990) and a 1.5% gain in cancer survival on an annual basis, the estimated decrease in cancer mor- tality rates is 50%, if one assumes that the prevention, 74 screening, and treatment objectives are all met. Under these circumstances, the approximate reductions by cancer con— trol activities are as outlined in table 1-4: prevention, 23%; screening, 3% for females only; and treatment, 26%. SUMMARY Meeting the objectives will lead to a dramatic reduction in the cancer mortality rate from the projected levels for a number of cancer sites. Even greater reductions are possi- ble if smoking levels fall to below 20% by the year 1990, and if the cancer survival rate increases at an even greater rate than observed during the past decade. A Computer-based Model for Designing Cancer Control Strategies 1 David M. Eddy 2-3 ABSTRACT—An analytic framework for designing cancer control strategies and a computer model based on the framework are described. The framework and associated computer-based model, titled CAN*TROL, enable a planner to 1) specify a popu- lation and the cancers that affect that population, 2) define a set of cancer control activities for the population, and 3) calculate the effect of the set of activities on the incidence, prevalence, mortality, and cost of cancers in the population in future years. For an analysis of realistic cancer control problems, it enables the planner to identify subpopulations that differ from the rest of the population with respect to risk factors, incidence rates, stage at detection, effectiveness of treatment, costs, and behavioral characteristics that might affect the impact of a cancer control program. A planner can specify virtually any combination of ac- tivities (spanning primary prevention, screening, early detection, and treatment), at different times, for different age groups, with different acceptance and adherence rates. This computer-based model has been used by the Cancer Unit of the World Health Organization and the Ministry of Health of Chile. The frame- work is also the basis for the computer model used by the National Cancer Institute to help set and analyze the cancer con- trol objectives for the year 2000.—NCI Monogr 2:75-82, 1986. The design of comprehensive and efficient strategies for controlling cancer in a population is complicated by sev- eral factors. For any population, there are dozens of can- cers, each with its own age-specific incidence rates and mortality rates, risk factors, and high-risk groups. For each cancer, there is a wide range of activities which can be ap- plied that span prevention, screening, early detection of signs and symptoms, diagnosis, treatment, surveillance, re- habilitation, and pain control. Within each of these major ABBREVIATIONS: NCI=National Cancer Institute; BCDDP= Breast Cancer Detection and Demonstration Project; HIP= Health Insurance Plan. I Supported by The Charles A. Dana Foundation, the National Cancer Institute, and the Cancer Unit of the World Health Organization. 2Center for Health Policy Research And Education, World Health Organization Collaborating Center for Research in Cancer Policy, Duke University, Box GM, Duke Station, Durham, NC 27706. 3I want to thank Judy Eddy and Cynthia Nichols of the Center for Health Policy Research and Education; Drs. Edward J. Sondik, Larry G. Kessler, and David L. Levin, Division of Cancer Prevention and Control, National Cancer Institute; Rosalie Anderson and Robert Smucker at Information Management Systems, Inc., Rockville, MD; Drs. Augusto Schuster, Cecelia Sepulveda, and Hector Rodriguez of the Ministry of Health, Santiago, Chile; Dr. Usha Luthra of the Indian Council of Medi- cal Research, New Delhi, India; and Drs. Jan Stjernswérd and Kenneth Stanley of the Cancer Unit of the World Health Organization, Geneva, Switzerland. categories are dozens of specific activities; decisions must be made about specific subpopulations, age groups, fre- quencies of examinations, choices of diagnostic tests, types of treatments, and other factors. Altogether, hundreds of activities are possible for controlling cancer in a popu- lation. Finally, all the possible activities have different benefits, risks, and costs, and take effect over different periods. This paper describes an analytic framework and com- puter model, CAN*TROL, that can be used by health care planners to help estimate the impact of different cancer control programs on the incidence, prevalence, mortality, prevalence of risk factors, and cost of different cancers and cancer control programs in future years. The paper also il— lustrates use of the model with a program for screening women over 65 in the United States with annual physical examinations and mammography. The underlying framework for CAN*TROL is a simple but powerful conceptual model for calculating the effects of cancer control activities, developed for the Cancer Unit of the World Health Organization (1). The framework and model have been used by this Cancer Unit and by the Ministry of Health in Chile. The framework also is the basis for the computer model used by the NCI to help set cancer control goals for the year 2000 in the United States (2,3). A version of CAN*TROL is to be made available through the NCI to states. DESCRIPTION OF THE FRAMEWORK The framework for the model is illustrated in figure 1 and consists of a set of “states” (denoted as circles in fig. 1) that partition the population that is the target of cancer control programs. State 0 consists of people who have not been diagnosed to have cancer. The model keeps track of the number of people in this state by sex and 5-year age groups and by calendar year. Each year, a proportion of these people will be diagnosed as having a particular cancer (a)," as determined by age- and sex-specific incidence rates. These cancers will be found in different stages (b), with the proportion of new cases found in stage 3, e.g., denoted as s; in figure 1.5 For any particular stage, some people will die before the end of the year (c). Cancer patients who do not die in the first year will be in states 1, 2, 3, 4, or 5 at the end of the year, each state corresponding to a stage of the cancer, as determined at the time of diagnosis. People in 4The letters in parentheses refer to parts of the framework illustrated in figure 1. 5The definitions of the stages depend on the particular cancer being analyzed. 75 . 0!) STATE 6 Dead of FIGURE LAConceptual framework of the model. state 0 also have a chance of dying of other causes (d), as determined by age- and sex-specific mortality rates. People who neither develop cancer nor die of other causes remain in state 0 at the end of the year (e). In any year, people with cancers detected in various stages have a probability of dying of their cancer (1), as determined by stage-specific relative mortality rates (represented in the figure as m. through m5). Some cancer patients will also die of other causes (g). If a patient neither dies of the cancer nor of other causes in the year, he or she will remain at the end of the year in the state labeled for the stage at the time of diagnosis. In any year, a patient’s cancer might advance to a later stage, but this framework does not deal explicitly with this progres- sion of disease. Rather, the biologic effect of a cancer pro- gressing from stage to stage is reflected in the framework through stage-specific relative mortality rates. Because the relative mortality rates are a function of stage at the time of diagnosis, the probability that the cancer patient is alive at the beginning of any year is determined by the stage at the time of diagnosis, and there is no need to model the progression of disease through stages. The framework accounts for people who die of cancer or other causes through states 6 and 7 (h). The framework is designed to facilitate calculating (by sex and age group) for each year in the future I) the number of people in each state at the beginning of the year, 2) the proportions of people who will make transi- tions between various states, and 3) the number of people in each state at the end of the year. The framework con- cludes the calculations for each year by incrementing the 76 age of everyone in each state by 1 year and by adding newborns to the population in state 0 (i). The effects of a wide variety of cancer control programs can be introduced into this framework. For example, primary prevention affects the incidence rates of cancers (a), screening and early diagnosis affect the proportions of cancers discovered in each state (b) [and might also affect the relative mortality rates of cancers labeled as being in a particular stage (0], and treatment affects the stage-specific relative mortality rates (f). One can estimate the effect of a cancer control activity by designating its effects on inci— dence rates, stage proportions, and/ or relative mortality rates; executing the model; and comparing the outcomes with those obtained in the absence of the program. The framework can calculate several outcome measures. For example, the program can estimate the 1) annual num— ber of new cases by calculating the number of people who move from state 0 into one of the five cancer states each year; 2) prevalence of cancers by calculating the number of people in states 1—5 at any time (both these outcome mea- sures can be calculated by sex, 5-year age group, and stage); and 3) annual number of cancer deaths (by sex and 5-year age group) by calculating the number of people who move from any of the five cancer states to state 6 each year. Incidence, prevalence, and mortality rates (crude and age-adjusted) can be obtained readily when the number of new cases, the number of prevalent cases, and the number of deaths are divided by the appropriate population—at— risk. The framework is sufficiently flexible to be applied to the analysis of cancer control programs that have special NCI MONOGRAPHS, NUMBER 2, 1986 targets. For example, if a colostomy education program is designed for a subset of people diagnosed with late-stage rectal cancer, the number of people to be served can be calculated from the number of new cases diagnosed in the late stages, and the proportion of people in each of those stages who receive colostomies. If a social service is de- signed to assist grieving families of patients who die of a cancer, the number of such families can be calculated. If a rehabilitation program is designed for patients over 65 with breast cancer, the number of such people can be cal- culated. If a proportion of patients with a late-stage cancer are to be given a particular diagnostic test (e.g., positron emission tomography), the number of examinations can be calculated. All these measures can be calculated for any year in the future, and discounting can be included. The framework can also be used to model the effect of cancer and cancer control activities on quality of life" Two types of factors that affect quality of life can be evaluated: factors that affect day-to-day quality of life (“state quality factors”), and factors that affect transitions between states (“transition quality factors”). State quality reflects the rela- tive desirability of various states that a person can be in, such as alive and free of cancer, diagnosed with cancer of a particular stage, under treatment with chemotherapy, dead of cancer, and dead of other causes. Transition quality reflects the desirability of various transitions a person might make between states, as caused by the following types of events: getting diagnosed as having cancer in a particular stage, receiving chemotherapy, dying of cancer without pain control, and so forth. By designating mea- sures of the quality of life associated with various state and transition quality factors, one can use the framework to calculate summary measures of the quality of life for the population as a whole or for individuals and to estimate how these measures are affected by cancer control pro- grams that affect quality of life. Finally, a variety of costs can be