Effectiveness and Costs of Osteoporosis Screening and Hormone Replacement Therapy, Vol

Home , Benzestrol, Estrogen patch, Hormone therapy, Norethisterone acetate

Effectiveness and Costs of Osteoporosis Screening and Hormone Replacement Therapy, Vol. II: Evidence on Benefits, Risks, and Costs

August 1995 OTA-BP-H-144 GPO stock #052-003-01424-6 Foreword

enopause typically occurs in women around age 50. Accompany- ing this life event is a decline in estrogen levels and an increase in the rate of decline in women’s bone density. This rapid bone loss increases women’s subsequent risk of developing osteoporosis, a diseaseM characterized by low bone density and increased bone fragility. Among the most serious consequences of osteoporosis is fracture of the hip, which may result in substantial morbidity, prolonged hospitalization, and death. Estrogen can prevent bone loss after menopause by replacing the body’s own estrogen. Given the serious consequences of osteoporosis, some osteoporosis experts have recommended that women have their bone mineral density measured at the time of menopause and those with the lowest bone mineral density be offered hormone replacement therapy, comprising estrogen given alone or in combination with the hormone progestin. This background paper, Effectiveness and Costs of Osteoporosis Screen- ing and Hormone Replacement Therapy, assesses the medical benefits and costs of both screening and hormone replacement therapy. It is divided into two volumes. The first volume, Cost-Effectiveness Analysis, presents the results of a model that estimates the cost per year of life gained from osteoporosis screening and hormone replacement therapy in postmenopausal women. The second volume, Evidence on Benefits, Risks, and Costs, provides the basis for the assumptions about the costs and effects of screening and hormonal replacement therapy used in the cost-effectiveness model. This background paper is one of three documents resulting from OTA’s assessment of policy issues in the prevention and treatment of osteoporosis. This assessment was requested by the Senate Special Committee on Aging, Senator Charles Grassley and Senator John Glenn, and the House Select Committee on Aging, Representative Olympia J. Snowe, Representative Benjamin A. Gilman, and former Representatives Brian J. Donnelly, Thomas J. Downey, and Patricia F. Saiki. Two background papers in this series have been issued, both in July 1994: Public Information about Osteopo- rosis: What’s’ Available, What’s Needed?, and Hip Fracture Outcomes in People Age Fifty and Over.

ROGER C. HERDMAN Director

iii Preface

his volume, “Evidence on Benefits, Risks, and Costs of Hormonal Replacement Therapy,” is a companion to the volume “Cost-Effec- tiveness Analysis” of the OTA background paper “Effectiveness and Costs of Osteoporosis Screening and Hormone Replacement Therapy.”T This volume reviews evidence on the impact of hormonal replacement therapy (HRT) on bone density, fractures, breast cancer, endometrial cancer, gallbladder disease, and heart disease that underlies the assumptions used in OTA’s cost-effectiveness analysis. This volume also includes information about hormonal replacement therapy dosage regimens; reviews the relationship between bone mineral density and hip fracture; and summarizes the costs of bone mineral density screening, intervention, and diseases affected by HRT. This volume is organized as a series of appendices. Several appendices review HRT’s impacts on disease, including:

■ Appendix B: “Evidence on Hormonal Replacement Therapy and Fractures,” ■ Appendix F: “Evidence on Hormonal Replacement Therapy and Breast Cancer,” ■ Appendix G: “Evidence on HRT and Endometrial Cancer, ” ■ Appendix H: “Evidence on HRT and Gallbladder Disease,” and ■ Appendix I: “Evidence on HRT and Coronary Heart Disease.” Appendix D, “Summary of Hip Fracture Prediction Methods,” details the method for predicting the number of hip fractures used in OTA’s cost-effectiveness analysis. The appendix describes the specific parameter assumptions and sources of data regarding the longitudinal distribution of bone mass in menopausal women from ages 50 to 90. The appendix also describes the specific parameter assumptions and sources of data regarding the short-term relationship of bone mass to fractures at each age. Appendix E, “Hormonal Replacement Therapy Regimens,” describes the types of estrogens and progestins used for hormonal replacement therapy, their doses, and their administration. The appendix also describes the impact of hormonal replacement therapy on menopausal symptoms, and adverse effects of HRT, such as bleeding and premenstrual-tension-like symptoms. The appendix also describes the impact of these various dosage regimens on compliance with HRT. Appendix J, “Methods for Estimating Costs,” provides the basis for OTA’s assumptions concerning the costs of bone mineral density measurement, hormone replacement therapy, heart disease, hip fractures, gallbladder disease, endometrial cancer, and breast cancer.

v Advisory Panel

Robert P. Heaney Sylvia Hougland Robert Lindsay John A. Creighton Professor Dallas, Texas Chief, Internal Medicine Creighton University Helen Hayes Hospital West Haverstraw, New York Omaha, Nebraska Conrad C. Johnston Director Steven R. Cummings Division of Endocrinology & Betsy Love Research Director Metabolism Program Manager College of Medicine Indiana University School of Center for Metabolic Bone University of California Medicine Disorders San Francisco, California Indianapolis, Indiana Providence Medical Center Portland, Oregon Barbara L. Drinkwater Shiriki K. Kumanyika Research Physiologist Associate Director for Robert Marcus Pacific Medical Center Epidemiology Director Seattle, Washington Center for Biostatistics & Aging Study Unit Epidemiology Virginia Medical Center Deborah T. Gold College of Medicine Palo Alto, California Assistant Professor Pennsylvania State University Duke University Medical Center Hershey, Pennsylvania Lee Joseph Melton, Ill Durham, North Carolina Head, Section of Clinical Edward O. Lanphier, II Epidemiology Department of Health Sciences Susan L. Greenspan Executive Vice President for Research Director Commercial Development Mayo Clinic Osteoporosis Prevention and Somatix Therapy Corporation Rochester, Minnesota Treatment Center Alameda, California Beth Israel Hospital Boston, Massachusetts Donald R. Lee Gregory D. Miller Vice President Vice President Caren Marie Gundberg Procter and Gamble Nutrition Research/Technical Services Assistant Professor Pharmaceuticals National Dairy Council Department of Orthopedics Norwich, New York Rosemont, Illinois Yale University School of Medicine New Haven, Connecticut

Note: OTA appreciates and is grateful for the valuable assistance and thoughtful critiques provided by the panel members. The panel members do not, however, necessarily approve, disapprove, or endorse this report. OTA assumes full responsibility for the report and the accuracy of its contents. vi Morris Notelovitz Neil M. Resnick Mehrsheed Sinaki President and Medical Director Chief, Geriatrics Professor, Physical Medicine and Women’s Medical & Diagnostic Brigham and Women’s Hospital Rehabilitation Center & the Climacteric Clinic, Boston, Massachusetts Mayo Medical School Inc. Rochester, Minnesota Gainesville, Florida Gideon A. Rodan Executive Director Milton C. Weinstein William Arno Peck Department of Bone Biology Henry J. Kaiser Professor Dean Merck, Sharp & Dohme Research Health Policy and Management Washington University School of West Point, Pennsylvania Harvard School of Public Health Medicine Boston, Massachusetts St. Louis, Missouri

Diana B. Petitti Director, Research and Evaluation Kaiser Permanence Southern California Permanence Medical Group Pasadena, California

vii Project Staff

Clyde J. Behney PRINCIPAL CONTRACTORS PROJECT STAFF Assistant Director, OTA Dennis M. Black Robert McDonough Department of Clinical Study Director Sean R. Tunis Epidemiology Health Program Director University of California, San Judith L. Wagner Francisco Senior Associate ADMINISTRATIVE STAFF Elliott Pickar Katie Maslow Louise Staley Consultant, Senior Associ ate Office Administrator Rockville, MD Douglas Teich Carolyn Martin Senior Analyst Administrative Secretary

Laura Stricker Monica Finch Research Assistant Word Processing Specialist

William Adams Research Assistant

Julia Bidwell Research Assistant

Angela Schreiber Research Assistant

Viii contents

Appendix A: Acknowledgments 1

Appendix B: Evidence on Hormone Replacement Therapy and Fractures 5 References 17

Appendix C: Evidence on HRT and Bone Loss 19 Conclusions 30 References 31

Appendix D: Summary of Hip Fracture Prediction Methods 35 Requirements of the OTA Hip Fracture Prediction Model 35 Measuring Bone Mass 35 Predicting Bone Mass Over Time 36 Relationship of Bone Mass to Hip Fracture Risk 44 References 47

Appendix E: Hormonal Replacement Therapy Regimens 49 Estrogen Dosing Regimens 49 Therapy with PERT 53 Acute Indications for HRT 54 Evaluation and Followup of Women Taking HRT 55 Compliance with HRT 56 References 59

Appendix F: Evidence on Breast Cancer and Hormonal Replacement Therapy 65 Biological Plausibility 65 Case-Control Studies 66 Cohort Studies 67 Clinical Trials 68 Combined Estrogen-Progestin Therapy and Breast Cancer Risk 68 ix Stage of Breast Cancer at Diagnosis in HRT Users Versus Nonusers 70 Breast Cancer Mortality in HRT Users Versus Nonusers 71 Conclusions 72 References 73

Appendix G: Evidence on HRT and Endometrial Cancer 111 Duration and Dose of Estrogen 112 Recency of Use of Estrogen 112 Stage and Grade of Endometrial Cancer 112 Estrogen Use and Survival from Endometrial Cancer 114 Estrogen/Progestin Replacement Therapy 115 Implications for OTA’s Cost Effectiveness Model 117 References 118

Appendix H: Evidence on HRT and Gallbladder Disease 159 References 165

Appendix 1: Evidence on HRT and Coronary Heart Disease 167 How Estrogen May Affect CHD 167 Evidence on ERT and CHD 168 Evidence on PERT and CHD 189 Conclusions 191 References 193

Appendix J: Methods for Estimating costs 201 BMD Screening Cost 201 HRT Cost 201 Cost of Heart Disease 204 Cost of Hip Fracture 205 Cost of Cholecystectomy 205 Breast Cancer Costs 205 Endometrial Cancer Costs 207 References 210

Appendix K: Abbreviations and Glossary 211

x OTA wishes to express its appreciation to the following individuals and organizations that provided assistance with this volume. These individuals do not necessarily approve, disapprove, or endorse this report. OTA assumes fill responsibility for the volume and the accuracy of its contents.

Colette Z. Andrea Anne P. Clark Metra Biosystems Inc. Division of Research Grants Mountain View, CA National Institutes of Health Bethesda, MD Claude D. Arnaud Department of Medicine Robert S. Epstein School of Medicine Merck & Co., Inc. University of California West Point, PA San Francisco, CA

Martin Brown Bruce Ettinger Kaiser Foundation Research Institute National Cancer Institute Bethesda, MD Oakland, CA

Warren Browner Ronald G. Evens University of California Mallinckrodt Institute of Radiology San Francisco, CA Washington University Medical Center Saint Louis, MO Natalie Cannon Administration on Aging U.S. Department of Health and Human Services Seattle, WA

1 2 Cost Effectiveness of Screening for Osteoporosis

Susan C. Fox Dean K. Jenkins Women’s Health/Preventive Medicine Metra Biosystems Inc. Strang Cancer Prevention Center Mountain View, CA New York, NY Robert M. Kaplan Department of Family and Preventive Medicine Ted Ganiats University of California, San Diego Division of Family Medicine La Jolla, CA School of Medicine University of California, San Diego John C. LaRosa La Jolla, CA Office of the Chancellor Tulane University Medical Center Deborah Grady New Orleans, LA Departments of Epidemiology and of Medicine University of California Barbara Lindheim San Francisco, CA Therapeutics Strategic Marketing Sterling Winthrop, Inc. New York, NY Carolyn Green B.C. Office of Health Technology Assessment Nick Marmontello The University of British Columbia Wyeth–Ayerst Laboratories Vancouver, British Columbia Philadelphia, PA

Maria Greenwald Deborah Marshall Society for Clinical Densitometry SBU Norwich, NY The Swedish Council on Technology Assessment in Health Care Stockholm, Sweden Jay Hawkins Metra Biosystems Inc. Brian McCue Mountain View, CA Office of Technology Assessment Congress of the United States Washington, DC Elina Hemminki STAKES Joan McGowan National Research and Development Centre for Bone Biology Program Welfare and Health National Institute of Arthritis, Musculoskeletal, and Helsinki, Finland Skin Diseases Bethesda, MD Thomas Holohan Office of Health Technology Assessment Irma L. Mebane-Sims Agency for Health Care Policy and Research National Heart, Lung, and Blood Institute Rockville, MD Bethesda, MD Appendix A Acknowledgments 13

Alvaro Munoz Charles W. Slemenda School of Hygiene and Public Health Department of Medicine The Johns Hopkins University School of Medicine Baltimore, MD Indiana University Indianapolis, IN Eleanor M. Perfetto Wyeth–Ayerst Research Karen K. Steinberg Philadelphia, PA National Center for Environmental Health Centers for Disease Control and Prevention Mary Lake Polan Atlanta, GA Department of Gynecology and Obstetrics Stanford University Medical Center David B. Thomas Stanford, CA Division of Public Health Services Arnold Potosky Fred Hutchinson Cancer Research Center Applied Research Branch Seattle, WA National Cancer Institute Rockville, MD Anna N.A. Tosteson Department of Medicine Jerilynn C. Prior Dartmouth Medical School Department of Medicine Lebanon, NH The University of British Columbia Vancouver, British Columbia Wulf H. Utian Department of Obstetrics and Gynecology William Randolph University MacDonald Womens Hospital Bethesda, MD Cleveland, OH Lynn Rosenberg Slone Epidemiology Unit Richard D. Wasnich Boston University School of Medicine Hawaii Osteoporosis Center Brookline, MA Honolulu, HI

Philip D. Ross Joel B. Weinstein Hawaii Osteoporosis Center Hologic, Inc. Honolulu, HI Waltham, MA

Carl A. Roth Andrew R. Willan National Heart, Lung, and Blood Institute Hamilton, Ontario Bethesda, MD

Larry Shulman Bethesda, MD Appendix B: Evidence on Hormone Replacement Therapy and Fractures B

small number of studies have examined gion’s registry of hospital admissions. The mean directly the relationship between use of duration of observation of members of this cohort hormonal replacement therapy and risk of was 5.7 years. A hip fracture (table B-1). These studies, all Hormone use before 1977 and after 1980 was of which are of observational design, found a low- ascertained by mailing questionnaires to a random er incidence of hip fracture in estrogen users ver- sample of 735 women from the cohort in 1980 and sus nonusers, although the differences in inci- again in 1984 (14). Nine percent had not taken the dence did not always reach statistical significance. prescribed drug, half of the cohort had begun hor- The largest and most complete study of hormone replacement therapy before the beginning of monal replacement therapy and hip fracture inci- the study period, and the median duration of HRT dence was of a cohort of approximately 23,000 was approximately 3.5 years. Approximately one users of hormonal replacement therapy from the third of the HRT users were prescribed a com- Uppsala Health Care Region of Sweden (14). The bined regimen of estrogen and progesterone; the Uppsala Health Care Region comprises six coun- other two thirds used estrogen alone. ties and one-sixth of the total population of Swe- The incidence of hip fractures in the cohort was den. The authors included in the cohort all women compared with that of the background population living in the region who were age 35 years and old- of women age 35 and older in the region (14). The er (mean age 53.7 years at study entry) and who study demonstrated a statistically significant re- filled at least one prescription for a noncontracep- duction in risk of hip fractures in users of hor- tive estrogen-containing preparation between monal replacement therapy (relative risk 0.79). April 1977 and March 1980. HRT users were Women who received hormonal replacement identified through the region’s prescription record therapy with conjugated estrogens or estradiol database, which includes records of prescriptions showed a significantly reduced risk of hip fracture filled at all pharmacies within the region. The au- (relative risk 0.70), whereas virtually no protec- thors determined the incidence of first hip fracture tive effect was found in women prescribed es- in this cohort through 1983 by using each triols, which are much weaker than the conjugated woman’s unique national registration number to estrogens or estradiol typically prescribed for hor- link each woman’s prescription record with there- monal replacement therapy in the United States.

5 6 Cost Effectiveness of Screening for Osteoporosis

The data indicated that the protective effect of Compared to women who had never used HRT, the more potent estrogens was concentrated in current users of HRT had a significantly decreased women who initiated therapy within a decade after risk for wrist fractures (relative risk 0.46 (95% menopause (relative risk 0.55 for women less than confidence interval 0.29 to 0.72)) and all non- 60 years of age at initiation of therapy); no signifi- spinal fractures (relative risk 0.69 (0.57 to cant protective effect of hormonal replacement 0.83))1 (l). The relative risk of hip fracture was therapy was found for women 60 years or older at decreased but not statistically significant (relative the time of initiation of therapy (14). However, risk 0.80 (0.5 1 to 1.26)). By contrast, no associa- the data may not be sufficient to indicate whether tion was found between previous use of HRT and hormonal replacement therapy has a protective ef- the risk for either hip fractures (relative risk 1.00 fect for older women because of the relatively (0.72 to 1.07)) or for all nonspinal fractures (rela- small number of cohort members greater than 60 tive risk 0.97 (0.85 to 1.1 l)). Previous users had a years of age, and the relatively few older members 15-percent decrease in the risk for wrist fracture, of the cohort who were prescribed the more potent but the decrease was not statistically significant estrogens. (relative risk 0.85 (0.65 to 1.11)). Among current The study also found a greater protective effect users, the effect of PERT on fracture incidence of estradiol and conjugated estrogens for trochan- was similar to the effect of ERT. teric hip fractures (relative risk 0.60 (95 percent The investigators also found that the associa- confidence interval 0.35 to 0.96)) than for cervical tion between current HRT use and risk for wrist hip fractures (0.73 (0.55 to 0.95)) (14). This find- fractures and all nonspinal fractures was similar in ing supports earlier claims that trabecular bone, those younger and older than 75 years of age (1) which constitutes a larger part of the bone struc- (table B-l). They also found an 80 percent de- ture in the trochanteric than in the cervical part of crease in the risk for hip fractures among women the femur, is more rapidly affected by hormonal older than 75 years of age (relative risk 0.18 (0.04 changes than is cortical bone. to 0.77)), but found no effect on hip fracture in HRT has also been found to be effective in re- those 75 years of age or younger (relative risk 0.94 ducing other types of osteoporosis-related frac- (0.52 to 1.69)). tures. Cauley and colleagues, reporting on the Finally, the investigators found that HRT is Study of Osteoporotic Fractures cohort, found a more effective if initiated within 5 years of meno- decrease in risk of fractures in elderly women who pause and if used longer than 10 years (l). The in- currently used HRT, but not in elderly women vestigators found little effect of duration of use who previously used HRT (l). The Study of Os- among current HRT users on the occurrence of all teoporotic Fractures cohort includes 9,704 non- nonspinal fractures, but more than 10 years of use black women 65 years of age or older who were was associated with a substantial reduction in the recruited from population-based lists from four risk for wrist and hip fractures (table B-l). regions of the United States between September Women who began using HRT within five years of 1986 and October 1988. Members of the cohort menopause were found to have substantially were interviewed about HRT use and osteoporosis greater reductions in risk of wrist, hip, and all non- risk factors upon entry into the study, and women spinal fractures than women who began using were contacted every four months for up to 6.5 HRT more than 5 years after menopause (table years afterwards to determine whether they had a B-l). fracture.

] The investigators did not examine the incidence of spinal fractures in the cohort. Number of a Study participants Study design Type of fracture Description of study and population Results Gordan 120 on estrogen, Prospective cohort Fracture type Subjects were postmenopausal women, average CEE 1.25mg. 3 fractures/1 ,000 (1973) 100 on with internal unspecified age 62 at study entry, with osteoporosis as patient-years, androgens or controls determined by spinal x-ray evidence, who CEE 0.60mg. 25 fractures/ anabolic were seen at a San Francisco, CA, clinic 1,000 patient-years, steroids between 1948 and 1973. All subjects Androgens and anabolics 40 completed more than two years of hormone fractures/1 ,000 patient-years therapy. A total of 1,664 patient-years was No tests of statistical studied including 1,507 patient-years of CEE significance were provided. and 157 patient-years of androgens and anabolics.

Hammond 301 estrogen Retrospective Unspecified Subjects were patients at Duke Hospital or Estrogen-treated patients had (1979) users, 309 cohort with clinics (Durham, NC) between 1940 and 1979 an 8.6% fracture incidence controls Internal controls for diagnoses related to estrogen deficiency during study period. and who had been followed at Duke for at Patients not treated with least 5 years. estrogen had a 15.9% Estrogen use was defined as use greater than 5 fracture incidence during years. Average age at study entry for estrogen study period (relative risk users was 42.9 years, whereas average age 0.54). for nonusers at study entry was 49.6 years. No tests of statistical Average age of estrogen users at end of study significance were provided. was 56 years. Only 16 of the estrogen-treated group of 301 patients were black whereas 104 of the 309 patients not treated with estrogens were black.

Hutchinson 80 cases (107. Hospital-based Hip and distal Cases were admitted to Yale-New Haven Odds ratio “for protection” was (1979) estrogen case-control radius fractures Hospital (Connecticut) between 1974 and 3.0 (p = 0.01) for estrogen users); 80 1977. Controls were inpatients from the users versus nonusers, odds controls (25% orthopedic service during those same years, ratio increases to 3.5 if estrogen users) matched for race, age, and discharge date to estrogens are begun within 5 cases. All cases and controls were between years of menopause 40 and 80 years of age. Information was (p = 0.01) gathered through review of medical records and Interviews Estrogen use was defined as use greater than 6 months Number of a Study participants Study design Type of fracture Description of study and population Results Lindsay 58 mestranol- Prospective cohort Spine and radius Estrogen users and controls were post There was a significant (1 980) treated with internal fractures oophorectomy patients followed at a clinic in reduction in wedge patients; 42 controls Britain for a mean duration of 9 years. deformities of index controls Estrogen users were treated with mestranol, vertebrae (T4 and L2) 23.3mcg mean dose. All subjects were estrogen users. No tests of followed with spinal x-rays. All subjects were statistical significance were elderly patients with preexisting osteoporosis. performed.

Weiss (1980) 327 cases (34% Population-based Hip and distal Cases were white women, aged 50 to 74 years, Current use. relative risk 0.42 estrogen case-control radius fractures followed in 59 outpatient orthopedic clinics in (0.30-0.63) in hip fracture users), 576 King County (Seattle), WA, for hip fractures patients versus controls. controls (52% and wrist fractures that occurred between Long-term use (> 10 yr.). relative estrogen users) 1978 and 1979. Controls were of the same risk 0.46 (0.30-0,69) in hip ages as cases and selected from the same fracture patients versus region. All subjects were interviewed about controls. Decreased risk of estrogen use, fracture history, and hip fracture was seen only osteoporosis risk factors. with more than 5 years of hormonal replacement therapy.

Johnson 168 cases (29.2% Hospital based Hip fractures Cases and controls were members of the Relative risk of hip fracture in (1981) estrogen case-control Kaiser-Permanente Medical Program of estrogen users versus users); 336 Portland, OR, who were identified through nonusers was 0.72 controls (36% medical records. Cases were women between (0.48-1 .09) (p= 0.06). estrogen users) 52 and 80 years old hospitalized for fracture of Controls were more likely to the proximal end of the femur between 1965 have long duration (at least 36 and 1975. Two hospital controls were selected months) of exposure and for each case, matched for age and date of more likely to begin estrogen discharge, Estrogen use was ascertained from within three years of medical records. menopause than cases, Estrogen exposure was defined as written order although differences did not in the medical records of estrogens taken achieve statistical during or after the year of menopause and significance. prior to hospitalization Authors noted that the number of cases was sufficient only to detect a reduction of risk of about 50 percent or greater.

I Number of a Study participants Study design Type of fracture Description of study and population Results Paganini-Hill 91 Cases (49% Population-based Hip fractures Cases and controls were residents of Leisure Relative risk with long-term use (1981) estrogen case-control World Retirement Community, Los Angeles, (> 60 months) was 0,42 users); 166 CA. Cases were postmenopausal women who (0.18-0.98) in hip-fracture controIs (52% had hip fractures between 1974 and 1978, cases versus community estrogen users) were less than 80 years of age, and were controls. identified from local hospital records. Controls Protective effect of greater than were selected from the retirement community, 60 months estrogen use was matched for age, race, and date of entry into limited to oophorectomized the community. Estrogen use was recorded women, relative risk was 0.14 from outpatient medical records and personal (0.03-0.70), relative risk of interviews. estrogen use in hip fracture cases with natural menopause was 0.86 (nonsignificant).

Kreiger 98 cases; 884 Hospital-based Hip fractures Cases and controls were admitted to inpatient Relative risk in ever users (1982) controls (83 case-control surgical services at a Connecticut hospital versus nonusers was 0.5 trauma between 1977 and 1979. Controls were (0.3-1.1) for cases and controls, 801 admitted for nongynecologic diagnoses. Two trauma controls, 0.5 (0.3-0.9) nontrauma control groups were used for comparison: for cases and nontrauma control) trauma controls and nontrauma controls. Ever controls. use was defined as use greater than 6 months.

Riggs (1982) Five groups, Prospective cohort Vertebral Study subjects were postmenopausal women The fracture rate was 834 per including 45 with internal fractures referred to the Mayo Metabolic Bone Disease thousand person-years in the receiving no controls Clinic, Rochester, MN, between 1968 and untreated patients and 181 treatment and 1980. Estrogen users received 0.625 to 2.5 per thousand person-years 32 receiving mg/d of CEE with 1,000 to 3,000 mg/d of in the estrogen group (p < estrogen with calcium. Fifteen estrogen users had also 1X10 -6 calcium received vitamin D (50,000 units once or twice weekly). Mean age of estrogen users was 63.8, and mean age of untreated group was 62.9. Subjects were followed up in clinic, mean duration of followup for estrogen users was 4,5 years, and for subjects assigned to placebo, 2.0 years. Number of a Study participants Study design Description of study and population Results Williams 344 cases (34% Population-based Hip and forearm Study subjects were white women ages 50 to 74 The beneficial effect of estrogen (1982) estrogen case-control fractures who had sustained hip or forearm fractures use in preventing hip and users); 567 between 1976 and 1979. Cases were followed forearm fractures varied controls (52% by orthopedic surgeons in King County, according to a woman’s estrogen users) Washington. Controls were a random sample weight and smoking status, of white female residents of King County in the being greatest in thin women same age range. Estrogen use was who smoked cigarettes and ascertained by interviews with study subjects. near zero in heavy nonsmokers. Reduction in risk of forearm fracture in thin smokers by use of estrogen. 4.7 Reduction in risk of hip fracture in thin smokers by use of estrogen: 7.1 Reduction in risk of forearm fracture in average weight smokers by use of estrogen: 0.8 Reduction in risk of hip fracture in average weight smokers by use of estrogen. 4,4 No tests of statistical significance were done.

I Number of a Study participants Study design Type of fracture Description of study and population Results Ettinger 245 estrogen Retrospective Wrist, spine, and Study subjects were white postmenopausal Relative risk for osteoporotic (1985) users; 245 cohort with all fractures women, average age 73 years, identified from fracture was 2.2 (1 ,5-3.8) (in controls internal controls review of pharmacy records dated between nonusers versus users); 1968 and 1971. Black women were excluded relative risk for spine from the study because of the low incidence fractures was 2.7 (1 .0-8.1). of fractures in this group. Subjects were There was no significant followed for an average of 17.6 years. Study difference in risk of hip subjects were followed at the Kaiser fractures or wrist fractures Permanence Medical Center, San Francisco, between users and CA. Controls were matched for age and length nonusers, of membership in the health plan, Estrogen use was defined as use begun within three years of menopause and at least 5 years of estrogen use. Fracture incidence and continued estrogen use was determined by reviewing medical records.

Kiel (1987) 2,873 women Retrospective Proximal femur Study subjects were members of the Unadjusted relative risk in (667 estrogen cohort with fractures Framingham (Massachusetts) Heart Study women who have taken users) internal controls cohort between 1948 and 1985. Subjects estrogen within the previous were ages 30 to 62 years old at the 1st two years versus never users biennial examination held between 1948 to was 0,34 (0.08-0,64), 1951. Only women who reached menopause Unadjusted relative risk in during the study interval were included in this women with any estrogen analysis. Information on fractures was use versus nonusers was gathered from review of hospital records and 0.69 (0.46-1 .03). interviews of subjects at the cohort’s 18th Unadjusted relative risk for biennial examination (1983-1985). Subjects recent users of less than one had provided information about estrogen use year was 0.32 (0.09-1 .21); at most of the biennial examinations. relative risk in recent users for more than one year was 0.14 (0.03-0.76). —

Number of a Study participants Study design Type of fracture Description of study and population Results Naessn 23,246 women (all Prospective cohort Hip fractures Cohort included all women 35 years of age or Relative risk was 0.79 (0.68 to (1 990) HRT users) with external older from the Uppsala Health Care Region of 0.93) in estrogen users controls Sweden who received noncontraceptive compared with the estrogens from April 1977 to March 1980. background population o Comparisons were made with expected rates of incidence of hip fracture in women in the background population. Women were followed for an average of 5.7 years.