estimated. Given fixed and per-person costs for the initial and subsequent years of primary prevention programs, the per-person costs of screening programs, false-positive rates and workup costs of screening programs, the per-person cost of initial diag— nosis and treatment (by stage), the per-person cost of con- tinuing care (by stage), the per-person cost of terminal care, and the per-person cost of dying of other causes, the model can calculate the economic impact of a cancer con- trol program on a population for each of these compo— nents, as well as total cost, for any year. Inflation and discounting can be included. MATHEMATICAL FORMULATION The model consists of 20 Markov chains (one for each 5-year age group). Each Markov chain has eight states, as previously defined and illustrated in figure 1, plus a ninth state used to incorporate a mixed exponential function to model stage-specific survival rates (2). The model per- forms a sequence of calculations to simulate the passage of 6Although the framework has the potential to model quality of life, this feature is not available in the current computer soft- ware for CAN‘TROL (May I, I986). CANCER CONTROL OBJECTIVES FOR THE NATION: I985 2000 time in l-year intervals (e.g., 1986, 1987, . . .). At the be— ginning of any analysis, an initial state vector (No) records the number of people in each state. For each year, the model calculates the elements of a state transition proba- bility matrix (P0). These elements are the probabilities (Pij) that an individual in any one state (e.g., state i) will move to any other state (e.g., state j) during the next year. For each year, the state transition probability matrix is calcu- lated from 1) data on age-specific incidence rates, mortality rates, stage proportions, and stage-specific survival rates; 2) the estimated effect of the cancer control activities on incidence, stage proportions, and/ or survival rates; and 3) descriptions of the specific activities (e.g., subpopula— tions, risk factors, periods, age groups, proportion of people served, delay in effect of a primary prevention activity). The number of people in each state at the be- ginning of the next year (e.g., N.) can then be calculated by matrix multiplication, N. = N0 * P0 (where * denotes matrix multiplication). For each chronological year, the model performs these calculations for each 5—year age group. However, the model keeps track of the number of individuals in the population by 1-year age groups and uses this informa- tion, plus information on the expected number of births each year, to age the population 1 year between calcula- tions. The mathematics for calculating the state transition probability matrix for each chronological year (in a way that incorporates any combination of specific activities and a virtually limitless variety of subpopulations, periods, age groups, risk factors, proportions of people served, and other factors) is laborious, but it involves no new concepts. The mathematics used by CAN*TROL are described in more detail by Levin and his colleagues (2). Levin also describes the mathematics of some features that have been added to those in CAN*TROL to create the NCI model. The most important differences between CAN*TROL and the NCI model are that the latter model performs calcula- tions by 1-year age groups and also includes a Weibull function for modeling survival (in addition to the mixed exponential model used in CAN*TROL). The Weibull function provides a more accurate model for the survival rates for some cancers and requires more complicated com- putations. On the other hand, the mixed exponential model provides an equally or more accurate fit for most cancers, provides a reasonably close lit for virtually all can— cers, and enables CAN*TROL to be used on a personal computer. CAPABILITIES OF THE COMPUTER-BASED MODEL CAN*TROL The basic framework, described in figure 1, forms the basis for the computer model CAN*TROL which incor- porates a number of special features that allow planners to estimate the impact of cancer control programs in realistic settings. For example, because the populations are speci— fied by age, sex, incidence, and mortality distributions, virtually any population can be described; examples are the entire populations of France, Australia, or the Soviet Union; people over 20 in Shanghai County, China; His- panic Americans living in Los Angeles; or the employees of a particular corporation. 77 In addition, the population can be partitioned into any number of subpopulations, which may be the targets of different cancer control programs. (When CAN*TROL is used on a personal computer with 512K of random access memory, there is a practical limit of 5 subpopulations.) Subpopulations can be distinguished from the total popu- lation by differences in size, relative risk, stage propor- tions, stage-specific survival rates, and costs. Examples of problems for which this feature can be used include the analyses of programs aimed at different geographic regions or populations that have access to different levels of care; primary prevention programs for people exposed to par- ticular carcinogens (e.g., smokers, asbestos workers, sun- bathers); or groups of people who are expected to respond differently to a cancer control activity (e.g., different adher- ence rates). This version of the model can define a population con- sisting of a specific age group such as the Medicare popu- lation. Future versions will also define the population as a cohort of specific people who will be followed through time. A cohort is distinguished by the fact that once de- fined, no new individuals will enter the cohort. Examples are a specific group of people offered screening between 1986 and 1990 or the audience of a lecture on cancer pre- vention. This model can incorporate any cancer for which there are data on incidence rates, stage proportions, and stage- specific survival rates. Currently, for the United States population, CAN*TROL contains data on the same 40 cancer sites included in the NCI cancer control model (3). For any particular cancer, CAN*TROL can calculate si- multaneously the effects of any number and combination of activities. For any particular cancer control activity and any sub— population, the user can designate specific times that each activity will be conducted (e.g., 1987 through 1993). Any number of periods can be defined and, for any activity, subpopulation, and period, one can designate specific age groups that will be the target of the activity (e.g., all wo— men between 40 and 65). Any number of age groups can be defined, provided they encompass 5-year age groups. This feature is useful if an activity is to be restricted to a specific age group, or if the effectiveness or cost of an activity or the compliance of a population to an activity is different for different age groups. Within any activity, subpopu- lation, period, and age group, specification of the propor- tion of the people in a target group that will actually re- ceive the cancer control activity is possible. The program has been designed to analyze separately deaths from cardiovascular disease and deaths from non— cardiovascular diseases, to enable investigation of the im- pact of trends in cardiovascular mortality rates on cancer morbidity, mortality, or costs. Also possible is specifica- tion of future trends in cardiovascular mortality, mortality from other causes, incidence rates for any cancers, births, and the effectiveness of treatment. Given historical data, CAN*TROL can calculate functions that describe future trends. For a primary prevention activity, the user can identify a particular risk behavior or exposure to a carcinogen, cal- culate the relative risk of people in a specific risk group, calculate the maximum potential change (e. g., reduction) in risk that would occur if the risk behavior or exposure were 78 changed (e.g., eliminated), specify the proportion of that maximum change that is actually expected to occur if ex- posure is changed, and specify the delay in expression of the change. The program can also calculate how the activ- ity will change the prevalence of risk factors, i.e., the num- ber of people with the risk behavior or who are exposed to the carcinogen (e.g., the number of smokers). The program can also deal with cancer control activities that affect more than one cancer and calculates the com- ' bined effects of the activities across the various cancer sites according to changes in incidence, prevalence, mor- tality, prevalence of risk factors, quality of life, and costs. DATA SOURCES FOR CAN‘TROL This computer-based model requires information on the age and sex distribution of the population to be analyzed, the projected number of births over the period for which the effect of a program is to be calculated, age- and sex— specific incidence rates for the cancers to be analyzed, age- and sex-specific mortality rates from all causes, and age- and sex-specific mortality rates for the cancers to be analyzed. Levin et al. (2) describe the sources used to ana- lyze cancer control programs in the United States and the methods used to calculate the necessary data. To calculate the impact of a cancer control program on incidence, prevalence, mortality, or costs in a population, the user of CAN‘TROL also requires estimates of the im- pact of the specific cancer control activities on their im- mediate targets, such as the risk of developing a cancer, the stages in which cancers are detected, and stage-specific mortality rates. These numbers must be estimated from published research and, if necessary, expert judgments. The conclusions of four committees to evaluate the effectiveness of primary prevention, screening, treatment, and surveil— lance activities are incorporated in (3). USE OF CAN*TROL Software and a manual for CAN*TROL will soon be made available to people interested in planning cancer con- trol programs at the national, state, and local levels. The software runs on an IBM-PC, XT, or AT (or compatible PC) and performs six main functions. The first function permits a planner to define a popu— lation for which he or she wants to control cancer. Any population can be defined, and a planner can de- fine as many different populations as desired. The de- scriptions of a population can be stored for retrieval and use at any time. The second permits a planner to define the cancer(s) to be controlled in a population (e.g., incidence rates, stage proportions, stage-specific survival rates, and costs). These descriptions can also be stored and re- trieved at any time. The third function enables a planner to define the col- lection of activities to be implemented to control the cancers in the population. The activities are classified in the three general categories of primary prevention, screening, and treatment. Treatment activities include NCl MONOGRAPHS, NUMBER 2, 1986 those to improve quality of life. Any number of activ- ities can be defined and in any combination. The activ- ities can affect different cancers, subpopulations, and age groups, and can take place at different times. For convenience, CAN‘TROL enables a planner to define a “batch” of programs at one sitting. The fourth function displays or prints, or both, de- scriptions of the cancer control programs. In the fifth, CAN*TROL calculates the effects of the cancer control programs on the population, estimating their impact on incidence, prevalence, mortality, prev- alence of risk factors, quality of life,6 and costs. If a batch of programs has been defined, CAN*TROL will calculate results for the entire batch at one time (e.g., overnight). The last function allows a planner to display the re- sults, either graphically or in tables. BREAST CANCER SCREENING FOR WOMEN OVER AGE 65 The following example illustrates the use of CAN*- TROL. At the request of the staff of the Committee on Ways and Means, Subcommittee on Health of the House of Representatives, an analysis was conducted of the effec- tiveness and cost of annual mammography