Paganini-Hill 8,600 women Prospective cohort Hip fractures Cohort Includes postmenopausal women who Relative risk for ever use was (1991) (4,866 ever with internal were residents of Leisure World Retirement 1.02 (0.81-1 .27) estrogen users) controls Community, Los Angeles, CA. Mean age of Dose, residents was 73 years. Mailed questionnaires < 0.625mg CEE: 0.84 were sent in 1981, 1982, 1983, and 1985. (0.58-1 .21) Cohort was followed for area hospital > 1,25mg CEE: 0.91 (0.64-1.29) admissions and local health department death Duration: certificates through 1988. <3 years, 1,19 (0.89-1 .60) 4-14 years: 0,89 (0.63-1 .23) >15 years 0,88 (0,63-1 .24) Recency (years since last estrogen use): >15 years: 1,15 (0.88-1 .50) 2-14 years. 0.88 (0.63-1 .23) O-1 years. 0.80 (0.53-1 .21) Duration and Recency. <3 years, >15 yrs. 1,33 (0.97-1 ,82) 2-14 yrs. 0.79 (0.38-1 .60) O-1 yrs. 0.87 (0.28-2.73) 4-14 years: > 15 yrs. 0.95 (0.61 -1.49) 2-14 yrs. 0.86 (0.52-1 .42) O-1 yrs. 0.72 (0.31-1 ,64) > 15 years: > 15 yrs. 0.57 (0.1 8-1 .79) 2-14 yrs. 0,97 (0.61-1 .53) O-1 yrs. 0.85 (0.53-1 .38) Number of a Study participants Study design Type of fracture Description of study and population Results Kanis (1992) 2,086 cases, Population-based Hip (femoral The Mediterranean Osteoporosis Study Relative risk 0.45 (0,30 to 0,67) 3,532 controls case-control neck) fractures examined the incidence of hip fracture in men (P = 0.0001) in ever users and women aged 50 years or over from 14 versus never users. centers from six countries in Southern Europe. Adjusted relative risk 0,55 (0,36 Cases were women over age 50 (mean age to 0.85) (p = 0.01) (adjusted 78 years) who had a hip fracture over a for center, age, body mass one-year period (1988 to 1989). Cases were index, and previous fragility identified by surveillance of hospitals, private fractures). clinics, and nursing homes in the catchment Fracture risk stratified by age, area. Controls of the same age and who were <80 years, neighbors of cases or were sampled from population registers were identified. Adjusted relative risk 0.51 (0.31 Information about use of drugs affecting bone to 0.84) (p = 0.009) >80 years, was obtained by interview. Investigators from each center provided prospective information. Adjusted relative risk 0,70 (0.29 Only 1.9 percent of cases and 3.5 percent of to 1.66) (NS) controls had ever used estrogen. Duration: Less than median duration of estrogen use: relative risk 0.86 (0.51 to 1,46) Greater than median duration: relative risk 0.29 (0.1 3 to 0,61 )

Lufkin (1992) 75 women Randomized Vertebral Study subjects were postmenopausal white Eight new fractures occurred to clinical trial fractures women, 47 to 75 years of age, with seven women in the estrogen established osteoporosis (defined as BMD group, whereas 20 new below the 1Oth percentile of normal fractures occurred in 12 postmenopausal women and one or more women in the placebo group vertebral fractures) seen at Mayo Clinic, (relative risk 0.39 (0.16 to Rochester, MN. Subjects were randomly 0.95). assigned to placebo or treatment with transdermal estrogen (Estraderm patch) for 3 weeks out of a 4-week cycle, with 10 mg/d oral medroxyprogesterone acetate. The study duration was one year. Number of a Study participants Study design Type of fracture Description of study and population Results Spector 1,075 HRT users, Retrospective Osteoporotic HRT users were women, average age 52, who Relative risk of distal radius (1992) 1,471 controls cohort with fractures attended a Dulwich, South London, UK, fractures was 0.70 (0.32 to o external controls menopause clinic for hormone replacement 1,55) compared with therapy between January 1976 and pre-HRT fracture rates and December 1986. Controls were 0.63 (0.31 - 1.31) compared postmenopausal women of the same age from with nonusers. the registers of four general practices in Relative risk of osteoporotic Greater London. Information was gathered by fractures was 0.96 (0.55 - questionnaire and review of medical records. 1.68) compared with Most HRT users (65%) received subcutaneous pre-HRT fracture rates and estrogen implants (50 to 100 mg estradiol) 0.71 (0.43 to 1.16) compared every six months, usually in combination with with nonusers. 100 mg testosterone. Cyclical progestins were There was a trend toward given for 12 days each cycle to most estrogen decreased incidence of users who had not had a hysterectomy. fractures with increased Average duration of HRT use was 51 months. duration of use for osteoporotic fractures (P = 0.06) and wrist fractures (p= 0.03). There were 6 reported fractures of the hip in the nonusers compared with none among estrogen users (p = 0.15).

Grisso 144 cases (4% Case-control study Hip fracture Cases were black women admitted with first hip Fracture risk stratified by age*: (1994) HRT users), with both fracture to 1 of 30 hospitals in New York and Less than 75 years old: 218 community hospital and Philadelphia, Community controls were black adjusted odds ratio 0.05 controls (8% community women living in the community who were (0.01 -0.6) compared with HRT users), controls matched to cases by age and geographic community controls 181 area, Hospital controls were black women adjusted odds ratio 0.1 hospitalized matched by age and hospital. Information was (<0.1 -0.5) compared with controls obtained through personal interviews. hospital controls Number of a Study participants Study design Type of fracture Description of study and population Results Grisso-cent. Age 75 years or more: (1994) adjusted odds ratio 0.3 (0.1 -1 .2) compared with community controls adjusted odds ratio 1.1 (0.2-6.3) compared with hospital controls Duration*: 1 to 6 years: adjusted odds ratio 0.7 (0.2-3.1 ) more than 7 years: adjusted odds ratio 0.2 (0.1-1 .0) compared with community controls Recency (time since last use)*: less than 5 years: adjusted o odds ratio 0.0 (0-0.95) 5 or more years: adjusted odds ratio 0.6 (0,2-1 .7) compared with community controls ● All results are for HRT use for 1 year or more Lafferty 157 women (52% Prospective cohort Vertebral Study subjects were all white postmenopausal Relative risk of spinal (1994) estrogen users) with internal compression women between 43 and 60 years of age seen compression fracture 0.27 controls fractures and in a Cleveland, OH, private practice between (0.12-0.60) in estrogen users peripheral 1964 and 1983. Subjects were followed versus nonusers; relative risk fractures though 1989, All estrogen users received of peripheral fracture was 0.625mg/day CEE for the first 25 days per 0.23 (0.06-0.97); relative risk month After 1983, 5mg/day of of all fractures was 0.28 medroxyprogesterone acetate was given for (0.09-0.89). 12 days each month, Estrogen use was defined as use of at least 3 years duration (68% of ERT users took estrogens for more than 10 years). Number of Study participants Study design Type of fracture Description of study and population Results a Cauley 9,704 women Prospective cohort Hip fractures, Subjects were nonblack women 65 years of age Current versus past HRT use: (1995) (13.7% current with internal wrist fractures, or older who were members of the Study of Hip fractures: users) (27.4°A controls and all Osteoporotic Fractures cohort, a prospective current HRT use: ever users) nonspinal study conducted at four clinical centers in the age adjusted relative risk 0.80 fractures United States. The SOF cohort was recruited (0.51-1 .26) from population-based lists of women (voter risk-factor adjusted relative registration, driver’s license, and health risk’ 0.60 (0,36-1 .02 maintenance organization membership lists). past HRT use: Black women were excluded because of their age adjusted relative risk 1.00 low incidence of fractures. Information on HRT (0.72-1 .07) use and risk factors was gathered from risk-factor adjusted relative interviews at baseline and information about risk 1.03 (0.69-1 .55) incident fractures was gathered by postcard Wrist fractures: or telephone every 4 months. Subjects were current HRT use: recruited from September 1986 to October age adjusted relative risk 0.46 1988, and followed through March 1993. (0.29-0.72) Duration of followup ranged from 0.02 to 6.5 risk-factor adjusted relative years. risk 0.39 (0.24-0.64) past HRT use: age adjusted relative risk 0.85 (0.65-1 .11) risk-factor adjusted relative risk 0.81 (0,62-1 .07) All nonspinal fractures, current HRT use: age adjusted relative risk 0.69 (0.57-0.83) risk-factor adjusted relative risk 0.66 (0.54-0.80) past HRT use. age adjusted relative risk 0.97 (0.85-1 .11) risk-factor adjusted relative risk 0,94 (0,83-1 .08)

* Relative risk was adjusted for multiple — osteoporosis risk factors a95 percent confidence intervals are given in parentheses, unless otherwise specified KEY: BMD = bone mineral density; CEE = conjugated equine estrogen, ERT = estrogen replacement therapy; HRT = hormonal replacement therapy; SOURCE: Office of Technology Assessment, 1995 Appendix B Evidence on Hormone Replacement Therapy and Fractures 117

Providing further support for the proposition er, due to the relatively low incidence of hip frac- that hormonal replacement therapy reduces fractures relative to vertebral fractures, the long ture incidence are a number of clinical trials of duration between menopause and the age at which hormonal replacement therapy in postmenopausal most hip fractures occur (above 65 years), and dif- women which demonstrate statistically signifi- ficulties in maintaining compliance with hormon- cant reductions in incidence of osteoporosis-real replacement therapy over that long a time lated fractures other than hip fracture in users of period. hormonal replacement therapy (13,17) (table B-l). One controlled clinical trial demonstrated the effectiveness of hormonal replacement therapy in REFERENCES reducing the incidence of vertebral fractures in a 1. Cauley, J., Seeley, D., Ensrud, K., et al., “Es- group of women with established osteoporosis trogen Replacement Therapy and Fractures in (13). In this study, 75 postmenopausal women, 47 Older Women,” Annals of Internal Medicine to 75 years of age, with one or more vertebral frac- 122:9-16, 1995. tures due to osteoporosis, were randomly assigned 2. Ettinger, B., Genant, H. K., and Cann, C. E., to treatment with an estradiol patch and oral me- “Long-Term Estrogen Replacement Therapy droxyprogesterone acetate or a placebo patch. Prevents Bone Loss and Fractures,” Annals of Bone mineral density and vertebral fractures were Internal Medicine 102(3):319-324, 1985. assessed at the beginning of the study and after 3. Gordan, G., Picchi, J., and Roof, B., “Anti- one year. Bone mineral density was maintained or fracture Efficacy of Long-Term Estrogens for increased in the treatment group at all sites mea- Osteoporosis,” Transactions of the Associa- sured. Eight new fractures occurred in 7 women in tion of American Physicians 86:326-332, the estrogen group, whereas 20 occurred in 12 1973. women in the placebo group, yielding a signifi- 4. Grisso, J., Kelsey, J., Strom, B., et al., “Risk cantly lower vertebral fracture rate in the estrogen Factors for Hip Fracture in Black Women,” group (relative risk 0.39 (95 percent confidence New England Journal of Medicine 330(22): interval 0.16 to 0.95)). 1555-1559, 1994. Estrogen treatment would probably also de- 5. Hammond, C., Jelovsek, F., Lee, L., et al., crease hip fracture rates, as it does vertebral frac- “Effects of Long-Term Estrogen Replace- ture rates, because bone mineral density in the ment Therapy: I. Metabolic Effects,” Ameri- estrogen group increased at the hip sites studied. can Journal of Obstetrics and Gynecology The reduction in hip fracture rates may not be pro- 133(5):525-536, 1979. portional to the reduction in vertebral fracture 6. Hutchinson, T., Polansky, S., and Feinstein, rates, however, in part because of differences in A., “Postmenopausal Oestrogens Protect the qualitative features of the bone at both sites, Against Fractures of Hip and Distal Radius: A and because such factors as neuromuscular weak- Case Control Study,” Lance? 705-709, 1979. ness, postural instability, and the tendency to fall 7. Johnson, R. E., and Specht, E. E., “The Risk of play a greater role in fractures of the hip. Hip Fracture in Postmenopausal Females Controlled clinical trials of the relation be- with and Without Estrogen Drug Exposure,” tween HRT use and vertebral fracture have been American Journal of Public Health 71(2); possible because of the relatively large number of 138-144, 1981. vertebral fractures in postmenopausal women and 8. Kanis, J. A., Johnell, O., Gullberg, B., et al., the relatively early age at which vertebral fractures “Evidence for Efficacy of Drugs Affecting typically occur. Conducting a controlled clinical Bone Metabolism in Preventing Hip Frac- trial of hormonal replacement therapy and hip ture,” British Medical Journal 305(7): fracture presents much greater problems, howev- 1124-1128, 1992. 18 Cost Effectiveness of Screening for Osteoporosis

9 Kiel, D.P., Felson, D.T., Anderson, J. J., et al., tion of Hip Fracture: The Leisure World “Hip Fracture and the Use of Estrogens in Study,” Epidemiology 2(l): 16-25, 1991. Postmenopausal Women: The Framingham 16. Paganini-Hill, A., Ross, R. K., Gerkins, V.R., Study,” New England Journal of Medicine et al., “Menopausal Estrogen Therapy and 317(19):1169-1174, 1987. Hip Fractures,” Annals of Internal Medicine 10 Kreiger, N., Kelsey, J.L., Holford, T.R., et al., 95(1):28-31, 1981. “An Epidemiologic Study of Hip Fracture in 17. Riggs, B. L., Seeman, E., Hodgson, S.F., et Postmenopausal Women,” American Journal al., “Effect of the Fluoride/Calcium Regimen of Epidemiology 116(1): 141-148, 1982. on Vertebral Fracture Occurrence in Postme- 11. Lafferty, F.W., and Fiske, M. E., “Postmeno- nopausal Osteoporosis,” New England Jour- pausal Estrogen Replacement: A Long-Term nal of Medicine 306(8):446-450, 1982. Cohort Study,” American Journal of Medi- 18. Spector, T. D., Brennan, P., Harris, P. A., et al., cine 97(1):66-77, 1994. “Do Current Regimes of Hormone Replace- 12. Lindsay, R., Hart, D. M., Forrest, C., et al., ment Therapy Protect Against Subsequent “Prevention of Spinal Osteoporosis in Oo- Fractures?” Osteoporosis International phorectomized Women,” Lancet 1151-1154, 2:219-224, 1992. 1980. 19. Weiss, N. S., Ure, C. L., Williams, J. H., et al., 13. Lufkin, E.G., Wahner, H. W., O ‘Fallen, “Decreased Risk of Fractures of the Hip and W.M., et al., “Treatment of Postmenopausal Lower Forearm with Postmenopausal Use of Osteoporosis with Transdermal Estrogen,” Estrogen,” New England Journal of Medicine Annals of Internal Medicine 117(1):1-9, 303(21):1195-1198, 1980. 1992. 20. Williams, A.R., Weiss, N. S., Ure, C. L., et al., 14. Naessen, T., Persson, I., Adami, H. O., et al., “Effect of Weight, Smoking, Estrogen Use on “Hormone Replacement Therapy and the the Risk of Hip and Forearm Fractures in Risk for First Hip Fracture,” Annals of Inter- Postmenopausal Women,” Obstetrics & Gy- nal Medicine 113:95-103, 1990. necology 60(6):695-699, 1982. 15. Paganini-Hill, A., Chao, A., Ross, R. K., et al., “Exercise and Other Factors in the Preven-