and breast physical examinations for women covered by Medicare (i.e., women over age 65). Breast cancer is the most com— mon cancer in females, and about 123,000 cases are ex- pected in 1986. The lifetime probability that a woman will develop the disease is about 10%. About 45% of breast cancers occur in women over 65, and the probability a woman in that age group ever will develop breast cancer is about 4.2%. Population and Cancer Data Sources For this analysis, the population was defined to be all women in the United States over age 65, and all results will pertain to this population. For an accurate calculation of the impact of breast cancer and breast cancer screening in all women over age 65 in future years, data are also re- quired for the number of women currently in all age groups [as obtained from the Census Bureau (4)] because women currently under age 65 will gradually enter the over-65 age group, and CAN*TROL needs this information to make the appropriate calculations. Age-specific incidence rates were obtained from the Surveillance, Epidemiology, and End Results Program (5), and age-specific mortality rates were obtained from the National Center for Health Statis— tics. Stage proportions and stage-specific survival rates were estimated from the Surveillance, Epidemiology, and End Results Program data (unpublished). See Levin et al. (2) for a discussion of factors affecting the estimation of stage proportions and stage-specific survival rates. Financial Costs Current charges for mammography range approximately from $25 to $200 (6). However, the low charges tend to be found only in special programs in which extraordinary ef- forts have been taken to contain costs. For this analysis, a CANCER CONTROL OBJECTIVES FOR THE NATION: [985 2000 charge of $60 will be assumed, and the impact of other charges will be discussed. Screening can lead to otherwise unnecessary biopsies of noncancerous lesions. For example, in the BCDDP, the proportion of women over age 65 who had biopsies that were positive for cancer was 1 of 4, compared with about 1 of 2 in unscreened women (6). At any examination, wo- men over age 65 have about a 1% chance of having a biopsy for a noncancerous lesion. Charges stemming from such evaluations depend on the workup (e.g., needle biopsy vs. excisional biopsy), and range from about $50 to $1,000. For this analysis, a charge of $500 will be assumed. Varia- tions in these assumptions will be discussed. Screening can reduce some costs if cancers are found in earlier stages when treatment might be less expensive, and some terminal care costs can be saved if women do not die of breast cancer. However, these savings are diminished by the cost of treating persons who eventually die of other causes. For this analysis, the cost of treating cancers in various stages and terminal care costs were estimated from the Third National Cancer Survey, updated by the infla- tion rate of medical care costs from 1973 to 1986. This leads to assumptions that initial treatment of in situ, local, regional, and distant cancers costs $7,970, $8,219, $8,809, and $11,859, respectively, and terminal care costs $24,000. There is great uncertainty about these costs, but as will be discussed below, varying these assumptions over a reason— able range has virtually no effect on the estimated net cost of screening. For this example, net costs include costs of examination, workups, initial care, terminal care, and deaths from other causes. Effectiveness of Screenlng There are five main sources of evidence on the effective- ness of screening in reducing breast cancer mortality. In the late 19605, the HIP of Greater New York conducted a randomized controlled trial involving four successive an- nual examinations of women between the ages of 40 and 65 with mammograms and breast physical examinations (7). Despite the fact that only 65% of the women offered screening participated, and only about 50% received all the scheduled examinations, the trial showed a statistically significant reduction in mortality. After 5 years of follow- up, mortality was reduced 38%, with a sustained reduction of 28% after 10 years, and 22% after 14 years. The effect was greatest in women between 50 and 65 years; the reduc- tion was about 40% at 10 years. Because the upper age limit for this study was 65, it provides no direct evidence about the value of mammography screening in women over 65 years of age. In the 19705, the NCI conducted the BCDDP to demon- strate the feasibility of breast cancer screening (8). In this program, 280,000 women between ages 35 and 74 were screened for 5 years with mammography and breast physi- cal examinations. The BCDDP was not controlled and mortality was not observed, but it provides information on the proportion of cancers found by various modalities and on the proportion of cancers found in various stages (8) and enables investigators to make comparisons among dif- ferent age groups. Specifically, the proportion of cancers found by mammography and the proportion of minimal cancers were about the same for women over age 65 as for 79 women between the ages of 50 and 65, implying that, by these outcome measures at least, breast cancer screening is equally effective in women over age 65. A randomized controlled trial involving two counties (only the communities were randomized, not individuals) is in progress in Sweden. This trial differs from the HIP study in several ways: Modern mammography equipment is used; breast physical examinations are not included; single-view instead of double-view mammography is per- formed; and screening is offered every 2 or 3 years, instead of every year. After 7 years of follow-up, a 31% reduction in the probability of death was observed in the screened group (9). As in the HIP study, the evidence of effective— ness appears limited to women over the age of 50 (a 40% reduction was observed). Although the trial includes wo— men over age 65, the published reports do not present the results separately for this age group. Women aged 50—64 years were screened with mammog- raphy and physical examinations four times in a 2-year period in a case—control study conducted in Holland. The results indicate that screening reduced the risk of dying of breast cancer by about 30% (10). Whereas this study was limited to women under age 65, the effect of screening was observed to increase with age, which suggests that screening would be effective in older women. A case-control study has been reported from Nijmegen, Holland, where a group of 30,000 women over age 45 were screened with single-view mammography every 2 years. Early results from the study indicate that screening can reduce mortality approximately 50% (11,12). Results have not been reported by age. In summary, good evidence exists that mammography reduces mortality for women between the ages of 50 and 65, with the reduction in the range of 25% to 40%. Al- though little direct evidence attests to the effectiveness of mammography in women over age 65, indirect evidence from the BCDDP supports an assumption that mammog- raphy is approximately as effective in women over age 65 as it is in women between the ages of 50 and 65. For this analysis, new stage proportions were estimated to achieve an overall reduction in mortality in women over age 65 comparable to that observed in women ages 50 to 65. Spe- cifically, stage proportions were set to achieve a reduction in 5- and 10-year mortality of 32% and 28%, respectively, for screened women over age 65. The effects of other as- sumptions can also be calculated and will be discussed. Impact of Breast Cancer Screening on the Medicare Populatlon With the data and assumptions just listed, CAN*TROL can calculate the effects of breast cancer screening on the Medicare population up to any year (e.g., 2000). In 1986, in the absence of screening, about 47,400 new cases and about 19,550 deaths will occur from breast can- cer in women over age 65. By the year 2000, the number of new cases and deaths is projected to increase to about 53,100 and about 22,100, respectively. The impact of screening will depend on the age group for which screening is recommended and on the proportion of women in that age group who are actually screened. As an example, table 1 shows the effect on mortality and cost if screening is recommended for women ages 65 to 75 and 30% of them accept it each year.7 After 1 year of screening, ap- proximately 229 fewer deaths are anticipated, with the re- duction in number of deaths slowly growing to about 610 by 1990 and reaching about 719 by the year 2000. Screen- ing costs (including the cost of working up women with false-positive examinations) can be expected to be slightly over $178 million in 1986, growing to about $186.5 million in 2000. Some savings are achieved from lower treatment costs (from $11 to $16 million). However, these savings are outweighed by the costs of screening by a factor of more than 10, which leaves the net costs in the range of about $168 million each year. 7These results, and the results to be shown in table 2, do not include discounting. Interested readers can use the results of tables 1 and 2 and whatever rates they prefer for discounting future benefits and costs to calculate discounted measures. TABLE 1.—Effect of mortality and cost of breast cancer in women over 65 of a program to screen 30% of women aged 65 to 75 Deaths Changes in costs (in thousands of dollars) Year Cases‘1 Expected” Change" Screening Treatment Net 1986 47,408 19,548 —229 178,427 —8,214 170,210 1987 48,096 19,860 “385 179,430 —11,879 167,551 1988 48,712 20,141 —491 180,188 —14,196 165,992 1989 49,261 20,393 -552 180,728 - 15,315 165,143 1990 49,748 20,618 -610 181,087 - 16,322 164,765 1991 50,181 20,818 -643 181,316 -16,602 164,714 1992 50,568 20,996 —665 181,471 — 16,553 164,918 1993 50,917 21,155 —680 181,613 —16,273 165,340 1994 51,239 21,301 -690 181,800 -—15,828 165,972 1995 51,543 21,437 *696 182,090 -15,269 166,821 1996 51,839 21,566 —702 182,534 — 14,637 167,897 1997 52,136 21,692 —706 183,175 —13,963 169,212 1998 52,442 21,821 ‘710 184,049 —13,273 170,776 1999 52,767 21,953 —714 185,183 — 12,588 172,595 2000 53,1 16 22,095 —719 186,594 —11,925 174,669 " Values represent No. of new primary cases of breast cancer in all women over age 65. h Values are expected No. of deaths in all women over age 65 if no screening were done. " Change is given in the reduction in No. of deaths in women over age 65 if the screening program is conducted. 80 NC] MONOGRAPHS, NUMBER 2. 