I Appendix C: Evidence on HRT and Bone Loss c

large number of controlled clinical trials nopausal women who fail to respond to HRT ( 16). have demonstrated that hormone replace- Recent analyses have found that the proportion of ment therapy (HRT) is able to reduce the women who fail to respond to hormone replace- A rate of bone loss in postmenopausal ment therapy is relatively small (20). women. Most of the controlled clinical trials of Only a handful of studies of HRT and bone HRT on bone mass have been for a duration of mineral density have followed women more than three or fewer years (table C- 1 ). The first pages of three years after initiation of therapy (table C-2), table C-1 (“HRT and Bone Mineral Density: Clin- and these studies have shown that HRT maintains ical Trials of 3 or Fewer Years”) provide details of bone mass or reduces the rate of bone loss in post- the design and results of each of the studies. The menopausal women compared with placebo. In a percentage change in bone mass from the baseline retrospective cohort study, Meema and colleagues measurement to the end of the study is provided contacted postmenopausal women who had a so that we may compare bone mass data that are bone mass measurement at a university clinic four given in disparate units (e.g., bone mineral con- to 10 years previously and asked them to volun- tent (usually measured in grams per centimeter teer for a second bone mass measurement (36). (g/cm)) and bone mineral density (usually mea- Eighty two volunteers were identified, 29 of sured in grams per square centimeter (g/cm2)). whom had been treated continuously with estro- Virtually all of these studies have shown that gens. After an average followup period of six HRT, begun soon after menopause, maintains or years, the estrogen-treated women showed no sig- increases bone mass within the first three years af- nificant changes in bone mass and cortical thick- ter menopause. Although HRT may reduce the ness, whereas untreated women had significant rate of bone loss after menopause, HRT is not able decreases in bone mass and cortical thickness. In to substantially restore bone mass that is lost. The a cross-sectional study, Moore examined the bone increases in bone mass seen with initiation of ther- mineral density of 65 postmenopausal women be- apy soon after menopause are small, generally in tween 55 and 75 years of age who were at least 10 the range of 1 to 3 percent of the total bone mass. years from menopause (37). Long-term estrogen A number of investigators have questioned users were defined as those women who had be- whether there is a significant subgroup of postme- gun therapy within five years of menopause and

I 19 20 Cost Effectiveness of Screening for Osteoporosis Duration of monitoring of bone Study Number of participants density Type of treatment Results Christiansen part I: 43 treated; 51 3 years, Part Patients treated with Trisequens forte Part I: (1981 b) controls. I: first two (17-beta-estradiol (4mg) and estriol (2mg) days HRT (g/cm) 2.5% Part II: 35 treated, 42 years of 1-12, 17-beta-estradiol (4mg) estriol (2mg) and Placebo -3.8% controls study; Part norethisterone acetate 1 mg days 13-22, Part II: II: third 17-beta-estradiol (1 mg) and estriol (0.5mg) days HRT 3.7% year of 23-28), All patients received 500mg calcium per Placebo 0.2% study day. HRT -2.4% Placebo -5.7% Finn Jensen (1 982) 31 treated; 43 controls 18 mo. (6 Patients divided into four groups, Post six month run-in period, -1.91 %

mo. run-in 1,25(OH)2 D3 (0.50 mg/d) + 500mg calcium, 19 1,25 + calcium +3.62% period) patients; Trisequens + 500mg calcium, 11 calcium + hormones +3.06%

patients; 1 ,25(OH)2 D3 (0.50mg/d) + Trisequens + 1,25 + hormones + calcium -0.39% 500mg calcium: 20 patients; 500mg calcium 24 calcium patients

Lindsay (1984) 887 treated patients divided 2 years Patients assigned to either placebo group or to CEE Placebo -8.23% among four groups; 21 at one of four dosage levels: 0,15mg/day, 0.15mg/d -8.51 % controls 0.3mg/day, 0.625 mg/day, and 1.25mg/day 0.30mg/d -5.01 % 0.625mg/d -0.24% 1.25mg/d -0.00%

Christiansen (1984) 2 treatment groups and one 1 year Two treatment groups, 17 beta-estradiol and E2+P +0.52%

placebo group, E2 +E3 norethisterone acetate; 17 beta-estradiol, estriol, E2+E3+P +1.53%

+P: 22 patients; E2 +P: 20 and norethisterone acetate, placebo group Placebo -3.3% patients; placebo group: Daily doses used were, 17-beta-estradiol 2mg from 23 patients days 1-22, 1 mg from days 23-28, estriol: 1 mg days 1-22, 0.5mg days 23-28, norethisterone acetate, 1 mg days 13-22. All patients received 500mg/d calcium, —

Duration of monitoring of bone Study Number of participants density Type of treatment Results Caniggia (1984) 22 patients group 1: (n=5) 1 year Four groups: Placebo -8.O% +9.O% 1,25(OH) 2 D3, group 2: l ,25 (OH)2 D 3 (0.5mcg) 1 ,25(OH)2D3 ■ (n=5) estradiol valerate, ,25 (OH)2 D3 0.5mcg+ 1 ,25(OH)2D3+-E2 +6.0% +9.0% group 3: (n=7) 1,25(OH)2 estradiol valerate (2mg/d) E2 D & estradiol valerate ■ stradiol valerate (2mg/d) group 4: (n=5) placebo n placebo No statistical analysis of data.

Gotfredsen (1986) 52 treated, 52 controls 1 year Treated patients received either 17-beta-estradiol, Placebo: either percutaneously (one daily dose of 5g, Head -5. O+-1.5% corresponding to 3mg 17-beta-estradiol) or orally Chest -7.0+-2.0% (sequentially administered oral 17-beta-estradiol Arms -3.0+-1.0% 2mg and for 10 days each cycle 1 mg cyproterone Pelvis -5.5+-1.5% acetate). Legs -4.3+-0.3% Spine -3.0+-2.5% HRT: Head + 1.7+-1.3% Chest +2.3+-2.0% Arms +0.3+-0.5% Pelvis -1.0+-2.0% Legs -0.7+-0.25% Spine -2.1+-1.5% DPA lumbar spine. HRT + 1.0% Placebo 0.0% DPA total spine. HRT 0.0% Placebo -2.5% SPA forearm HRT o o% Placebo -2.0% Appendix C Evidence on HRT and Bone Loss 123 Duration of monitoring of bone Study Number of participants density Type of treatment Results Munk-Jensen 50 continuous estrogen and 18 months Group 1, continuous estradiol 2mg and Distal forearm, (1988) progestogen; 50 (including norethisterone acetate 1 mg; group 2, cyclic Estrogen and progesterone -0.8+-0.6% sequential estrogen and 6 mo. estradiol 2mg and 10 days per month 1 mg (continuous) progestogen, 51 placebo run-in norethisterone acetate; group 3, placebo. Estrogen and progesterone -2.0+-O.5% period) 6 month run-in period where all patients were (sequential) untreated Placebo -5.6+-0.55% Lumbar spine: Estrogen and progesterone +4,2+-0.8% (continuous) Estrogen and progesterone +3.2+-0.55% (sequential) Placebo

Riis (1988) 21 treated, 2 years Patients assigned to either continuous Forearm (prox,). 22 controls 17- b -estradiol, 2mg, and norethisterone acetate, 1 HRT + 1.0+-1 .9% mg, or placebo Placebo -4.5+-2.7% Forearm (distal): HRT +0.8+-3.8% Placebo -7.5+-3.8% Spine: HRT +5.4+-7.7% Placebo -3.7+-8.0%

Genant (1 990) 94 treated; 28 1 year 30 patients treated with 0.3mg estrone sulfate; 32 0.3mg estrone sulfate -3.22% placebo-controls patients treated with 0.65mg estrone sulfate, 32 0.625mg estrone sulfate + 1.38% patients treated with 1.25mg estrone sulfate. 1.250mg estrone sulfate +2.62% Purpose of study was to determine minimum Placebo -0.82% effective dose of estrogens. All patients given 1,000mg elemental calcium supplementation

Lindsay (1990) 22 estrogen treated; 18 2 years All treated and controls given calcium to bring their Vertebrae. controls total intake to 1,500mg/d, treated patients Controls -7.6% received CEE 0.625 mg/d, and those with an intact Treated +6.4% uterus received medroxyprogesterone 5 to 10mg Hip: for 12-14 days a month Controls -4.13% Treated +9.2% Duration of monitoring of bone Study Number of participants density Type of treatment Results Resch (1990) 9 treated, 9 controls 1 year Nine patients treated with Trlsequens; nine patients HRT +8.84% treated with placebo, all patients received 500mg Placebo 0.0% calcium; Trisequens is estradiol (2mg) and norethisterone acetate (1 mg) Stevenson (1990) 66 treated, 18 mos. 33 patients treated with transdermal 17-b estradiol Transdermal: 30 controls 0.05mg daily with transdermal norethisterone Spine (L2-L4) +3. 14% acetate 0.2mg to 0.3mg per day for 14 days a Femoral neck +3. 14% cycle; 33 patients treated with oral CEE 0.625mg Wards triangle + 1 .0% daily with dl-norgestrel 0.15mg daily for 12 of the Trochanteric 0.0% 28 days Oral HRT: Spine + 1 .71% Femoral neck + 1.00% Wards triangle +2.00% Trochariteric +2.66% Untreated: -1.93% Spine Femoral neck -3.16% Wards triangle -4.32% Trochanteric -2.15% 26 I Cost Effectiveness of Screening for Osteoporosis

I Duration of Number of Study monitoring Study participants design bone mass Type of treatment Results

Meema, et al. 29 control Retrospective 4 to 10 years Most frequently used hormone preparations Castrates: +1.92% (1 975) 53 treated cohort followup (6 were conjugated equine estrogens Estrogen-treated -7.78% years (0.625mg or 1.25 mg) usually Untreated +1.12% average administered cyclically Natural menopause: -6.30% followup) Estrogen-treated Untreated

Lindsay, et al. 14 controls; 15 Clinical trial 8 years Mean daily dose 27.6 mcg mestranol; 14 Placebo group -11.9% (1978b) treated 8 years; 14 patients placebo; 14 patients 4 yrs. Estrogen group -0.7% treated, then mestranol treatment then placebo 4 Estrogen, then withdrawal -10.070 treatment years; 15 patients received 8 years of after 4 years withdrawn mestranol treatment

Nachtigall, et al. 67 treated Clinical trial 10 years Treated patients received CEE 2.5mg/day <3 years from LMP: +8.67% (1979) 62 controls and 7 days each month Estrogen-treated -9.00% medroxyprogesterone acetate 10mg. Placebo control -0.5070 >3 years from LMP: -11.29% estrogen-treated placebo control

Lindsay, et al. 42 control Clinical trial Mean duration Treated with mestranol mean daily dose Placebo: -10.4% (1980) 58 treated 9 years 23.3mcg Metacarpal -9.45% Radius -1 .90% Estrogen-treated: -2.1 5% o Metacarpal Radius

KEY: LMP = last menstrual period. SOURCE: Office of Technology Assessment, 1995 28 I Cost Effectiveness of Screening for Osteoporosis

who continued for a duration of at least 10 years. and 3) beginning therapy at 65 and continuing for The mean duration of estrogen use among long- the remainder of life (12). Their model included a term estrogen users was 19.8 years. Controls were number of key assumptions, based on their review postmenopausal women who used estrogen for of studies of the impact of hormonal replacement less than one year. There was a significant differ- therapy in the elderly, including the assumption ence in mean spinal bone mineral density between that bone mass would increase by 5 percent to 10 long-term estrogen users (1.21 9 g/cm2) and con- percent in the first two years after initiating thera- trols (1.092 g/cm2), and this significant difference py in the elderly. The investigators concluded that was retained after controlling for age and type of women who begin therapy at menopause and stop menopause. at age 65 have only a small (8 percent) increase in In the only long-term prospective clinical trial bone density at ages 75 to 85, the ages of highest of HRT and bone mineral density, 84 pairs of post- hip fracture incidence, compared to never users, menopausal nursing home patients were random- which translates into a 23-percent reduction in ly assigned to estrogen and progesterone or place- fracture incidence. Women who begin therapy at bo (39). After 10 years, HRT-treated women had menopause and continue for the remainder of life no significant decrease in bone mass. Women who were predicted to have the highest mean bone den- began HRT within three years of menopause had a sity at ages 75 to 85, about 22 percent higher than small but significant increase in bone mass after never users, and the greatest reduction in fracture 1() years. Women assigned to placebo had a signif- incidence, a 73-percent reduction. But women icant decrease in bone mass. who began HRT at age 65 had almost as great an A number of studies have demonstrated that increase in bone density, from 14 to 19 percent, HRT is able to halt or possibly reverse bone loss and almost as great a reduction in fracture inci- even if it is started long after menopause (9,31, dence, from 57 to 69 percent, as women who be- 32,35,41,43,45). Gains in bone mass of 5 to 10 gan HRT at menopause and continue for the rest of percent or more have been found after initiation of their lives. HRT in the elderly. In a prospective study of 397 Ettinger and Grady argued that starting hor- postmenopausal women between the ages of 51 mone therapy later in life would halve the period and 80 years, Quigley found that estrogen replace- of hormone exposure, reducing the potential risks ment therapy reduced bone loss to about the same of very long-term estrogen therapy (12). rate for estrogen users regardless of age (43). There are several other reasons for beginning Ettinger and Grady predicted that beginning HRT in the elderly. Many of the early estimates of therapy later in life may provide almost as much the rate of bone loss with aging were derived from protection against osteoporotic fractures as start- cross-sectional studies, which may be biased if ing at menopause (12). Ettinger and Grady used there are cohort effects. Recent prospective stud- data on the effects of hormone replacement thera- ies of bone loss with aging demonstrate that bone py on bone density, and the association of bone loss may accelerate with aging. Jones and col- density to fracture risk to estimate and compare leagues reported on the rate of bone loss in 769 the expected benefits of three possible treatment residents of Dubbo, Australia, aged 60 years and scenarios: 1 1) beginning therapy at menopause older, followed between January 1989 and June and continuing for the remainder of life; 2) begin- 1993. They found that bone loss at the hip was al- ning therapy at menopause and stopping at age 65; most 1 percent per year in women, and about 0.8