1986 TABLE 2,—Effect on mortality and cost of breast cancer in women over 65 for the year 1995” Program Deaths Changes in costs (in thousands of dollars) Age, yr Acceptance. % Cases Expected Change Screening Treatment Net 657100 100 51,543 21,437 ~4,504 1,177,458 —82,270 1,095,188 65- 100 30 51,543 21,437 — 1,351 353,237 -24,577 328,660 657100 10 51,543 21,437 —450 117,746 —8,227 109,519 65—80 100 51,543 21,437 —3,249 837,348 —67,079 770,269 65780 30 51,543 21,437 —975 251,204 —20,123 231,081 65—80 10 51,543 21,437 —325 83.735 —6.708 77,027 65775 100 51,543 21 ,437 —2,322 606,967 —50,897 556,070 65-75 30 51,543 21,437 —696 182,090 —15,269 166,821 65—75 10 51,543 21,437 —232 60,697 —5,090 55,607 " See footnotes a—c of table 1 for definitions of cases, expected No. of deaths, and changes. With CAN*TROL, the user can also calculate the num- ber of person—years of life saved in the population (and the total cost) over time; between 1986 and 2000, annual screening of 30% of women ages 65 to 75 will add 56,362 person-years of life to this population. The cost during this period will be $2.5 billion. [It is not appropriate for one to divide the cost by the number of person-years to obtain the cost of adding one person-year of life to the population because the money spent on screening in the later years (e.g., 1990—2000) will continue to provide bene- fit beyond the year 2000. However, CAN*TROL can be used to calculate the cost of adding a person-year of life to a population. If one assumes a screening cost of $60 and the other assumptions listed, screening women between 65 and 75 adds a person-year of life to this population at an average cost of approximately $26,000.] The reduction in number of deaths (the person-years of life saved) and the costs are strongly affected by the age group for which screening is recommended and the propor- tion of women in that age group who are actually screened. Table 2 shows the effect on the Medicare population of screening three age groups (65—75, 65—80, and 65—100) with three acceptance rates (10%, 30%, and 100%) in the year 1995. Obviously, the more limited the program, the lower the financial cost and the lower the reduction in mortality. However, for all the programs, the costs of decreasing the number of deaths by 1 are about $230,000 each year. Sensitivity Analysis All these estimates depend on the values assumed for various parameters. In addition to the age group offered screening and the proportion of women who accept screen- ing, the two most important factors are the cost of the mammogram and the expected effectiveness of screening. The impact of different assumptions about both of these factors is easily estimated from tables 1 and 2. Because the cost of the mammogram comprises more than 90% of the total costs, different assumptions about the cost of screen- ing will cause approximately proportional changes in the total costs to Medicare. For example, if mammography is twice as expensive as assumed here (i.e., $120 instead of $60), the net cost of each program will be approximately twice that shown in tables 1 and 2. With respect to the ef- fectiveness of mammography, different assumptions will again cause proportional changes in the expected reduction CANCER CONTROL OBJECTIVES FOR THE NATION: 19852000 in deaths each year. Tables 1 and 2 were calculated on the assumption that breast cancer screening reduces 10-year mortality by about 28%. If the actual reduction is one-half as great (i.e., 14%), about one-half as many deaths will be prevented each year. If the actual reduction is twice as great, about two times as many deaths will be prevented each year. The effects of variations in the cost of treatment, the false-positive rate of screening, and the cost of workups for false-positive test results are negligible compared with varying assumptions about the cost of the mammogram. The estimates for this analysis are based on the informa- tion available in the published literature. More accurate ones can be obtained from Medicare records for the actual number of people covered in each age group, and specific data on the actual number of cases of breast cancer in 1986, cost of breast cancer treatment, cost of dying of other causes, cost of breast biopsies, and cost of a breast screening examination. The researchers in Sweden and Holland might also provide age-specific data on the effec- tiveness of screening women over age 65. Concerning the two most important sources of uncer- tainty and variability, one, the average cost of mammog- raphy to women covered by Medicare, is highly variable (ranging from about $25 to more than $200), easily deter— mined, and is rated high in priority for research. There is less uncertainty about the effectiveness of breast cancer screening (with a range of 25% to 40%). Any group con— sidering conducting new research to estimate the effective- ness of mammography in decreasing mortality in women over age 65 would find it extremely expensive and time- consuming. This deserves a lower research priority. For this particular analysis (breast cancer screening), uncer- tainty about false-positive rates and treatment costs has relatively small impact on policy decisions and deserves a low research priority. Discussion of Breast Cancer Screening The decision by researchers and health planners to rec— ommend breast cancer screening for women over age 65 should not be considered in isolation but should be com- pared with the benefits and costs that could be obtained with other possible activities, with any limits on costs taken into account. Specifically, a policy to cover breast cancer screening with annual mammography and physical examinations in Medicare should be based on the benefits per cost (marginal return) to be gained from this activity, 81 compared with the marginal return that could be gained from other activities. The impact on marginal return of changing the choice of tests and frequency of screening (as well as the age for stopping screening) should be explored. Furthermore, the marginal return of breast cancer screen- ing should be compared with the marginal return of other health activities that could be offered this population. With CAN*TROL, one could estimate the cost-effective- ness of other cancer control activities, such as cervical cancer screening, colorectal cancer screening, the use of special diagnostic tests to stage cancers prior to treatment, breast cancer chemotherapy and radiation therapy, lung cancer treatment, and the surveillance of cancer patients after treatment. In turn, researchers of other diseases can compare their results with the estimated marginal return of technologies such as organ transplants, renal dialysis, or coronary angioplasty. CONCLUSIONS Planning of national and regional cancer control strate- gies is difficult. At present, cancer experts and planners rely on global subjective judgments to estimate the effec- tiveness and costs of different programs and to set priori- ties. The complexity of the problem makes this approach vulnerable to oversimplification and error. Computer mod- els can be powerful aids to planning, enabling decision makers to break a problem into parts for which data exist and then reconstructing the parts to estimate the effects of a virtually limitless variety of cancer control activities on a large number of important clinical and economic out- comes. Large data bases and statistical projections provide information about the population, births, and deaths. Data are also available on the incidence and staging of can- cers and stage-specific survival rates. From decades of re- search, we have information about the effects of a wide variety of cancer control activities on their immediate tar- gets. For example, we know how 1) a chemotherapy pro- tocol affects survival of patients with a particular cancer in a particular stage; 2) screening affects staging or mortality, or both; 3) a particular anti—tobacco education program affects the number of people who stop smoking; or 4) treat- ment affects the pain from cancer. In addition to this, prac- titioners, experts, and planners can make judgments about factors that affect the distribution, use, cost, and other ef— fects of specific cancer control activities. The purpose of the framework described in this paper and its embodiment in CAN*TROL and the NCI computer 82 model is to integrate all this information. With these models, a planner can define a population, the cancer(s) to be controlled, and a virtually limitless variety of cancer control activities. He or she can also estimate the effects of the activities on cancer risk factors, incidence, prevalence, mortality, quality of life, and costs. This information is es- sential in the designing of comprehensive and efficient can- cer control strategies. REFERENCES (1) EDDY DM: Setting priorities for cancer control programs. JNCI 762I87—l99, 1986. (2) LEVIN DL, GAIL MH, KESSLER LG, et al: A model for pro- jecting cancer incidence and mortality in the presence of prevention, screening, and treatment programs. NCI Monogr 2:83—93, 1986. (3) GREENWALD P, SONDlK EJ (eds): Cancer control objectives for the nation: 19854000. NCI Monogr 213—74, I986. (4) BUREAU OF THE CENSUS: General population characteris- tics. In I980 Census of Population: Characteristics of the Population, United States Summary, chap B, vol I, Part 1 (PC80—I-BI). Washington, DC: US. Govt Print Off, 1983. (5) YOUNG JL JR, PERCY CL, ASIRE AJ (eds): Surveillance, Epidemiology, and End Results Program: Incidence and Mortality Data: I973—77. Natl Cancer Inst Monogr 57:17 1082, 1981. (6) BAKER LH, CHIN TD, WAGNER KV: Progress in screening for early breast cancer. J Surg Oncol 30296402, 1985. (7) SHAPIRO S, VENET W, STRAX P, et al: Ten- to fourteen- year effect of screening on breast cancer mortality. JNCI 69:349—355, 1982. (8) BAKER LH: Breast Cancer Detection Demonstration Proj- ect: Five-year summary report. CA 32:194—225, 1982. (9) TABAR L, FAGERBERG CJ, GAD A, et al: Reduction in mortality from breast cancer after mass screening with mammography. Randomised trial from the Breast Cancer Screening Working Group of the Swedish National Board of Health and Welfare. Lancet 1:829—832, 1985. (10) COLLETTE HJ, DAY NE, ROMBACH JJ, et al: Evaluation of screening for breast cancer in a non-randomised study (the DOM project) by means of a caseicontrol study. Lancet 1212244226, 1984. (ll) VERBEEK AL, HENDRIKS JH, HOLLAND R, et al: Reduc- tion of breast cancer mortality through mass screening with modern mammography. First results of the Nij- megen Project, 1975—1981. Lancet 121222—1224, 1984. (12) VERBEEK AL, HENDRIKS JH, HOLLAND R, et al: Mammo- graphic screening and breast cancer mortality: Age-specific effects in Nijmegen Project, 1975*1982. Lancet I2865—866, I985. A Model for Projecting Cancer Incidence and Mortality in the Presence of Prevention, Screening, and Treatment Programs David L. Levin, "* Mitchell H. Gail, 1-2 Larry G. Kessler, 3 and David M. Eddy“ ABSTRACT—We present a model for projecting cancer in- cidence and mortality under various cancer control programs in the presence of competing causes of death other than cancer. To use the model, one must first develop baseline projections that allow for secular trends in cancer incidence and mortality rates and in mortality from causes of death unrelated to the cancer under study. The expected effects of prevention, screening, and treatment interventions are then measured against these baseline projections. The calculations are modular, so that improved knowledge about the effects of the interventions can be incor- porated without substantial modification to the basic recursive formulas. Although the individual projected rates may vary depending on modeling assumptions and changes to the input data, the projected effects of specific types of intervention, expressed as percent reduction in mortality rates, are quite robust and consistent within each intervention activity.—NCI Monogr 2:83-93, 1986. To assess progress toward the goal of reduced cancer mortality, the NCI has established a set of quantitative objectives with the assistance of three Working Groups composed of experts in cancer prevention, screening, and treatment (I). The NCI has recommended a number of specific interventions. Our purpose is to describe a mathe- matical model of cancer incidence and mortality that allows one to estimate the changes in future cancer mortal- ity from application of these measures to prevent cancer, to screen patients for early diagnosis, and to improve the treatment of patients with cancer. Described herein are the elements of the model, the data sources used, and the types of information required by the user to run an inter- active computer program, written in the Fortran pro- gramming language to run on the DEC-10 computer at the National Institutes of Health. ABBREVIATIONS: NCI=NationaI Cancer Institute; SEER=Sur— veillance. Epidemiology, and End Results (Program); HIP= Health Insurance Plan (New York). |Biometry Branch, Division of Cancer Prevention and Control, Na- tional Cancer Institute, National Institutes of Health, Department of Health and Human Services. 2 Present address: Biostatistics Branch, Division of Cancer Etiology, National Cancer Institute. 3Surveillance and Operations Research Branch, Division of Cancer Prevention and Control, National Cancer Institute. 4Center for Health Policy Research and Education, World Health Organization Collaborating Center for Research in Cancer Policy, Duke University, Durham, NC. ‘Address reprint requests r0: David L. Levin, M.D., Biometry Branch, Division of Cancer Prevention and Control, National Cancer Institute, Blair Building, Room SAOS, Bethesda, MD 20892-4200. The model described is a descendent of an earlier model developed by Eddy (2) for the Cancer Unit of the World Health Organization. This earlier work first incorporated the general conceptual framework used here as well as specific models to assess the effects of the recommenda- tions of each of the Working Groups. Another direct de- scendent of the earlier model, written in the APL pro- gramming language to run on a personal computer, is described elsewhere (3). THE MODEL Figure 1 depicts the broad outlines of our model. Per- sons are assigned to one of four states, WELL, DIAG- NOSED, CANCER DEATH, and OTHER DEATH. Dur— ing each annual cycle, they remain in a given state or shift to another state according to transition probabilities that may remain constant over time, vary according to secular trends, or be modified as a result of cancer control inter- ventions. The model in figure I is applied to individual sex—site combinations. Thus if we are studying colon cancer in males, the term “cancer” refers only to colon cancer; cancers of other sites are included in the “other causes of death” category. The states in the model are based on epidemiologic and statistical concepts, not on biologic disease entities. The earliest state is called the WELL state. Persons in this state may, in fact, have other diseases or may even have developed preclinical cancer, but they have not been diag- nosed as having cancer. The WELL state is more precisely defined as “alive without a diagnosis of cancer.” At the start of each annual cycle, new births are added to the WELL state. Persons diagnosed with cancer shift into the DIAG- NOSED state. The probability of transition from WELL to DIAGNOSED is a function of the cancer incidence hazard, which we denote by the symbol “."0 Cancer pre- vention programs are directed toward reducing the number of new cancer cases by lowering the incidence hazard. Once the total number of incident cases are calculated, they are immediately distributed among up to five stages of disease in proportions 1n, 1r2, . . . m, (where 2k) 7rk= 1) according to the stage distribution at the time of diagno- sis. Transitions among stages are not modeled because available survival data, measured from the date of diagno- sis, are tabulated according to the initial stage at diagno- sis. During any cycle of the model, persons beginning in the WELL state are allocated to the appropriate DIAG- NOSED state for the start of the next annual cycle. For some sites such as cervical carcinoma, stage I represents in 83 DIAGNOSED / CANCER Stage 1 DEATH 9 1r 4: Stage 2 ( fl Stage 3 ‘ \ )— \ \ 3:32: \ OTHER g ~ DEATH < FIGURE l.—Conceptual model of cancer incidence and mortality. situ cancer, a preinvasive stage. To run the computer pro- gram, we set the hazard of dying from in situ cancer at zero. The effect of screening is a change in the distribution 1r, . . . Tl’k into a new distribution with higher proportions in the earlier stages with more favorable survival. Persons in the DIAGNOSED state are subject to the risk of dying from their cancer, denoted by the symbol “H.” The effect of a treatment intervention is a reduction in this hazard, leading to prolonged survival. The number of persons entering the CANCER DEATH state each year is tabulated and used in reports of the annual number of cancer deaths and the annual cancer mortality rate. All living persons in the model are at risk of death from causes other than the cancer of interest. We use the sym- bol “M” to indicate the hazard of transition into the OTHER DEATH state. Note that transitions from WELL to CANCER DEATH take two cycles of the model, one to go from WELL to DIAGNOSED and the second to go from DIAGNOSED to CANCER DEATH. Although this leads to underesti— mates of the immediate cancer mortality, it does not sig- nificantly affect the long-term projections. This picture is oversimplified in several respects. First of all, the model is age specific; therefore, we divide the pop- ulation of these states and the transition rates separately into 100 one-year age grOups from less than 1 year of age, 1, 2, . . . 97, 98, 99+. Persons entering the last age group (99+) remain there until they die. Secondly, the model is year specific to account for the effects of starting and stopping interventions and for external secular trends in incidence and mortality. The computer program written to implement the model keeps track of 41 individual years from 1980 through 2020. Statistical data from 1980 are used as the initial values for the population distribution and the underlying incidence, mortality, and survival rates. Thirdly, the DIAGNOSED state is further subdivided according to duration of illness. This elaboration of the DIAGNOSED state permits us to use a survival hazard H which changes with duration of illness and calendar year. Although the conceptual model allows subdivision of each state in many different ways, in practice we only divide to match the available input data. Thus the WELL state is subdivided by age and calendar year to match annual age-specific population data, whereas the DIAG- NOSED state is subdivided by stage of disease, calendar year, and duration of illness to correspond with annual stage-specific survival data. Note that the DIAGNOSED state is not divided according to age because reliable sur- vival data broken down by both age and stage are not available. MATHEMATICAL FORMULATION Notation and definitions are presented in table 1. Throughout, the indices i, j, k, and I shall refer, respec- tively, to age, calendar year, initial cancer stage, and dura— tion of illness. The model is based on basic recursive equations which define how people move among the various states. We TABLE l.—Explanation of terminology” Term Definition WLIJ No. of WELL people (alive without diagnosed cancer) age i at the start of calendar yr j. DXU“ No. of DIAGNOSED patients, age i at the start of yr j, whose initial cancer diagnosis in stage k was made [yr ago. Incident cases occurring in yrj are tabulated in yr j+l as having cancer for 1:] yr. CD“- No. of deaths from the cancer under study in I yr among those who were age i at the start of yr j. OD”- No. of deaths from other causes, including cancer of other sites, in 1 yr among those who were age i at the start of yr j. ICIJ No. of incident cancers in I yr among those who were age i at the start of yr j. BJ No. of births in yr j, added to the population at the start of that year. C”- No. of cancer cases in age group i at the start of yrj that were assigned to the fictitious CURED state and are at risk only of non- cancer hazards in yr j. Incident cases occurring in yrj are assigned to the CURED state at the start of yrj+ I. 1ij Proportion of cancers incident in yrj diagnosed in stage k, PC}.K Proportion of cases diagnosed at stage k in yrj that can be assigned to the fictitious CURED state. 0,J Annual sex- and site-specific cancer incidence hazard in yrj for people age i. p” General mortality hazard. excluding death from the cancer under study. in yrj for people age i. Hm Pure hazard of dying from cancer in yr j, under the appropriate survival model, for a person whose cancer was diagnosed I yr ago in stage k. RSI-k, Relative I—yr survival rate in yrj for people with cancer which was diagnosed in stage k. Pl,j = exp{—(mj+0,j)}. Probability that a person who is well and age i at the start of yrj remains alive and cancer free for | yr. P2,,“ = epr—(pij+HJ-k,)}. Probability that a person who is age i at the start of yrj and who developed stage k cancer I yr ago survives to the start of yr j+l. II P3.J not die of other causes) for I yr. exp(—pij). Probability that a person age i at the start of yrj who is not subject to the risk of developing cancer remains alive (does " i = index for age: i = 0, I, 2, . . . 99;j = index for calendar year: j = I980, I981, , . . 2020; k = index for initial stage of cancer: k = I, 2, . . . 5; 1: index for No. of elapsed years since cancer diagnosis: I: l, 2, . . . 41. 84 NCI MONOGRAPHS, NUMBER 2. I986 assume constant hazards within each yearly interval, though hazards may change from year to year. We model transitions as competing events according to standard methodology (4). The risk of a transition if no other tran- sitions were possible is referred to as the “pure” hazard. The transition probabilities developed below are based on these pure hazards. As defined in table 1, the terms “P1, P2, and P3” refer to the probability of a person’s remain- ing in a given state during any annual cycle. The annual number of incident cancers is given by .. _ _. JL. _ .. 1C, — WLU #u + on (1 Pl”), 1) where WLij is the number of WELL persons subject to the transition probability for the specific age and year. These incident cases are then allocated into k stages according to rrjk in preparation for the next annual cycle. The number of persons who have been diagnosed [=1 year ago as of the start of year j +1 is Dxi+l,j+l,k_1=ICij'Trij- 2) For [+122, the number of persons in each subdivision of the DIAGNOSED state is Dxi+l,j+l,k,l+1 = Dxijkl' Pzijkls 3) the number of cancer deaths in yearj among those age i is H4 ..= ._ ._L. _ .. CD, f§{DX-m #11 + HM (1 PAM}. 4) and the number who die from other causes is _ “ii i- _ WL,»~——- l—P .- OD] , MUM“ ( 1,) + 2 z {oxr L-1—P2.- } 5) k I jkl flij+ijI ( jkl) Further details of these recursions, including equations one can use to prevent losing track of people over 99 years old and those with cancer for more than 41 years, are given in Addendum 1. SPECIFICATION OF INITIAL DATA These projections require specification of births (8,), cancer incidence hazards (0.1-), mortality hazards from other causes of death apart from the cancer(s) under study (flij), secular trends in these hazards, stage allocation pro- portions (n-jk), and survival hazards (Hm). Also, the num- bers of subjects in the initial population are required. The present section outlines the baseline estimates of these parameters useful for projections in the absence of specific interventions. We will then describe how baseline hazards may be modified to assess the likely effect of various intervention scenarios on future mortality rates. Blrths Rather than attempt to project trends in age-specific fer- tility rates and then model the resultant numbers of annual births, we obtained projections of the number of future births, 8,, for l980£j_<_2020 by sex from the Bureau of the Census (5,6) and used their “middle” series. Note that we CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 thus assume that the influence of modeled changes in cancer on future births is negligible. Cancer Incidence All cancer incidence data are derived from the SEER Program of the NCI. Data from cancer registries in Ha- waii, Connecticut, Iowa, New Mexico, Utah, Altanta, Detroit, Seattle-Puget Sound, and San Francisco-Oakland are included. The SEER incidence rates are used in the model as estimates of 0, the pure incidence hazard. Inci— dence rates are for first primary tumors in the years 1978—82, which are similar to those published for the years 1973—81 (7). For each sex—site category studied, data are given in 5-year age groups, 0—4, . . . 