] OTA’s estimates of the impact of hormonal replacement therapy on fracture risk were calculated in a similar manner. See appendix D. Appendix C Evidence on HRT and Bone Loss 29

percent in men, and that bone loss increased with ued, and these benefits dissipate after cessation of advancing age in both sexes (23).2 therapy. Studies of bone mass in elderly women Recent data on the relation of bone mass to frac- support this prediction. Felson and colleagues ture risk in the elderly show that there continues to measured bone mass in 670 elderly women (mean be a strong relationship of bone mass to hip frac- age 76 years) in the Framingham study cohort to ture risk, even after age 80, so that therapies that determine whether their bone mass was affected slow bone loss will reduce fracture risk in this age by earlier estrogen use (14). They found that, group(3). among the 212 women who had received estrogen In addition, there is evidence from prospective therapy, only those who had taken estrogen for studies that the rate of bone loss immediately after seven or more years had significantly higher bone menopause may not be as great as previously mass than women who had not taken estrogen. thought, and the period of accelerated bone loss The differences in bone mass between long-term may not last as long as was predicted from cross- users and nonusers was greatest among women sectional studies (3). Finally, at age 65, densito- under 75 years old (11.2 percent). Among long- metry can more precisely estimate the subsequent term estrogen users 75 years old or older, bone risk of hip fractures and target treatment more ef- density was only 3.2 percent higher than in fectively (3). women who had never taken estrogen around the There are, however, a number of reasons to time of menopause, and even those who had taken question whether this type of model overestimates estrogen for 10 years had ceased therapy by the the number of fractures avoided by preserving time they were 60 to 65 years old. Of the 24 bone mass in the elderly. Reports are inconclusive women 75 years old or older who had taken estro- regarding how HRT initiated after substantial gen therapy for at least seven years, only two had bone is lost affects fracture incidence (25). (See begun therapy at 60 years of age or later, and only appendix B for discussion.) three were still taking estrogen when their bone In addition, progressive bone loss is associated density was measured. with erosion and perforations in the trabecular HRT has been found to reduce postmenopausal structure, or struts, in cancellous bone (24,33). bone loss regardless of the route of administration These perforations decrease the structural integri- (45,49,50). Lufkin and colleagues compared bone ty of bone out of proportion to the amount of bone loss in 75 osteoporotic women randomly assigned lost. Interventions such as estrogen that reduce to transdermal estrogen patches and progesterone bone resorption are at best capable of thickening tablets or to placebo patches and tablets (32). They the trabecular elements that remain, but are un- found that bone mass was significantly greater in likely to be able to repair perforated trabeculae. those who received the transdermal estrogen patch Finally, such a strategy would not be as effec- compared with those who received placebo. tive in preventing wrist and vertebral fractures, Those women receiving transdermal estrogen had which have a peak incidence earlier in menopause a median annual increase in bone mass of 5.3 per- than hip fractures.3 cent in the lumbar spine, compared to an increase After cessation of therapy, bone loss acceler- of 0.2 percent for women receiving placebo. In a ates to a rate equivalent to that of untreated women two-year clinical trial, Ribot and colleagues ran- at menopause (7,30,43). Thus, one would predict domly assigned 94 postmenopausal women to a that the benefits of HRT on bone mineral density transdermal estrogen patch, a topically applied es- are maintained only so long as therapy is contin- trogen gel, or to a placebo (46). At the end of the

2 They reported no significant bone loss at the spine, which was perhaps due to the presence of spinal arthritis (23). 3 Wrist fractures and vertebral fractures, however, cause relatively little morbidity compared with that incurred by hip fracture. 30 Cost Effectiveness of Screening for Osteoporosis

study, bone mineral density had increased signifi- treatment with the progestin Provera R (medroxy - cantly for the transdermal estrogen patch group progesterone acetate), the estrogen Premarin and the percutaneous estrogen gel group, but not (conjugated equine estrogen), Premarin plus Pro- for the placebo group. There was no significant vera, or placebo (16). The group receiving Prema- difference in the percent increase in bone density rin plus Provera received half the dose of estrogen between the transdermal estrogen patch group and as the Premarin only group and half the dose of the percutaneous estrogen gel group. progestin as the Provera only group. After two The combination of estrogen and progestin, ei- years, bone mass of the spine (composed primari- ther given sequentially or as continuous combined ly of trabecular bone) was maintained in the Pre- therapy has been found as effective as estrogen marin group and the Premarin plus Provera group, alone in reducing postmenopausal bone loss but was lost in the Provera group and the placebo (5,6,7,10,13,19,29,31,34,38,40,41 ,42,45,47,50). group. Bone density of the wrist (composed pri- In fact, a number of studies have demonstrated marily of cortical bone) was lost in all four groups, that progestins alone are effective in reducing but was least in the Premarin only, Provera only, bone loss in postmenopausal women (1,16,26, and Premarin plus Provera groups, and was great- 29). Lindsay and colleagues demonstrated the est in the placebo group. For both cortical bone ability of progestins to reduce bone loss in a clini- and trabecular bone, Premarin alone was better cal trial involving 30 postmenopausal women ran- able to maintain bone mass than Provera alone. domly assigned to the progestin gestranol, the estrogen mestranol, or placebo (29). Women treated CONCLUSIONS with gestranol showed no significant change in Controlled clinical trials have demonstrated that bone mineral density after one year, and women HRT is able to halt bone loss and perhaps increase treated with mestranol showed a nonsignificant bone mass in postmenopausal women. For this increase in bone mineral density. Women assigned analysis, OTA has assumed as a base case that to placebo, however, showed a significant decline HRT maintains bone mass for as long as it is taken. in bone mineral density after one year. There is less information about whether HRT is Abdalla and colleagues showed that progestin able to maintain bone mass over the long term. was able to increase bone mineral density in post- OTA also assumed that initiation of HRT at age 65 menopausal women in a cohort study of the pro- was able to maintain bone mass. Two recent re- gestin norethisterone versus placebo (l). Women views of studies of bone density have concluded assigned to norethisterone were referrals to a that bone mass is lost in long-term HRT users, but Glasgow, Scotland menopause clinic, and con- at a rate that is one-half to one-third that of nonus- trols were patients chosen from placebo groups of ers (3, 12). OTA assumed as a worst case that bone other clinical trials matched to the treatment group mass in HRT users is lost at half the rate of nonus- for age, years since menopause, and initial bone ers. Studies have demonstrated that bone loss is mass. After two years, the bone mass of women halted or reduced only as long as HRT is used. assigned to norethisterone increased by 3.3 per- OTA assumed that, upon cessation of HRT use, cent, whereas the bone mass of the matched con- bone mass is lost at a rate similar to the rate of bone trols declined by 5 percent. The difference in bone loss at menopause. mass between the two groups after two years was Because there are relatively few data on the re- statistically significant (p < 0.002). duction of fracture in long-term estrogen users, Although progestins have been demonstrated OTA used data on the effects of HRT on bone den- to prevent bone loss in postmenopausal women, sity and the association of bone density on fracture they do not appear to be as effective as estrogens in risks to estimate the risks of hip fracture in HRT maintaining bone mass, especially mass of trabec- users at each age. This assumption is discussed in ular bone. Gallagher and colleagues randomly as- more detail in appendix D. signed 81 postmenopausal women to four groups: Appendix C Evidence on HRT and Bone Loss 31

REFERENCES 10. Christiansen, C., and Rodbro, P., “Does Oes- 1. Abdalla, H. I., Hart, D. M., Lindsay, R., et al., triol Add to the Beneficial Effect of Com- “Prevention of Bone Mineral Loss in Postme- bined Hormonal Prophylaxis Against Early nopausal Women by Norethisterone,” Obstet- Postmenopausal Osteoporosis?” British rics & Gynecology 66(6):789-792, 1985. Journal of Obstetrics and Gynecology 2. Aitken, J., Hart, D., and Lindsay, R., “Oestro- 91(5):489-493, 1984. gen Replacement Therapy for Prevention of 11. Civitelli, R., Agnusdei, D., Nardi, P., et al., Osteoporosis After Oophorectomy,” British “Effects of One-Year Treatment with Estro- Medical Journal 3:515-518, 1973. gens on Bone Mass, Intestinal Calcium Ab- 3. Black, D. M., “Why Elderly Women Should sorption, and 25-Hydroxyvitamin D-1 Alpha- be Screened and Treated to Prevent Osteopo- Hydroxylase Reserve in Postmenopausal rosis,” American Journal of Medicine Osteoporosis,” Calcified Tissue International 98( Suppl. 2A):67S-75S, 1995. 42:77-86, 1988. 4. Caniggia, A., Delling, G., Nuti, R., et al., 12. Ettinger, B., and Grady, D., “Maximizing the “Clinical, Biochemical, and Histological Re- Benefit of Estrogen Therapy for Prevention of sults of a Double Blind Trial with 1,25-dihy- Osteoporosis,” Menopause l(l): 19-24, 1994. droxyvitamin D3, Estradiol, and Placebo in 13. Farish, E., Hart, D. M., Gray, C. E., et al., “Ef- Postmenopausal Osteoporosis,” Acta Vita- fects of Treatment with Oestradiol/Levonor- minologica et Enzymologica 6(2):117-130, gestrel on Bone, Lipoproteins and Hormone 1984. Status in Postmenopausal Women,” Clinical 5. Christiansen, C., Christensen, M. S., McNair, Endocrinology 31:607-615, 1989. P., et al., “Prevention of Early Postmenopau- 14. Felson, D.T., Zhang, Y., Hannan, M.T., et al., sal Bone Loss: Controlled Two-Year Study in “The Effect of Postmenopausal Estrogen 315 Normal Females,” European Journal of Therapy on Bone Density in Elderly Clinical Investigation 10:273-279, 1980. Women,” New England Journal of Medicine 6. Christiansen, C., Christensen, M. S., Rodbro, 329(16):1141-1146, 1993. P., et al., “Effect of 1,25-dihydroxy-vitamin 15. Finn Jensen, G., Christiansen, C., and Transbol, I., “Treatment of Post-Menopausal D 3 on Itself or Combined with Hormone Treatment in Preventing Postmenopausal Os- Osteoporosis. A Controlled Trial Comparing teoporosis,” European Journal of Clinical In- Oestrogen/Gestagen, 1,25-dihydroxy-vita- vestigation 11:305-309, 1981a. min D3 and Calcium,” Clinical Endocrinolo- 7. Christiansen, C., Christensen, M. S., and gy 16:515-524, 1982. Transbol, I., “Bone Mass in Postmenopausal 16. Gallagher, J., Kable, W., and Goldgar, D., Women After Withdrawal of Oestrogen/Ges- “Effect of Progestin Therapy on Cortical and tagen Replacement Therapy,” Lancet Trabecular Bone: Comparison with Estro- 1:459-461, 1981b. gen,” American Journal of Medicine 90: 8. Christiansen, C., Mazess, R., Transbol, I., et 171-178, 1991. al., “Factors in Response to Treatment of Ear- 17. Genant, H., Baylink, D., Gallagher, J., et al., ly Postmenopausal Bone Loss,” Calcified “Effect of Estrone Sulfate on Postmenopausal Tissue International 33:575-581, 1981c. Bone Loss,” Obstetrics & Gynecology 9. Christiansen, C., and Riis, B.J., “17-Estra- 76(4):579-584, 1990. diol and Continuous Norethisterone: A 18. Gotfredsen, A., Nilas, L., and Riis, B. J., Unique Treatment for Established Osteoporo- “Bone Changes Occurring Spontaneously sis in Elderly Women,” Journal of Clinical and Caused by Oestrogen in Early Postmeno- Endocrinology and Metabolism 71(4): pausal Women: A Local or Generalised Phe- 836-841, 1990. 32 Cost Effectiveness of Screening for Osteoporosis

nomenon?” British Medical Journal 292: 29. Lindsay, R., Hart, D. M., Purdie, D., et al., 1098-1100, 1986. “Comparative Effects of Oestrogen and Pro- 19. Grey, A. B., Cundy, T. F., and Reid, I. R., gestogen on Bone Loss in Postmenopausal “Continuous Combined Oestrogen/Progestin Women,” Clinical Science and Molecular Therapy is Well Tolerated and Increases Bone Medicine 54: 193-195, 1978a. Density at the Hip and Spine in Post-Meno- 30. Lindsay, R., Hart, D. M., MacLean, A., et al., pausal Osteoporosis,” Clinical Endocrinolo- “Bone Response to Termination of Estro- gy 40:671-677, 1994. gen Treatment,” Lancet 1(8078):1325-1327, 20. Hassager, C., Jensen, S. B., and Christiansen, 1978b. C., “Non-Responders to Hormone Replace- 31. Lindsay, R., and Tohme, J. F., “Estrogen ment Therapy for the Prevention of Postme- Treatment of Patients with Established Post- nopausal Bone Loss: Do They Exist?” Osteo- menopausal Osteoporosis,” Obstetrics & Gy- porosis International 4:36-41, 1994. necology 76(2):290-295, 1990. 21. Horsman, A., Gallagher, J. C., Simpson, M., 32. Lufkin, E. G., Wahner, H. W., O’Fallen, et al., “Prospective Trial of Oestrogen and W.M., et al., “Treatment of Postmenopausal Calcium in Postmenopausal Women,” British Osteoporosis with Transdermal Estrogen,” Medical Journal 2:789-792, 1977. Annals of Internal Medicine 117(1): 1-9, 22. Horsman, A., Jones, M., Francis, R., et al., 1992. “The Effect of Estrogen Dose on Postmeno- 33. Manolagas, S. C., and Jilka, R. L., “Bone Mar- pausal Bone Loss,” New England Journal of row, Cytokines, and Bone Remodeling— Medicine 309(23):1405-1407, 1983. Emerging Insights into the Pathophysiology 23. Jones, G., Nguyen, T., Sambrook, P., et al., of Osteoporosis,” New England Journal of “Progressive Loss of Bone in the Femoral Medicine 332(5):305-311, 1995. Neck in Elderly People: Longitudinal Find- 34. Marslew, U., Overgaard, K., Riis, B.J., et al., ings from the Dubbo Osteoporosis Epide- “Two New Combinations of Estrogen and miology Study,” British Medical Journal Progestogen for Prevention of Postmenopau- 309:691-695, 1994. sal Bone Loss: Long-Term Effects on Bone, 24. Kanis, J. A., “Treatment of Osteoporosis in Calcium, and Lipid Metabolism, Climacteric Elderly Women,” American Journal of Medi- Symptoms, and Bleeding,” Obstetrics & Gy- cine 98( Suppl. 2A):60S-66S, 1995. necology 79(2):202-210, 1992. 25. Langer, R. D., and Barrett-Connor, E., “Ex- 35. Marx, C. W., Dailey, G. E., HI, Cheney, C., et tended Hormone Replacement: Who Should al., “Do Estrogens Improve Bone Mineral Get It, and for How Long?” Geriatrics Density in Osteoporotic Women Over Age 49(12):20-24, 1994. 65?” Journal of Bone and Mineral Research 26. Lee, J. R., “Osteoporosis Reversal with 7(11):1275-1279, 1992. Transdermal Progesterone” (letter), Lancet 36. Meema, S., Bunker, M. L., and Meema, H. E., 336:1327, 1990. “Preventive Effect of Estrogen on Postmeno- 27. Lindsay, R., Hart, D. M., and Clark, D. M., pausal Bone Loss,” Archives of Internal Med- “The Minimum Effective Dose of Estrogen icine 135(11): 1436-1440, 1975. for Prevention of Postmenopausal Bone 37. Moore, M., Bracker, M., Sartoris, D., et al., Loss,” Obstetrics & Gynecology 63(6): “Long-Term Estrogen Replacement Therapy 759-763, 1984. in Postmenopausal Women Sustains Verte- 28. Lindsay, R., Hart, D., Forrest, C., et al., "Pre- bral Bone Mineral Density,” Journal of Bone vention of Spinal Osteoporosis in Oopho- and Mineral Research 5:659-664, 1990. rectomised Women,” Lancet 2(8205): 38. Munk-Jensen, N., Pors Nielsen, S., Obel, 1151-1154, 1980. E. B., et al.. “Reversal of Postmenopausal Ver- Appendix C Evidence on HRT and Bone Loss 33