80e84, and 85 and over. Second (or later) primaries are not counted as inci- dent cancers, so that the survival data, which associate cancer deaths with the first diagnosed primary, are con- sistent with the incidence data. Mortality From Other Causes Total mortality data for 1980 were obtained from the data tapes of Vital Statistics collected by the National Center for Health Statistics. Sex-specific data are pro- vided for ages <1, 1—4, 5—9, . . . and subsequent 5-year intervals through age 95+. The last interval refers to all persons age 95 and over. To derive the underlying mortality rates from “other causes,” I‘m for each sex—site category, we also need initial sex-, site-specific cancer mortality data. These are also obtained from the data of the National Center for Health Statistics. For each individual age and year, we subtract the sex-, site-specific mortality from the total mortality to obtain estimates of an. Secular Trends The rates described above are used as initial baseline values for the year 1980. They can be held constant for allj years or allowed to change with time. One of the ways we model these secular changes is to use projections based on arithmetic or geometric series. For arithmetic series, one projects incidence rates from 0i.j+1 = 0U + F'Gmm, 6) where F is the annual linear fractional increase or decrease. Geometric series are likewise projected from Bi.j+l = 0ij'(I+F)a 7) where F is the annual rate of change. A review of recent literature on trends in cancer inci- dence (8,9) and examination of SEER incidence rates from 1973 through 1982 indicate significant trends in mel- anoma and cancers of the lung, colon, stomach, prostate, cervix, and uterus. For each of these sites, we determined the constant F in equation 6 or 7 by fitting both equations to reported age-adjusted rates for the years 1973—81 and selecting the equation which best fit the data. Then we used these constants with the baseline 1980 data to pro- duce incidence hazards through the year 2020. The same constant F was used for all i ages. Whereas a linear series was used for projection of trends in increasing incidence for melanoma and cancers of the colon and prostate, a 85 geometric series was used in decreasing incidence for cancers of the cervix and uterus. One can also include secular trends by producing a table of age- and year-specific rates with other methods and then using the table in the calculations. This technique was used for lung cancer, which exhibits complex incidence patterns as a result of many component factors, including changes in cigarette composition and smoking habits (10—13). These trends cannot be expressed in the simple mathematical terms of equations 6 and 7. To model these complex trends, NCI investigators estimated a pair of sex- specific, age—period—cohort models based on long-term lung cancer mortality data (Brown CC, unpublished data). Because lung cancer survival is low, we used current ratios of mortality to incidence to convert these mortality projec— tions into incidence projections. Trends in non-cancer mortality rates may also be speci- fied either in arbitrary arrays hi] or in an arithmetic or geometric series. The total age-adjusted mortality rate for cardiovascular disease in the United States has declined from 425.6 per 100,000 persons in 1950 to 238.9 per 100,000 in 1982, an average decrease of 1.4% per year (14). To incorporate these changes in the model, we partition the total mortality into two components, labeled cardio- vascular and noncardiovascular. A trend can be entered for either component or for the combined total with the use of equations 6 and 7. Once the trend data are pro- cessed, the two components are added to get age- and year-specific total mortality. Then, as previously described, by substracting the sex-, site-specific cancer mortality fig- ures, one derives the individual in), the mortality rates from “other causes.” Stage Distribution The stage distribution at diagnosis is obtained from 1978 to 1982 SEER Program data. The definitions of stage used by the SEER Program are consistent with the Manualfor Staging of Cancer published by the American Joint Committee on Cancer (15). When clinical data sub- mitted to SEER are insufficient to determine stage, an “unknown stage” code is assigned. Usually, the proportion with unknown stage is less than 5%. For purposes of pro- jecting overall population mortality, we reallocated pa- tients with unknown stage into known stages of disease by an algorithm that preserves overall 5-year relative survival and maintains roughly the same distribution among known stages (Kessler LG, Parker R, Ries LG, unpublished manu— script). The resulting distribution is used for all rrjk unless modified by screening intervention. Cancer Survival Hazard Specification of the survival hazard depends on the spe- cific cancer survival model used. To estimate the survival model parameters, we would ideally use cause-specific survival data, but they are not readily available. Instead, we use the relative survival rate as an estimate of the pure cancer hazard (16). Relative survival is a measure of mor- tality among a specific group compared with that of the general population. Flattening of the relative survival curve indicates that patients with the cancer have attained the same mortality rate as the general population. We consider the following “pure” survival models (4) to de- 86 scribe the probability of surviving with cancer in the absence of other causes of death. Mixed Exponentlal Model R811]: PCjk + (l—PCjk)C_()‘jkl) . 8) This model is useful when the survival curve approaches a horizontal asymptote. For stage k cases diagnosed in year j, PCjk is the level of the asymptote to the survival curve. One interpretation of this model is the assignment of a percentage of the total cases to a fictitious CURED state with zero hazard of death due to cancer. Then PCjk represents the percent cured, M, is the constant pure hazard for the “noncured" part of the population (also for stage k cases diagnosed in year j), and l is the number of years elapsed since diagnosis. The corresponding hazard function for the noncured proportion of cases is then ij1 = Ajk . 9) Although it is obvious that at the time of diagnosis one cannot actually determine which patients are cured, intro- duction of a CURED state is a useful conceptual proce- dure that greatly simplifies the mathematical calculations. This model has been used for many years (17). Note that under the model the hazard does not depend on I, the time since diagnosis. Also note that by assuming that the cancer specific death hazard m, is independent of age i, we are ignoring any age~stage relationship; this assump- tion is necessary for use of the available stage-specific sur- vival data. Use of the fictitious CU RED state requires minor adjust- ments to the recursion formulas. The number of persons entering the DIAGNOSED state, as given in equation 2, is multiplied by the factor (l—PCJ-k). The number of persons in the CURED state remaining there for another year is represented by P3; thus the number of persons in the CURED state with any stage of cancer includes both per- sons remaining there from the previous cycle and those being newly assigned as a proportion of the new incident cases. The equation for the number of CU RED persons at the start of any cycle then is given by Ci+1yj+|= Cij'P3ij ‘i‘ E. {ICij'fl'jk'PCjk}. 10) The formula for the number of non—cancer deaths, equa- tion 5, has the additional term +Cij-(1— P3”). Welbull Model stk, = e‘WW‘W’, 11) where M, and 7“. are the scale and shape parameters for a stage k case diagnosed in year j. The corresponding sur- vival hazard is then ijI = Mk ‘ 'ij ‘ (1_0~5)(71k_ U, 12) where the term “0.5” is used to approximate mid-year hazard rates. Under the Weibull model, patients are never cured of their cancer, but the hazard of cancer death may become arbitrarily small. This model is appropriate when there is no horizontal asymptote to the survival curve. NCl MONOGRAPHS, NUMBER 2, I986 Our approach was to fit these two models to 2- and 5-year relative survival data for each sex—site combination and each stage of disease at diagnosis and to choose the model that provides the better overall fit to the data. We fit the early 2- and 5-year data because the heaviest cancer mortality is concentrated in the first 5 years for most cancers and because the most current data do not in- clude long-term survival figures. Details for estimating the parameters PCjk and M, for the mixed exponential model and M, and 35;, for the Weibull model from historical data on relative survival are presented in Addendum 11. Figure 2 shows the historical relative survival data and the fitted curves for two representative sites examined. For buccal cavity cancer in women, we selected the Weibull model; for colon cancer in men, we chose the mixed expo- nential model. Starting Populatlon We obtained the total United States population by sex for individual years of age from the 1980 decennial census (18) to establish the initial WELL population, WLi‘lggo. If one starts the calculations naively assuming that all per- sons are “well” in 1980 and that initially all DXW=0, then estimates of cancer deaths will be too small in the early 19808. However, after 20 years of propagating cancer inci- dence and deaths, the proportions with prevalent disease in the DIAGNOSED states DXW stabilize. To obtain more realistic starting population figures, we use the naive assumptions and generate the proportions in the DXW state for j=2000. We then apply these proportions to the 1980 census figures to obtain the absolute numbers in the various DXW states initially (j=l980) and subtract the corresponding numbers from the census data to produce modified values for each WLnggo. These procedures lead to more realistic estimates of cancer deaths in the early 19805 and otherwise have little effect on longer term projections. No reliable estimates of cancer prevalence in the nation are available, although a prevalence study by Feldman et al. (submitted for publication) is now in progress, and data on prevalence in Finland have been published (19). The initial prevalence figures obtained above are generally consistent with the preliminary prevalence data of Feld- man and co-workers. MODELS OF INTERVENTION The programming philosophy of expressing cancer inci- dence, prevalence, and deaths as functions of general transition rates based on hazards 6,}, W, and Hi“, and dis- tribution 1ij permits one to write separate program modules to define initial values of the basic parameters 0, p, H, and 1r, and then modify them with intervention modules for prevention, screening, and treatment. These modules may then be refined as new information becomes available. Preventlon The following model of cancer prevention calculates a fraction Fij, where OSFUSI, which reflects the effective- ness of prevention on age group i in yearj. The prevention program reduces the incidence hazard from the baseline value 6,,- to the new value FiJ-Bij. In the following discus- sion, we suppress the indices i and j. We suppose that a CANCER CONTROL OBJECTIVES FOR THE NATION: I98572000 100 | Buccal Cavity - Female —\ \ 804 \ S3: —1 ._: < 2 60~ in ‘\o~\ En) —< \\\\ LL! “‘~‘ 2 40—1 “\“ r— T‘ < d — o: .0. Reported Survival 20_ --Weibull — Mixed Exponential .4 Olllllllllllllllllll 0 5 10 15 20 SURVIVAL TIME (YEARS) 100 Stage] Colon-Male 80 33 ._l < 2 60 > «z D m E 40 g. < ._l Lu 9:. 20 ‘-_'_____ Olllllllllllllllllll SURVIVAL TIME (YEARS) FIGURE 2.—Rep0rted relative survival compared with the fit from two alternative survival models. proportion of the age- and year-specific population is exposed to a risk factor, e.g., smoking, and that the pre- vention program serves a proportion of this exposed pop- ulation (table 2). Then P(E) = the proportion of the total population exposed to the risk factor, P(S) = the proportion served by the prevention pro- gram, P(ES) = the proportion both exposed and served, P(E)_ = the proportion unexposed, and P(ES) = the proportion exposed but not served. In the absence of prevention, one observes a baseline incidence rate 0 = lo{P(E) + R-P(E)}, 13) where 10 is the incidence rate for an unexposed person and R is the relative risk for an exposed individual. At max- imal effectiveness, the prevention program reduces R to 87 TABLE 2.