tebral Bone Loss by Oestrogen and Progesto- ment Therapy on Peripheral Bone Mineral gen: A Double Blind Placebo Controlled Content in Patients with Osteoporotic Spine Study,” British Medical Journal 296(6630): Fractures,’ ’Acta Endocrinologica 123: 14-18, 1150-1152, 1988. 1990. 39. Nachtigall, L.E., Nachtigall, R. H., Nachti- 46. Ribot, C., Tremollieres, F., and Pouilles, gall, R. D., et al., “Estrogen Replacement J. M., “Prevention of Postmenopausal Bone Therapy I: A 10-Year Prospective Study in the Loss by Long-Term Parenteral Administra- Relationship to Osteoporosis,” Obstetrics & tion of 17 Estradiol: Comparison of Percuta- Gynecology 53(3):277-281, 1979. neous and Transdermal Route,” Osteoporo- 40. Nielsen Television Research Services, 1990 sis: Physiological Basis, Assessment, and Report on Television (New York, NY: Nielsen Treatment, H.F. DeLuca and R. Mazess (eds.) Media Research, 1990). (New York, NY: Elsevier Science Publishing 41. Nordin, B. E. C., Horsman, A., Crilly, R., et Co., Inc., 1989). al., “Treatment of Spinal Osteoporosis in 47. Riis, B.J., and Christiansen, C.. “Continuous Postmenopausal Women,” British Medical Oestrogen-Progestogen Treatment and Bone Journal 280:451-455, 1980. Metabolism in Post-Menopausal Women,” 42. Prince, R.L., Smith, M., Dick, I. M., et al., Maturitas 10:51-58, 1988. “Prevention of Postmenopausal Osteoporo- 48. Riis, B., Thomsen, K., and Christiansen, C.. sis: A Comparative Study of Exercise, Cal- “Does Calcium Supplementation Prevent cium Supplementation, and Hormone-Re- Postmenopausal Bone Loss? A Double- placement Therapy,” New England Journal of Blind, Controlled Clinical Study,” New Eng- Medicine 325(17 ):1189-1 195, 1991. land Journal of Medicine 316(4): 173-177. 43. Quigley, M.E.T., Martin, P. L., Burnier, A. M., 1987a. et al., “Estrogen Therapy Arrests Bone Loss 49. Riis, B., Thomsen, K., Strom, V., et al., “The in Elderly Women,” American Journal of Ob- Effect of Percutaneous Estradiol and Natural stetrics and Gynecology 156(6): 1516-1523, Progesterone on Postmenopausal Bone 1987. Loss,” American Journal of Obstetrics and 44. Reeker, R.R., Saville, P. D., and Heaney, R.P., Gynecology 156(1):61-65, 1987b. “The Effects of Estrogens and Calcium Car- 50. Stevenson, J., Cust, M., Gangar, K.. et al., bonate on Bone Loss in Postmenopausal “Effects of Transdermal Versus Oral Hor- Women,” Annals of Internal Medicine 87: mone Replacement Therapy on Bone Density 649-655, 1977. in Spine and Proximal Femur in Postmeno- 45. Resch, H., Pietschmann, P., Krexner, E., et pausal Women,” Lancet 336:265-269, 1990. al., “Effects of One-Year Hormone Replace- Appendix D: Summary of Hip Fracture Prediction Methods D

his appendix describes the assumptions based on earlier work by Black and colleagues on used in OTA’s analysis of the impact of the relationship between bone mass at menopause hormone replacement therapy (HRT) on and lifetime risk of hip fracture (3). In that model, hip fractures.l as in the present one, BMD at any age is predicted TOTA’s model assumes that HRT affects fracture from BMD at menopause. (OTA used age 50 as a risk through its impact on bone mass as measured reasonable proxy for the age at menopause.) The by bone mineral density (BMD). The rationale for predicted BMD at each age is then used to esti- this assumption is twofold. First, the causal rela- mate the risk of fracture at that age. tionship between HRT and bone loss is well-es- The parameters required for such a model fall tablished and precisely estimated, at least in the into two general categories: 1) those related to the short-run. In contrast, the evidence of a direct rela- longitudinal distribution of BMD; and 2) those re- tionship between HRT and fracture rests on stud- lating BMD to the short-term risk of fracture. ies with relatively weak designs that do not lend Most of the data available to estimate these rela- themselves readily to precise estimates of effect tionships are based on studies of white women, size. (See appendices B and C.) Second, the rela- the group at highest risk of osteoporosis and the tionship between BMD measured at each age and only ethnic-sex group for whom data are available the risk of hip fracture has been quantified in some for estimations of sufficient precision for model- recently reported studies. ing. Where data on racial or ethnic groups or sexes other than white women are available, however, REQUIREMENTS OF THE OTA HIP their findings are described in this appendix. FRACTURE PREDICTION MODEL OTA’s model predicts the probability of hip frac- MEASURING BONE MASS ture at every age between 50 and 90 as a function Different technologies are available for measuring of an individual’s BMD at age 50. The model is bone mass at different sites in the body. How and

1 This appendix is based on a contract report prepared in 1992 for OTA by Dennis Black (1). The data in that report reflected information available in 1992 .That report describes methods for predicting wrist, spine, and all fractures as well as hip fractures. 35 36 Cost Effectiveness of Screening for Osteoporosis

where bone mass is measured can affect its predic- PREDICTING BONE MASS OVER TIME tive power for hip fracture.2 OTA’s model is based The OTA model assumes that bone mass at any on bone mineral density measured at the proximal age follows a normal distribution with an age-spe- radius (the lower forearm above the wrist) using cific mean and standard deviation. The evidence single photon absorptiometry as the measurement to support this assumption is summarized in the technology. The primary reason for this choice is next section. If BMD at any age is normally dis- that the proximal radius is the only site for which tributed over the cohort of individuals in the age there are sufficient data from a number of sources category, then the joint probability distribution of to make reasonable estimates of all parameters of BMD at any two ages can be assumed to be bivari- the model. In particular, the proximal radius is the ate normal. This implies that a woman’s BMD at only site for which the longitudinal pattern of a given age, t, is related to her BMD at the pre- bone mass measurement over time has been char- vious age, t -1, according to the following formu- acterized. 3 Because OTA’s model requires consis- las: tent data on both changes in bone mass and the CMBMD = + ( BMD - relationship between bone mass and fracture, no t t t t-1 t - 1 other site is feasible for modeling at present. 2 2 1/2 CSDVt = t (1 - t )) There has been some discussion in the literature BMD = Z * CSDV + CMBMD about whether bone mineral content (gm/cm) t t t (BMC) or density (actually areal density, gm/cm2) where: is a better predictor of fracture risk. For predicting BMD = a woman’s BMD at age t; hip fracture, an analysis of data from the Study of = the mean BMD in the population of women m t Osteoporotic Fractures (SOF) showed that BMC at age t; was approximately the same as bone mineral den- s t= the standard deviation of BMD in the popula- sity in predicting fracture (5). For predicting all tion of women at age t; fractures, other analyses performed on the SOF r = the correlation between a woman’s BMD at data have shown similar results. Although most age t and her BMD at age t–1. studies have reported results in terms of BMD, CBMDt = the conditional mean of the probability some have reported BMC. This appendix treats distribution of BMD values at age t in women with these results as interchangeable, although the pre- BMD value at age t–l of BMD t-l; dictive model is measured in terms of BMD. CSDVt = conditional standard deviation at age t;

2 Although a paper from the Hawaii Osteoporosis Center (29) has suggested that bone mass measurements taken at the calcaneus (heel) predicted all fractures better than did measurements taken at the radius (a bone of the lower part of the arm) or in the spine, the Study of Osteopo- rotic Fractures (the only other study which has measured bone mass at the calcaneus) has shown a relationship of approximately equal magnitude between bone mass and fracture risk at all sites (radius, calcaneus, spine and hip) for all fractures and for wrist fractures (3). For hip fractures, the three appendicular sites (proximal radius, distal radius and calcaneus) have been shown to be approximately equal as predictors (5) although recent data have suggested that bone mass at the proximal femur (thighbone) is abetter predictor of hip fracture risk than bone mass at the other sites (2,6). Unfortunately, no data on the longitudinal distribution of bone mass at the hip or the long-run predictive accuracy of any densitometry method are yet available. 3 After the proximal radius, the bone mass site studied most frequently is the spine. At present, however, the information available to estimate the parameters of the model are insufficient for three reasons. First, most studies of bone mass at the spine are either small, have a very wide age range, or have been performed on samples of women who are unrepresentative of the general population of women. Second, bone mass at the spine is measured by several techniques, including quantitative computed tomography, dual photon absorptiometry, and dual x-ray absorptiometry, each of which might show a unique longitudinal pattern or relationship to fracture risk. Third, as women age and develop anatomical abnormalities (e.g., vertebral deformities, osteophytes, etc.,) the spine presents special difficulties as a site for bone mass measurement. Appendix D Summary of Hip Fracture Prediction Methods 37

1 0.415 0.412 0.372 0.396 5 0.484 0.484 0.438 0,441 10 0.521 0.522 0.473 0.476 25 0.582 0.584 0.532 0.529 50 0.651 0.650 0.598 0,596 75 0.719 0.719 0.663 0,659 90 0.781 0.780 0.722 0,722

Mean (gm/cm2) 0.6507 0.5978 Standard deviation 0.1015 0.09705 a Predicted values based on a normal distribution with the observed mean and standard deviation. SOURCE: D. M. Black, “Cost Effectiveness of Screening for Osteoporosis’ Review of Bone Mineral Density and Fracture Parameters Required for Model,” University of California, San Francisco, CA, unpublished OTA contract report, Nov. 17, 1992

Z = a random number drawn from the standard observed percentiles for bone mass at the proxi- normal distribution. mal radius to the predicted percentiles based on a normal distribution. There is close agreement Thus, knowledge of the mean and standard devi- between the observed and predicted values, in- ation of the BMD distribution at each age, and the dicating that the normal distribution provides coefficient of correlation between the two dis- an excellent approximation to the observed, tributions, permits the generation of a BMD tra- empirical distribution (1). jectory for an individual woman over her lifetime. ■ University of Indiana: A study of bone mass in This section reviews the evidence on the fol- 583 women showed that the fit of bone mass lowing aspects of the BMD prediction formula data to a normal distribution was excellent (19). given above: The investigators of this study have reported that they have found no evidence of significant ■ Age-specific distribution of BMDs skewing or other nonnormality in their cross- ■ Correlation between BMD values at successive ages sectional data (15). ■ University of Iowa: In a cross-sectional study of bone mass in 217 Caucasian women, bone ❚ Age-Specific Distribution of BMDs mass variables were found to have a normal The age-specific distribution of BMDs is defined distribution (24). by the general shape of the distribution (i.e., whether it is a normal, or bell-shaped curve, or de- These findings and the lack of any report suggest- fined by some other general form) and, if it is a ing that bone mass departs from a normal distribu- normal distribution, its mean and standard devi- tion, strongly suggest that at any age the ation. distribution of bone mass across women is normal. Shape of the BMD Distribution Age-Specific Means Three large studies of bone mass are available to Ideally, bone loss could be estimated directly from assess the shape of the BMD distribution. longitudinal data on cohorts of women followed ■ Study of Osteoporotic Fractures (SOF): The for long periods of time. However, few such stud- unpublished analysis in table D-1 compares the ies with sufficiently large numbers of subjects are 38 Cost Effectiveness of Screening for Osteoporosis

available. In the absence of such longitudinal data, 20 measurements) over followup periods ranging bone loss can be estimated from age-specific from six weeks to seven years (mean = 4 years). means derived from cross-sectional data. OTA The samples have been used for a number of dif- used a combination of longitudinal and cross-sec- ferent analyses. The exact participants and mea- tional studies to estimate change in bone mass surements used have differed in the various with age. reports of the study depending on the research It is possible to establish age-specific means question being addressed. from cross-sectional studies of bone mass in pop- One analysis of these data compared longitudi- ulation-based samples of individuals, but there are nal with cross-sectional estimates of mean BMC a number of potential problems in using cross-sec- in post-menopausal women (19). This analysis tional data to estimate longitudinal changes in showed that the cross-sectional results agreed bone mass.4 First, the sample maybe biased. Se- closely with longitudinal results. The results cond, the sample must be large enough in each age showed a quadratic relationship between BMC category to allow for sufficient precision. Third, and age which was essentially linear in the age cross-sectionally estimated changes in bone mass range 50 to 70. The average rate of bone loss de- may differ from those estimated from longitudinal creased after about age 70, and there is a sugges- studies if there are cohort effects, such as nutrition of an increase in bone mass after age 70. The tional factors or medication use patterns that vary actual rates of loss in the data are not useful for the with age. purposes of the OTA study, because they were ad- Although many cross-sectional studies address justed for body weight without reporting enough the relationship of bone mass to age, most of these information to calculate overall population data are not very informative because they are means. from small studies without population-based A more recent analysis of the same data looked samples. Two studies described in this section are in detail at bone loss in the period zero to five years exceptions. after menopause and five to 10 years after meno- Data from six separate studies of bone mass pause (18). The large number of repeat measure- formed the basis for estimates of age-specific ments over time gives very precise estimates of mean BMD. Each is described in detail below. bone loss. For the period zero to five years after menopause, a total of 89 women were available University of Indiana for analysis with an average of 11 measurements during the five-year period. These women showed A total of 268 women were studied longitudinally an average loss of about 1.6 percent per year over and 583 were studied cross-sectionally. They were the five years. For the period five to 10 years after recruited as two distinct samples. The cross-sec- menopause, a total of 47 women were used with tional sample was younger (under age 65) and an average of eight measurements each during the consisted primarily of gynecology clinic patients five years. These women showed an average loss and employees of the Indiana University Medical of about 1.2 percent per year over the five years. Center. The longitudinal sample consisted of a group of older subjects (over age 55, mostly over age 65) who were residents of a retirement home. University of Iowa The older women had repeated bone mass mea- A sample of 217 Caucasian women from a rural surements (between three and 45, with a mean of community in Iowa between the ages of 22 and 80