—Rela!ive risks and proportions for subpopulations” defined by whether or not the person is exposed and whether or not the person is served by the prevention program Proportions” Risk‘ Served Served Yes No Yes No Yes P(ES) P(ES) P(E) Yes R“ R Exposed _ _ Exposed No P(ES) P(ES) P(E) No 1.0 1.0 P(S) P(S) " Separate tables of this type are needed for each age group i and calendar year j. b P(E) = proportion exposed; P(ES) = proportion both exposed and served; P(E) = proportion not exposed, etc, ’ R = relative risk in the absence of intervention; R“ = relative risk at full prevention program effectiveness. Risk of 1.0 indicates baseline risk for persons not exposed to the risk factor. R* for served, exposed individuals. Therefore, the reduced incidence rate is 0* = F*0 = Io{P(E) + R- P(ES) + R*-P(ES)}. 14) The maximal prevention effectiveness factor, F*, is then calculated from equations 13 and 14. Note that 10 divides out of the calculation of F* and therefore does not need to be specified. The population served by the prevention program may require several years to achieve maximal effect. The com- puter program user may specify a phasein period during which F decreases linearly from 1.0 to F*. Likewise, fol- lowing cessation of the prevention program, the effective- ness of prevention may be allowed to persist at the maxi- mal level F*, or the user may specify a phaseout period during which F increases linearly from F* to 1.0. Another choice specifies how the intervention is actually conducted. In one case, a cohort such as all persons age 50—60 in 1986 can be identified and followed for a speci- fied period. That single cohort moves through time and is subject to intervention, but new people are not added to the study group. An alternative method is for the age group under study to be identified and new people added to the cohort each year as they reach the starting age. The current computer program module allows either type of intervention. The user must specify the parameters in equations 13 and 14, the age group affected, the years the intervention is in effect, phasein and phaseout periods, and whether new persons are added to the cohort. The required probabilities in equations 13 and 14 are specified by the user as P(E), P(S), and P(EIS)=the proportion of the served popula- tion with the risk factor. Then the program calculates P(ES)=P(S)' P(EIS), P(ES)=P(E)—P(ES), and P(E)=l- P(E). The relative risk R may be supplied directly or expressed indirectly as a function of P(ElCancer)=the proportion of cancer patients with the risk factor with the formula : P(CancerIE) : P(EICancer) 4 P(E) __ __ . _. 15 P(CancerlE) P(EICancer) P(E) ) 88 The user may specify R“ directly or indirectly by specify- ing X, the overall percent reduction in incidence among the served population. Then R* can be calculated from _ (R—R*)-P(ES) : (R—R*)- P(EIS) P(ES) + R' P(ES) P(EIS) + R- P(EIS). 16) Screenlng We model the effect of screening as altering the distribu- tion of initial stage of disease at diagnosis. We assume that a maximally effective screening program causes the stage distribution for screened individuals to change from 1m to 113:, where the latter distribution assigns greater probabili- ties to early stages and applies uniformly to the entire screened population. As with the original distribution, 2 1n? = 1. In the following discussion of screening, we k again suppress year index j. To allow for screening of part of the population, we partition the population as in table 2, except now S indi- cates that a subject is screened, rather than served by the prevention program. Such partial screening of a popula- tion might arise in an investigator’s effort to screen only high-risk subjects or in a screening program with limited resources. In the absence of prevention, the total number of cancers is given by equation 13. At maximal screening effectiveness, the overall proportion of the total cases diagnosed in stage k, denoted here as 1r}, is therefore given by T _ mmfiHR- P(ES)} + 1rf[P(ES)+R- P(ES)} 17) 1” P(E) + R- P(E) ' Note that if no one is screened, NIT-m; if everyone is screened, ”1:111“. Equation 17 may be rewritten as Wi=7rk +p'P(S)'(1n?‘—7Ti), 18) where : P(EIS) + R‘P(EIS) _ 19) P(E) + R~ P(E) is the relative risk of cancer in the screened group com- pared with the general population. To allow for phasing the screening program in, one specifies a period over which each 7r: varies linearly from its initial value, m, to the value given by equation 18. This technique is equiva— lent to an assumption that 1r;k in 18 “phases-in” linearly from an initial value m to the value at maximal effective— ness, mi“. A similar phaseout period can also be specified. To use the screening algorithm, one specifies the years during which the screening intervention occurs, the age range affected, and the parameters of equation 18, i.e., m1", P(S) and p. The new stage distribution, mi", applies to per- sons actually screened. The probability of being screened, P(S), may be specified directly or as the product of three factors, i.e., the proportion in the age range offered the screening program, the proportion of those who accept the offer, and the probability of compliance, given acceptance. The relative risk of the screened compared with the total population, p, may be entered directly or calculated from equation 19 if one wishes by specifying R, P(E), and P(EIS). NCl MONOGRAPHS, NUMBER 2. 1986 Treatment We model treatment by changing the stage-specific pa- rameters of the survival hazard equations. In the following discussion, we now suppress the indicesj and k. The model assumes that a new treatment goes into effect at a given year, applies uniformly to all people in the affected stage, and applies to that group for the rest of their cancer experience. To calculate the new survival, one specifies a new relative survival value for a specific point on the sur- vival curve. For example, one may specify a new value for 5—year relative survival. In addition, one specifies whether the Weibull or the mixed exponential model is to be used and which parameter of that model will be changed from baseline to reflect the improved survival. For the Weibull model, for a given stage and year of diagnosis, equation 11 can be solved explicitly for the new parameter, either N" or 7*. If the scale parameter A is affected by treatment, one solves RS,‘=exp{—)\* - (17)} for A‘, using the specified quantities l and RS,* and the baseline shape parameter 7. Likewise, one can solve for 7* if it is assumed that the effect of new treatments is to alter -y. The effect of new treatment may be phased in by linear interpolation of 1-year survival from RS, to the maximal value, RS,*. Then N“ or 'y“ is calculated sepa- rately as above for each year of the phasein period. For the mixed exponential model. for a given stage and year of diagnosis, equation 8 can be expressed as RS,=PC+(1 -PC)exp(—)\l). Depending on which param- eter is to be changed, the new values M or PC“ are then calculated from RS,*=PC+(l—PC)exp(—)\*l) or RS,*= PC*+(l—PC*)exp(—)\I). The new treatment effect is phased in similar to the Weibull model. Multlple Interventlons The model allows interventions in prevention, screening, and treatment of the cancer under study. The computer program allows any or all of these to occur simultaneously. When using more than one type of intervention, we do not consider interactions between the interventions. For ex- ample, cancer prevention changes the number of cancer cases. Cancer screening affects the stage distribution of cases which occur. Both interventions require input data on the proportion of the population at risk, but we do not consider how prevention affects the proportion of the population served by the screening program. The user may make those calculations external to the program and mod- ify the input data accordingly. Screening also interacts with cancer survival. Although there may be a difference between the survival of stage 11 cases detected as a result of screening and those identified clinically, we do not automatically change the stage-spe- cific survival as a result of screening. Because we allow the treatment hazard to change as a result of treatment inter- vention at the same time as we allow screening to occur, the user must decide the net effect on survival and enter the appropriate data. PROGRAM OUTPUT The model tracks the number of persons in each of the four states given in figure l for individual ages and calendar years. In addition, the number of new cancer cases can be determined. These numbers can then be used CANCER CONTROL OBJECTIVES FOR THE NATION: 1985 2000 to generate output from the model. The current computer program records the number of incident cases, cancer deaths, and deaths from other causes by 5-year age groups for each year from 1980 through 2020. The population-at- risk of developing cancer or dying is also recorded. From these data, we obtain age-adjusted incidence and mortality rates by applying direct adjustment to the 1980 United States standard population. Alternative standard popula- tions may be specified. This output was used in prepara- tion of the NCI's cancer control objectives (1). Any other output based on the recorded data items can also be generated. The computer program is currently being used for research on cancer prevalence, cancer inci- dence and mortality in the elderly, and cancer experience in subpopulations of the United States. When sufficient input data are not readily available, they may be generated or estimated. BREAST CANCER AS AN EXAMPLE One of the few cancers for which prevention, screening, and treatment activities all have the potential to decrease mortality in this country is breast cancer. The recommen- dations of the three Working Groups in these areas of intervention were used as input to the computer program described above. Although a more detailed report on the development of the cancer control objectives is available (1), we review the input data necessary to evaluate the effect of suggested cancer control strategies. These recom- mendations are summarized in table 3. Reduction of dietary fat was recommended as a strategy for reducing breast cancer incidence and subsequent mor- tality. The target age group of women over the age of 45 was suggested in the development of the cancer control objectives; this group includes the bulk of incident breast cancer cases. At maximal effectiveness, we expect a 25% reduction in the age-specific incidence rates as dietary fat intake drops. This is the value of the variable X required TABLE liSummary of interventions/0r breast cancer mortality projections” Prevention Target population is all women aged 45 and above. New women are added as they reach age 45. Target reduction in incidence is 25%. Screening Target population is women aged 50770 yr. Proportion served is 63.4% Relative risk (screened vs. total) = 1.0. Stage distribution In situ l 11 III IV Old 0.046 0.128 0.549 0.201 0.076 New 0.150 0.600 0.150 0.100 0.000 Treatment Parameter of mixed exponential model changed is “percent cured." Five-yr survival In situ 1 II 111 IV Old 1.000 0.960 0.826 0.600 0.167 New 1.000 0.964 0.879 0.693 0.175 " All interventions begin in 1985 with 5-yr phasein period. 89 in equation 16. The time required to attain maximal effect is estimated at 5 years. We assume that the dietary mes- sage directed to 45-year-old women would begin in 1985, continue through the end of the century, and be transmit- ted to each new cohort entering the targeted age group. The HIP of New York study (20) has provided us the strongest evidence that screening practices can reduce long—term mortality from breast cancer. This study, several case—control studies (21,22), and a recently published trial in Sweden (23) provide a strong foundation for the devel- opment of screening objectives. Because stage shift data are unavailable from these stud- dies, the effect of screening on the stage distribution of incident breast cancer cases must be estimated. The shifted stage distribution shown in table 3 produces a 30% reduc- tion in mortality for women aged 50—70 who were fol- lowed for 15 years. Although other equally plausible alter- natives exist, this distribution was chosen by the Working Group as a plausible stage shift which matches the mortal- ity data in the literature (20). Several other input parame- ters were needed; these were derived from the HIP expe- rience. We assume that all women aged 50—70 are offered the screening package (breast physical examination and mammography), two-thirds avail themselves of the offer, and 95% of those with positive results adhere to a workup. Although there is some indication from the HIP study that the women who actually were screened have a higher relative risk of breast cancer than the unscreened popula- tion, the more conservative assumption of no excess risk was chosen. The Treatment Working Group concluded that some gains can be made in the 5-year survival rates for each of TABLF 4.