4 Several studies have compared cross-sectional to longitudinal data sets for estimating bone 10SS with age (7,19). Only Davis and colleagues found the two approaches to lead to different results, although the methodology in their paper is difficult to interpret. Also, the data set which Davis used (a cohort of Japanese American women) may have cohort effects not present in other data sets. Appendix D Summary of Hip Fracture Prediction Methods 39

Age at first Number of Percent loss per BMC at the proximal radius measurement women year Mean Percent loss 45-49 19 1.3% Age N (gm/cm) per year 50-54 36 2.O% 43-49 11 .849 — 55-59 60 1.2% 50-54 38 ,797 1.2% 60-64 55 1.2% 55-59 196 ,790 1.2% 65-69 60 1.3% 60-64 411 ,745 1.1% 70-74 43 1.2% 65-69 306 .687 1.2% 75-79 53 1.3% 70-74 108 .669 0.5% 75-79 24 .577 2.7% SOURCE: M. Sowers, M. Clark, B. HoIIis, et al., “Prospective Study of Radial Bone Mineral Density in a Geographically Defined Population KEY BMC = bone mineral content of Postmenopausal Caucasian Women, ” Calcified Tissue Interna- SOURCE K Yano, R.D. Wasnich, J.M. Bogel, et al , “Bone Mineral tional 48:232-239, 1991. Measurements Among Middle-Aged and Elderly Japanese Resi- dents in Hawaii,” American Journal of Epidemiology 119751-764, had bone mass measured at the radius in 1984, and 1984. 181 of them had repeat measurements made five years later in 1989 (24,25). The strength of these- bone mass measurements. One analysis examined data is that they were collected longitudinally over change in bone mass with age among post-meno- a long period of time (five years). However, the pausal women who did not use estrogen (7). These precision of estimates based on the data is limited loss rates were adjusted for height, weight, and due to the small numbers of participants. The av- bone width. Longitudinal analyses (n = 636 erage annual bone loss for women who were post- women, mean length of followup = 3.2 years) in menopausal at the time of followup is given in the same paper showed that the rate of loss was table D-2. about 1.5 percent per year at age 55 and declined For ages 50 and over, the confidence intervals to about 0.8 percent at age 75. Cross-sectional around the mean percent loss are approximately analyses of bone density in 677 women showed a plus or minus 0.3 to 0.4 percent. The loss rates in decrease in mean bone mass of approximately 1 this study are approximately the same as other percent for each year of age for women around the studies. There is an approximate doubling of the age of 55. The mean decrease by age increased to rate of loss during the five years after menopause. about 1.25 percent for each year of age for women around 75 years old. University of Copenhagen The results of another cross-sectional analysis 121 women who were six months to three years on the same sample, which did not exclude estro- post-menopausal and who were 45 to 54 in 1977 gen users, are shown in table D-3 (30). Through had BMD measured in the forearm. Their BMD the early post-menopausal years, the results are measurements were repeated 12 years later in essentially constant at about 1.2 percent per year. 1989 (14). The mean loss averaged 1.7 percent per The results for age 70 to 74 are at odds with the re- year over the 12 years. mainder of the data and suggest either a typo- graphic error (e.g., .699 should be .629) or Hawaii Osteoporosis Center imprecision due to small sample size. A cohort of 1,098 Japanese-American women, all An important caveat in interpreting analyses of post-menopausal, was established in 1981. This these data is that the sample is drawn from a very cohort has been extensively followed with repeat special population (Japanese-American women 40 Cost Effectiveness of Screening for Osteoporosis

living in Hawaii) which may not reflect loss rates in a larger population. However, the general pattern of change in bone density with age is helpful in confirming the pattern found in other data sets. Bone mass at the proximal Study of Osteoporotic Fractures radius The largest cross-sectional study (9,704 women Mean Percent loss Age N 2 over age 65) of bone mass, the SOF should be rea- (gm/cm ) per year sonably representative of healthy white women 65-69 1,864 .650 — over age 65 years in the U.S. (5). Data are avail- 70-74 1,507 .620 0.9% 75-79 907 .598 0.7% able for bone mass measured by SPA in the proxi- 80-85 537 .566 1.O% mal radius, distal radius and calcaneus, but only 85-89 140 .541 0.9% in women over the age of 65. SOURCE:S.R. Cummings, D.M. Black, M C. Nevitt, et al , “Appendic- Steiger and colleagues recently reported an ular Bone Density and Age Predict Hip Fracture in Women, ” Journa/ analysis of cross-sectional bone loss in SOF (26). of the American Medical Association 263:665-668, 1990 However, that paper does not exclude current estrogen users. The data shown in table D-4 are the sistent given the inherent imprecision due to lim- same as those in the Steiger paper but exclude es- ited sample size. Some of the discrepancies trogen users. among the studies may also be due to differences Framingham Osteoporosis Study in the study population, methods of analysis, or The investigators in the Framingham study per- differences in measurement technique. formed cross-sectional analysis of bone mass at The only longitudinal study in the age range of various sites, including the proximal radius in 708 65 and over is the study from Iowa which showed women over age 68 years. At the time of prepara- a mean loss of about 1.2 percent per year. Two tion of this report, no data had yet been published. large cross-sectional studies are available of However, preliminary results have suggested a women over age 65 years. SOF is much larger than constant loss rate of about 0.9 percent per year any other study and shows a constant rate of loss from ages 68 to 90 (1 ,13). after 65 of 0.8 percent per year. The results from the Framingham study are consistent with those OTA’s Estimate from SOF showing an average loss of about 0.9 Qualitatively, most studies have shown a slightly percent per year after age 65. higher rate of bone loss at the appendicular sites Based on the results of these studies, OTA de- just after menopause, which slows after about age veloped a base-case set of assumptions about the 55 or 60. One interesting consistency among the rate of change in mean bone mass of a population data presented above is that the acceleration in of women as they age. These assumptions are bone loss just after menopause at these sites is shown in table D-5. Alternative assumptions re- only slight. flecting reasonable upper and lower bounds on the For the age range of 50 to 65, the various stud- bone loss rate are also shown in the table. ies provide consistent results. The longitudinal In addition to the percentage of bone loss in data from Copenhagen suggest an average rate of each year, the OTA model requires an estimate of loss of 1.7 percent per year for the 12 years after mean BMD at each age. Although all the studies menopause. The Iowa data show a 1.6 percent loss described above are consistent in their estimates for the five years after menopause with a 1.2 per- of loss rates, recorded bone density levels vary cent loss for the next five years. The estimated with each densitometer. Consequently, it is not rates of loss from these three studies are quite con- possible to pool mean values from various Appendix D Summary of Hip Fracture Prediction Methods 41

Annual rate of loss Assumption set Age interval Base case Slow loss Fast loss Age Base slow loss Fast loss

50-54 1.8% 1.6% 2.0% 50 0.814 0.796 0.844 55-59 1.3% 1.2% 1.5% 51 0.799 0.783 0.827 60-64 1.0% 0.9% 1.2% 52 0.785 0.770 0.811 >65 0.9% 0.8% 1.2% 53 0.771 0.758 0.794 54 0.757 0.746 0.778 SOURCE: Office of Technology Assessment, 1995. 55 0.743 0.734 0.763 56 0.733 0.725 0.751 sources. OTA used the estimated mean BMD val- 57 0.724 0.717 0.740 2 58 0.714 0.708 0.729 ue from SOF for ages 65 to 69 (0.650 gm/cm ) as 59 0.705 0.699 0.718 an anchor value for the age-specific BMD levels. 60 0.696 0.691 0.707 The means at the other ages were calculated from 61 0.689 0.685 0.699 this value at age 67 using the loss rates given in 62 0.682 0.679 0.690 table D-5. The derived age-specific means are 63 0.675 0.673 0.682 64 0.669 0.667 0.674 shown in table D-6 for the base assumptions and 65 0.662 0.661 0.666 two alternative assumptions. Under the base as- 66 0.656 0.655 0.658 sumption, there is an overall loss of 35 percent be- 67 0.650 0.650 0.650 tween ages 50 and 90. For the slow loss 68 0.644 0.645 0.642 69 0.638 0.640 0.634 assumption set, there is an overall loss of 32 per- 70 0.633 0.635 0.627 cent and under the fast loss assumption set, the 71 0.627 0.629 0.619 overall loss is 42 percent. 72 0.621 0.624 0.612 73 0.616 0.619 0.605 74 0.610 0.614 0.597 Age Specific Standard Deviations 75 0.605 0.610 0.590 The requirements for estimating the standard 76 0.599 0.605 0.583 deviations of the distribution of BMDs are similar 77 0.594 0.600 0.576 to estimating their means: a large, randomly cho- 78 0.588 0.595 0.569 sen sample in which estrogen users have been ex- 79 0.583 0.590 0.562 80 0.578 0.586 0.556 cluded. Longitudinal data are not required, 81 0.573 0.581 0.549 however. Again, the problem of scaling of bone 82 0.568 0.576 0.542 density values taken from different densitometers 83 0.562 0.572 0.536 makes comparisons across studies difficult, and 84 0.557 0.567 0.529 0.552 0.563 0.523 the values of a given study must be used as an an- 85 86 0.547 0.558 0.517 chor. An important question in analyzing the data 87 0.542 0.554 0.511 available on standard deviation is whether it va- 88 0.538 0.549 0.504 ries with age. 89 0.533 0.545 0.498 Because of its size and relatively representative 90 0.528 0.540 0.492 sample, the SOF study provides the best estimates SOURCE: Office of Technology Assessment, 1995; D.M. Black, “Cost of standard deviation for women overage 65. Un- Effectiveness of Screening for Osteoporosis: Review of Bone Mineral published data from that study for women who Density and Fracture Parameters Required for Model, ” University of California, San Francisco, CA, unpublished OTA contract report, have never used estrogen are shown in table D-7. NOV. 17, 1992. Although these data suggest that, at least for women over 65 years of age, the standard devi- variation with age. Data from the Indiana Univer- ation is fairly constant, other studies suggest some sity sample of 268 post-menopausal women sug- 42 Cost Effectiveness of Screening for Osteoporosis

the standard deviation of bone mass at the proximal radius is 0.10 gm/cm2 and does not change with age in the age range 50 to 90 years.

❚ Correlation Between Values of Bone Mass at Two Ages The model requires an estimate of the correlation 65-69 1,864 .102 between bone mass at the age at which BMD 70-74 1,507 .100 screening takes place and at later ages. For exam- 75-79 907 .097 80-84 537 .096 ple, if screening for bone mass occurs at age 50, 85+ 140 .095 the model requires the correlation between bone mass at age 50 and bone mass at ages 51, 52, etc. KEY: BMD = bone mineral density; SOF = Study of Osteoporotic Fractures, The correlation required is the correlation be- SOURCE: D. M. Black, “Cost Effectiveness of Screening for Osteopo- tween the true values of bone mass in successive rosis: Review of Bone Mineral Sensity and Fracture Parameters Re- years, not the measured values, because it is the quired for Model, ” University of California, San Francisco, CA, un- true values that predict fracture. For long-term published OTA contract report, Nov. 17, 1992 studies (e.g., at least five to 10 years), the correlation between the true values will be about the same gested that there is an increase in standard devi- as that between the measured values. However, in ation with age, but no estimates of standard devi- studies of shorter duration, measurement error ation were reported (19). Data from the University plays a larger role artificially deflating the correla- of Copenhagen study, on the other hand, showed tion. a decrease in the standard deviation between the The accuracy of the estimate of this model pa- first measurement (age 45 to 54) and the second rameter is important, because changes in the esti- (age 57 to 66), by about 0.5 percent per year. The mates would have large effects on the resulting Framingham cross-sectional data also showed a fracture rates. Fortunately, sufficient data exist decrease—about 0.6 percent per year-in women (see below) to restrict its possible values, and over age 68. The University of Iowa 1983 cross- within this range its effect on outcomes is only sectional sample showed age-specific standard moderate. deviations of BMD that varied from a high of To estimate the long-term correlation, longitu- 0.119 gm/cm2 (age 70 to 74) to a low of 0.081 gm/ dinal data must be collected over as long a period cm2 (age 65 to 69). There was no clear trend with of time as possible. For example, to estimate the age, although the precision of the estimates is lim- correlation between bone mass at age 50 and age ited by the small numbers within each age group. 65, 15 years of followup data are needed on a co- Estimates of standard deviations are less pre- hort who were age 50 at the initial measurements. cise than estimates of means; it is therefore diffi- For the correlation between BMD at age 50 and cult to conclude from these data whether there is a BMD at age 80, a 30-year followup period is nec- real decrease in the variation of bone mass with essary. The ideal data set would have bone mass age. In addition, the value of the standard devi- measured on a large random sample of women ation of bone mass depends on the technique used from the age of 50 to 90. Clearly, such data do not to measure bone mass as well as the population (yet) exist. from which the sample was drawn. The data are Three studies have reported the correlation be- most consistent with a slight decrease of standard tween bone mass measurement at widely sepa- deviation with age. However, the small decrease rated intervals. The University of Indiana analysis suggested would have a negligible effect on any of post-menopausal women estimated the correla- results of the model. Therefore, OTA assumed that tion between bone mass measured within two Appendix D Summary of Hip Fracture Prediction Methods 43

years of age 50 with bone mass measured about 10 years later based on the experience of 34 women (18). The data showed a correlation of about 0.81 between measurements taken at ages 50 and 60 and a correlation of about 0.7 between measure- Improved tracking ments made at ages 50 and 70. For women over age 55, the University of Iowa 0.8 study found that the correlation between the two (A) measurements taken five years apart was greater than 0.9 (24). However, the exact 5-year correla- (B) tion was not given. Finally, in the University of Copenhagen study (C) of 121 post-menopausal women aged 45 to 54 at I entry, the correlation between the first measure- I ment and the second taken 12 years later was 0.8 (95 percent CI: 0.7 to 0.9) (14). The two long-term data sets on early post- 50 55 60 65 70 75 80 85 menopausal women agree closely in showing a Age 10-year correlation of about 0.8. OTA used this value as the base assumption about the longitudi- SOURCE D.M. Black, “Cost Effectiveness of Screening forOsteoporo- SIS: Review of Bone Mineral Density and Fracture Parameters Required nal correlation from age 50 to age 60. For ages 50 for Model,” University of California, San Francisco, CA, unpublished to 70, we used the estimate of 0.7 from the Univer- OTA contract report, Nov 17, 1992 sity of Indiana. Beyond age 70, for the base assumption, a quadratic function was fit under the assumption that the degradation in correlation base-case pattern. This might occur if a more pre- continues after age 70 in the same pattern as before cise measurement of bone mass was made. age 70. There is little published data specifically ad- Alternative assumptions are possible. Figure dressing the correlation between bone mass at age D-1 shows three alternative correlation trajecto- 65 and later ages. However, the correlation of ries. Under the base-case assumptions, the extrap- bone mass between age 65 and subsequent ages olation of correlation beyond age 70 (pattern B in will almost certainly be better than correlations figure D-1) continues to decrease along the same with age 50, because a relatively high rate of peri- quadratic pattern as before age 70. As a woman menopausal bone loss after age 50 adds greater ages and becomes less active and more ill, howev- variability to predicting later values. Thus, the er, a second acceleration in bone loss may occur. correlation between bone mass at age 65 and ages Since this increased bone loss would be associated above 65 will almost certainly be higher than the with factors that could not be predicted from bone correlation between bone mass at age 50 and later mass at age 50, a decreased correlation between ages. bone mass at age 50 and bone mass beyond age 65 Because the OTA model uses year-to-year cor- would result. Pattern C represents the decreased relation estimates, the estimates projected in fig- correlation that might be associated with in- ure D-1 probably represent too steep a loss of creased bone loss associated with severe immobil- correlation in later years. Consequently, we re- ity and/or illness or extreme old age. vised the correlation pattern after age 65 to ac- Another correlation trajectory (pattern A in fig- count for the higher correlation pattern at older ure D-1) maps a correlation of bone mass at each ages. Figure D-2 contains the results. Pattern D age with bone mass at age 50 that is higher than the represents a correlation after age 65 that is slightly 44 Cost Effectiveness of Screening for Osteoporosis

then the correlations between any two points can be found simply by multiplying the correlations in between. For example, the correlation between age 50 and 60 is 0.9 x 0.9 or 0.81. Under the autoregressive model, the long-term pattern closely approximates pattern C. Similar sets of short-term correlations could be developed for the other long- term correlation patterns above. Reduced tracking I .7 RELATIONSHIP OF BONE MASS TO HIP FRACTURE RISK At any age, the OTA model must predict the probability of hip fracture as a function of the woman current BMD. Following work of Black and colleagues, OTA assumed a logistic relationship be-