~Projected breast cancer mortality in the year 2000 after multiple interventions compared with no intervention” TABLE 5.—Projected breast cancer mortality in the year 2000 after single intervention, compared with no intervention No. of deaths Age—adjusted rate‘1 No. of deaths Age-adjusted rate” Intervention Intervention Year No Yes Difference No Yes Difference. % 1980 38,027 38,027 0 30.0 30.0 0.0 1981 38,961 38,961 ” 30.2 30.2 ” 1982 39,868 39,868 " 30.5 30.5 ” 1983 40,751 40,751 " 30.7 30.7 ” 1984 41,611 41,611 ” 30.9 30.9 ” 1985 42,443 42,443 ” 31.1 31.1 " 1986 43,247 42,604 —643 3 I .2 30.7 — 1.6 1987 44,025 42,292 —1,733 31.4 30.1 —4.2 1988 44,785 41,644 —3,|4I 31.5 29.1 -7.4 1989 45,534 40,754 —4,780 3 I .6 28.1 — 1 1.2 1990 46.265 39,683 -6,582 31.7 26.9 —15.2 1991 46,981 38,478 -8,503 31.8 25.6 -I9.3 1992 47,679 37,577 —10,102 ” 24.6 —22.6 1993 48,368 36,891 —11,477 31.9 23.8 -25.4 1994 49,054 36,359 - 12,695 " 23.1 —27.7 1995 49,729 35,940 - 13,789 32.0 22.5 —29.7 1996 50,399 35,616 —14,783 ” 21.9 —3I.5 1997 51,063 35,372 —15,691 32.1 21.5 —33.0 1998 51,719 35,179 -l6,540 " 21.1 —34.3 1999 52,374 35,034 ~17,340 ” 20.7 —35.4 2000 53,023 34,935 — 18,088 ” 20.4 —36.5 ‘7 The multiple interventions are prevention, screening, and treatment. b Rate is per 100,000 persons, age—adjusted to 1980 United States population. 90 Inter- Intervention vention Type of _ —— intervention No Yes Difference No Yes Difference, % Prevention 53,023 45,473 —7,550 32.1 27.4 —l4.8 Screening ” 46,027 —6,996 " 27.0 — 16.0 Treatment " 45,465 -7,558 " 27.5 — 14.5 " See footnote b, table 4. four stages of breast cancer. The exact therapies recom- mended varied by stage, menopausal status, and estrogen receptivity of lymph nodes. Improvements in the observed survival rates from SEER were estimated from large-scale clinical trials, and these percentage improvements were translated to improvements in relative survival by roughly the same proportionate changes. We assume that the change in 5-year survival improves the probability of a cure from breast cancer, which subsequently reduces the deaths due to the cancer. The results of the model, shown in table 4, estimate a 36% reduction in breast cancer mortality by the year 2000 from all interventions combined. This would reduce the age-adjusted mortality rate from 32 to 20 per 100,000 women. Data for each intervention alone (table 5) suggest that each intervention has approximately the same pro- portionate effect, 15%, on the age—adjusted mortality rate projected for the year 2000. As might be expected, the activities are not additive: A case of cancer prevented is not amenable to improved therapy. In the case of breast cancer, it appears that each of these cancer control activi- ties has an important role to play in reducing mortality. DISCUSSION The health status of a population depends on actions taken by health professionals, institutions, and the public. The model presented by Eddy (2) and further amplified here offers a rational framework for utilizing available data on cancer status of a population and for estimating the impact of cancer control activities. Such a framework is helpful in the planning of research activities and in recommendations for cancer control. A feature we have stressed is the modular program design, which permits the incorporation of new algorithms to describe any of the interventions: screening, prevention, or treatment. Furthermore, general specification of base- line arrays for the incidence and mortality parameters and for the cancer survival data can be incorporated. A major benefit of constructing this model is that it forces those involved in cancer control decisionmaking to consider what data are available and what aspects require further data and research. To achieve the full potential of identifying which cancer control efforts might be expected to yield important public health benefits, we need to test a projection model for sensitivity to assumptions and to revise it to incorporate new knowledge about the effects of various interventions. Two concerns that might be ex- NCI MONOGRAPHS, NUMBER 2, I986 TABLE 6.*Comparison of program effects under two alternative survival models" No. of deaths Age-adjusted rate” Intervention Intervention Yes Difference No Yes Sex Site Model No Difference, % Male Lung, small cell Mixed exponential 18,556 14,114 —4,442 16.2 12.3 —24.0 Weibull 18,354 14,401 —3,953 16.0 12.5 —21.6 Male Colon Mixed exponential 25,837 11,838 — 13,999 21.7 9.6 —55.6 Weibull 28,046 13,458 —- 14,588 23.5 11.0 —53.4 Female Colon Mixed exponential 25,989 11,456 —14,533 14.5 6.4 —55.8 Weibull 28,748 13,423 —15,325 15.9 7.4 —53.4 Female Breast Mixed exponential 53,023 34,935 — 18,088 32.1 20.4 —36.5 Weibull 69,308 44,382 —24,926 42.4 26.2 —38.3 " Model projections are always for the year 2000, with projections beginning in 1980 and the cancer intervention scenarios as described in (1) starting in 1985. ” See footnote b, table 4. pressed are related to the input data for the model and the description of the cancer control scenarios. Alternate estimates of the effectiveness of prevention, screening, and treatment interventions translate directly into changes in the projected relative effectiveness of these activities. For example, in the scenario for preven- tion of breast cancer used in the text, an estimated inci- dence reduction of 15% rather than the suggested 25% results in cancer mortality reductions of only 8.9% rather than 14.8%. A treatment improvement one-half as large as that suggested by the Working Group results in one- half the mortality reduction from breast cancer, i.e., from a reduction of 14.5% to 7.1%. For any of the interven- tions, the length of the phasein period is also directly translated into changes in the program’s effectiveness. Limitations are also inherent in the algorithms used. The current screening algorithm is perhaps the least realis- tic of the intervention models because it does not allow for changes in the incidence rate caused by the introduction of a large screening program in an unscreened population, nor does it have any provision for including an estimate of lead time bias, rather than true screening effect. Both problems can be handled by the user with careful con- struction of the input data; however, the development of a more sophisticated model and corresponding computer program module is also feasible. The prevention model is perhaps the most realistic of the three models, incorporating many of the considera- tions important in planning cancer prevention programs. For example, what proportion of the population has a certain risk factor and how much of the population can the program reach? Unfortunately, as in screening, devel- opment of sophisticated models is limited by the availa- bility of reliable population data. The heavy data re- quirements are not easily met, and it is difficult for investigators to reach a consensus opinion on reasonable input parameters. The severest limitations of the treatment component are its simplicity and, again, the lack of population data to be used as input. Estimating future relative survival rates for a specific target population is difficult. Data from a study or clinical trial in one group may not be appropriate when applied to another. In addition, it is difficult for anyone to estimate the effects of changes in Medicaid or other parts of the medical care system that influence cancer patient CANCER CONTROL OBJECTIVES FOR THE NATION: 198572000 survival even in the absence of clear changes in cancer treatment. Although the model is expected to be sensitive to changes in the intervention parameters, it should be robust to changes in the baseline data. The sensitivity of the model‘s projections to possible errors in the stage-specific survival data and the choice of survival model were sys- tematically evaluated. Sensitivity analyses were conducted for cancers for which patients have relatively poor survival (small cell lung cancer in males), intermediate survival (colon cancer in both sexes), and good survival (breast cancer in females). The choice of the survival model (table 6) has a marked effect on the absolute numbers of deaths projected and the age-adjusted cancer mortality rates, but the relative effects of specific interventions, expressed as percent difference of the rates, are not substantially per- turbed. This effect is most clearly seen in breast cancer; the difference between the two models is large in every column except that for the percent differences in the age- adjusted rate. Variation of the baseline incidence data or of the 2- or 5-year survival probability produced corresponding changes in long-term, age-adjusted mortality for the site being investigated but again did not affect the relative effect of the interventions. For the purpose of evaluating the relative effects of different interventions, the model is therefore robust. More work needs to be done on assess- ment of the sensitivity of the overall projections. The planning of cancer control programs is a complex task. Without a rational approach on the part of those involved in evaluating these programs, inefficient or waste- ful decisions may result, and opportunities to reduce can- cer morbidity and mortality may be lost. The model we presented here offers an approach to evaluating public health programs in cancer control as an aid to the difficult decisionmaking process. ADDENDUM 1. BASIC RECURSIONS FOR THE NONABSORBING STAGES, WL” AND DXW The definitions of the following parameters are given in table 1 and figure 1. Persons are not allowed to age beyond 99 years old. Thus a 99-year-old person who sur- vives 1 year remains in the 99+ age group in this model. Likewise, a person who developed cancer more than 41 91 years ago is still regarded as having incident cancer 41 years ago. The initial WELL population WLi_|930 is defined from the input data for all OSiS99. At the end of each annual cycle (198OSj52019), the numbers of WELL persons for the subsequent cycle are given by the following recursions: WLo.j+1 = j+l WLi+Lj+l : WLij ‘ Plij WL99J+I = Wng‘j ‘ Plgg'j + Wqu‘j ‘ Flog-j. [for 03397] As given in formulas l and 2 of the text, the numbers of new incident cases in each of the k subdivisions of the DIAGNOSED state, for I98OSjSZOI9, are governed by 0.. DXi ‘ =WLi-' -——-—U - I—Pli- ' - f 03.397, +l,j+l.k,l J I“) +0ij ( J) 1r“, [or 1 ] and oosj 99.j+|.k.| L98.) #98.) + 998,1" ( 98.)) m 099,’ +WL --—'— l-Pl -- -. 99'] I499,j + 099.j ( 9%) 771k For OSiS97 and lSl.<_39, the number of persons re— maining in the DIAGNOSED state is Dxi+l,j+l.k,l+l = DXijkI'PZijkI To accommodate “edge effects,” we have for OSiS97: Dxi+l,j+l,k_4l = Dxi,j_k,40 ' P2i,j_k,40 + Dxi,j,k,4l ' P2i.j,k.4l; for 131339: DX99.j+l.k,l+l = DX98.j.k.l' P298.j,k.l + DX99,j,k,l' P299,j,k./: and finally DX99.j+l.k.4l = DX98.j.k.-10 ‘ P298.j.k.40 + DX98.j.k.4l ' P29X,j.k,4l + Dx99.j.k.40 ' P299.,.k.4o + DX99.j.k.4l ' P299.j.k,4|~ These formulas complete the recursions needed for non- absorbing states. ADDENDUM II. FITTING PURE CANCER SURVIVAL MODELS TO TWO- AND FIVE-YEAR RELATIVE SURVIVAL DATA The mixed exponential and Weibull models for pure cancer survival were fitted to observed 2- and 5-year rela- tive survival rates. In the following, we consider two arbi- trary survival times from cancer diagnosis, t.