0.2 ! I I I I 1 [ tween BMD and hip fracture risk (4). A logistic is 65 70 75 80 85 given by the following formula: Age 1 SOURCE: D. M. Black, “Cost Effectiveness of Screening for Osteoporo- P=

SiS: Review of Bone Mineral Density and Fracture Parameters Required 1 + e t+ t x for Model,” University of California, San Francisco, CA, unpublished OTA contract report, Nov. 17, 1992, where: P is the probability of hip fracture at a given age; higher than that assumed after age 50 (pattern B in figure D-l). This pattern was used as the base-case a is a constant term that varies with age; assumption. Pattern E is analogous to the data b is a term that varies with BMD, but not with age; from Hui for peri-menopausal women showing a and correlation of 0.90 after five years (at age 70), a x is the individual’s BMD at the age in question. correlation of 0.80 after 10 years, etc. This pattern probably represents a lower limit on the correla- When the risk of fracture is less than about 10 tion of bone mass after age 65 (18). percent (as is the case for all hip fracture risks con- The OTA model requires correlations between sidered in this appendix), the logistic relationship bone mass at each age and bone mass at the next between bone mass and risk is essentially linear. highest age. The patterns above include only the Therefore, data fitted to any other functional form correlation between age 50, or age 65, and subse- that is similarly linear would yield essentially the quent ages. However, the cost-effectiveness mod- same results as those obtained from fitting data to el requires the correlation between, for example, estimate the parameters (a,b ) of the logistic mod- bone mass at ages 60 and 65 and between ages 80 el. However, if a nonlinear relationship (e.g., and 85. We required a pattern of short-term cor- threshold model) were the true relational form be- relations that approximates the long-term correla- tween BMD and hip fracture, the logistic assumptions shown in the figures above. tion would yield substantially erroneous results. It To approximate the long-term pattern in the is therefore important to establish the validity of base-case (pattern C in figure D-1) we assumed the logistic (or linear) relationship. that the correlation between bone mass at any age There is only one source of data (SOF) that has and subsequent bone mass five years in the future published data relating bone mass to risk of hip is 0.90 forages 50 to 60 and 0.95 for ages over 60. fracture (5). The results of that analysis suggest If the correlations follow an autoregressive model, that the relationship of bone mass to hip fracture Appendix D Summary of Hip Fracture Prediction Methods 45

RRa Age group N Fractures SD 95% CI

65-74 6,896 934 0.10 1.36 (1 .27,1 .45) 75+ 2,441 470 0.10 1.33 (1 .21,1 .47) a Relative hazard per standard deviation decrease in bone mass. SOURCE: D. M. Black, “Cost Effectiveness of Screening for Osteoporosis’ Review of Bone Mineral Sensity and Fracture Parameters Required for Model,” University of California, San Francisco, CA, unpublished OTA contract report, Nov. 17, 1992 risk follows a linear pattern and that the logistic cal significance. A recent analysis of SOF data of model is therefore a consistent description of the radical BMD prediction of hip fracture found no true relationship. A statistical test for the good- difference between women over 80 years of age ness of fit of the data to a logistic model did not and those under 80 years of age (23). Therefore, reject the null hypothesis that the relationship is OTA assumed that the relative risk relating to logistic. bone mass and hip fracture risk is constant across The relationship between bone mass and frac- all ages in the model. ture probability is usually estimated in terms of a standardized relative risk: the ratio of risk of a per- ❚ Relative Risk of Bone Mass son whose bone mass is one standard deviation and Hip Fracture below the mean to the risk of a person whose bone Several sources of data are available on the short- mass is at the mean. run risk of hip fracture as a function of bone mass. A critical assumption of the OTA model is that SOF published data using bone mass at three the relative risk of fracture is constant across age sites to predict hip fracture in the sample of 9,704 groups (or, alternatively, that b is constant across women (5). The average followup was 1.7 years. all age groups). For example, if the relative risk of The standardized age-adjusted relative risk for fracture of a 70-year-old woman whose BMD lies BMD (gm/cm2) at the proximal radius was 1.4 1 standard deviation below the mean BMD of (1.1 to 1.9). Analysis using BMC as the measure 70-year-old women is 3, then the relative risk of a found slightly lower relative risks. 50-year-old woman with BMD at 1 standard devi- The University of Indiana reported on a total of ation below mean BMD of 50-year-old women is 23 first hip fractures in 135 residents of a retire- also 3. This assumption is equivalent to stating ment home (17). Bone mass (gm/cm2) at the prox- that there is no interaction between bone mass and imal radius was used as the predictor of hip age in predicting hip fractures. Virtually all pub- fracture. The relative risk (95 percent CI) was 1.9 lished analyses in this field have included this as- (1.3, 2.8) per 0.1 gin/cm of BMC (approximately sumption (e.g., 3,5,1 1,12,17,22,28). 1 SD). Age was not a significant predictor of hip Fortunately, data available to test the interac- fracture after adjustment for BMC. tion between bone mass and age suggest that the An analysis of the data on 1,076 women in Mal- assumption is valid. Table D-8 shows data pro- mo, Sweden, found that after adjusting forage, the vided by the SOF on the relationship of BMD relative risk was 1.8 (95 percent CI: 1.3 to 2.4) for (measured in the proximal radius) to hip fracture BMC of the mid-radius. However, since these risk in ages 65 to 75 compared with ages 75 and findings were not age-adjusted, they overestimate over. There is a very slight suggestion of decreas- the age-specific relationship of bone mass to risk ing strength in the relationship of bone mass to hip and therefore may not be of direct relevance to this fracture risk with increasing age. However, the study. (20,2 1). A recent analysis of a cohort of 304 large overlap in the two confidence intervals women in Rochester, Minnesota, who were fol- shows that the differences do not approach statisti- 46 Cost Effectiveness of Screening for Osteoporosis

Annual incidence of hip fracture Relative risk per standard Age (per 100.O00) deviation 1.65 Corresponding value of 50-54 69.5 beta (6) -5.0078 55-59 135.4 60-64 169.6 Values of alpha (a) 314.3 65-69 Bone loss = base case 70-74 493.5 assumption 75-79 1,033.2 80-84 1,669.3 Age Value of a 85-89 2,552.5 50-54 -3.465 SOURCE M.E. Farmer, L R., White, J, A., Brody, et al., “Race and Sex 55-59 -3.10272 Differences in Hip Fracture Incidence, ” American Journal of Public 60-64 -3.0874 Health 74:1374-1380, 1984. 65-69 -2.62888 70-74 -2.32062 75-79 -1,70997 lowed for an average of eight years, found a stan- 80-84 -1.35219 dardized relative risk of hip fracture of 2.7 85-89 -1.04624 (95-percent CI: 1.5 to 5.0) (21). Bone loss = slow Black and colleagues performed a meta-analysis to calculate a pooled estimate and confidence Age interval for the relative risk of hip fracture forage- 50-54 -0.2460 adjusted levels of bone mass in the radius, the only 55-59 -0.2283 measurement site that was common to all three 60-64 -0.2870 65-69 -0.2533 studies (4). Standard meta-analytic methods per- 70-74 -0.3520 mit estimation of a pooled treatment effect and 75-79 -0.3655 confidence intervals for data from numerous clini- 80-84 -0.1783 cal trials (10). This pooled relative risk is calcu- 85-89 -0.3261 lated as the weighted mean of the individual Bone loss = fast standardized relative risks from each study, using Age the squares of the inverse of the standard error of 50-54 -3.3348 the relative risk as the weights. The resulting com- 55-59 -3.0226 bined relative risk was 1.65, with a 95 percent con- 60-64 -3.0473 fidence interval of between 1.4 and 2.0. On the 65-69 -2.6289 basis of this meta-analysis, OTA used a value of 70-74 -2.3657 1.65 as the base assumption for relative risk of hip 75-79 -1.8001 80-84 -1.4824 fracture. 85-89 -1.2015

SOURCE: Off Ice of Technology Assessment, 1995, D M Black, ❚ Calculation of the Constant Term “Cost Effectiveness of Screening for Osteoporosis Review of Bone for the Logistic Model Mineral Density and Fracture Parameters Required for Model, ” Uni- versity of California, San Francisco, CA, unpublished OTA contract We have assumed that the relative risk relating report, Nov. 17, 1992. bone mass to fracture is the same for all age groups. The absolute risk of fracture does increase Under the assumption of an age-invariant rela- with age, however. The constant term in the logis- tive risk, the constant term can be estimated using tic model, a, must be estimated for each age to ad- age-specific hip fracture incidence data (i.e., no just the absolute risk for differences in age. age-specific data on the relationship of bone mass Appendix D Summary of Hip Fracture Prediction Methods 47

to fracture is required). This method has been de- 5. Cummings, S.R., Black, D. M., Nevitt, M. C., scribed elsewhere (4). et al., “Appendicular Bone Density and Age Briefly, the overall age-specific incidence of Predict Hip Fracture in Women,” Journal of fracture (P(Ft)) is the mean of the bone-mass-spe- the American Medical Association 263: cific incidence (P(Ft|BMt)) weighted by the age- 665-668, 1990.

specific distribution of bone mass f(BMt)) or: 6. Cummings, S.R., Black, D. M., Nevitt, M. C., et al., “Bone Density at Various Sites for Pre- diction of Hip Fractures,” Lance?, 341:72-75, January 1993. If we have data on the age-specific incidence of 7. Davis, J.W., Ross, P. D., Wasnich, R. D., et al., fracture and on the age-specific distribution of “Comparison of Cross-Sectional and Longi- bone mass (both of which are readily available tudinal Measurements of Age-Related from cross-sectional studies) and we know the rel- Changes in Bone Mineral Content,” Journal ative risk parameter (b) for the logistic model, the of Bone Mineral Research 4:351-357, 1989. only unknown in the logistic function equation is 8. Farmer, M.E., White, L.R., Brody, J. A., et al., the parameter a . t “Race and Sex Differences in Hip Fracture In- OTA used data from the National Hospital Dis- cidence,” American Journal of Public Health charge Survey to estimate the age-specific inci- 74:1374-1380, 1984. dence of hip fracture (8). Table D-9 shows those 9. Fisher, E., Baron, J., Malenka, D., et al., “Hip incidence estimates. Other population-based Fracture Incidence and Mortality in New studies have yielded similar annual incidence England,” Epidemiology 2(2): 116-122,1991. rates of hip fracture among white women (9,27). 10. Fleiss, J., and Gross, A., “Meta-Analysis in Based on these data, the values of the parameters a Epidemiology, with Special Reference to and b in the logistic function are as given in table Studies of the Association Between Exposure D-10. to Environmental Tobacco Smoke and Lung Cancer: A Critique,” Journal of Clinical Epi- REFERENCES demiology 44(2):127-139, 1991. 1. Black, D. M., “Cost Effectiveness of Screen- 11. Gardsell, P., Johnell, O., and Nilsson, B.E., ing for Osteoporosis: Review of Bone Miner- “Predicting Fractures in Women by Using al Density and Fracture Parameters Required Forearm Bone Densitometry,” Calcified Tis- for Model,” University of California, San sue International 44:235-242, 1989a. Francisco, CA, unpublished OTA contract re- 12. Gardsell, P., Johnell, O., and Nilsson, B. E., port, Nov. 17, 1992. “Predicting Various Fragility Fractures in 2. Black, D. M., “Why Elderly Women Should Women by Forearm Bone Densitometry. A be Screened and Treated to Prevent Osteopo- Follow-Up Study,” Calcified Tissue Interna- rosis,” American Journal of Medicine 98 tional 44(4):235-242, 1989b. (Suppl. 2A):67S-75S, 1995. 13. Hannan, M.T., Felson, D.T., and Anderson, 3. Black, D., Cummings, S., Genant, H., et al., J.J., “Bone Mineral Density in Elderly Men “Axial and Appendicular Bone Density Pre- and Women: Results from the Framingham dict Fractures in Older Women,” Journal of Osteoporosis Study,” Journal of Bone and Bone Mineral Research 7(6):633-638, 1992. Mineral Research, 7(5):547-553, May 1992. 4. Black, D., Cummings, S., and Melton, L., 14. Hansen, M. A., Overgaard, K., Riis, B.J., et III., “Appendicular Bone Mineral and a al., “Potential Risk Factors for Develop- Woman’s Lifetime Risk of Hip Fracture,” ment of Postmenopausal Osteoporosis—Ex- Journal of Bone Mineral Research amined Over a 12-Year Period,” Osteoporosis 7(6):639-646, 1992. International 1:95-102, 1991. Appendix E: Hormonal Replacement Therapy Regimens E

he proportion of post-menopausal women 14 days and a three-day drug-free period) or in who use hormone replacement therapy continuous combination with estrogen. These (HRT) has increased in the United States progestin/estrogen therapies (PERTs) alter the during the last two decades (13). While benefit-risk profile of HRT. Tuse of HRT has increased, the average dose and In the United States, conjugated equine estro- duration of use of postmenopausal estrogens has gen (CEE) (Premarin, Wyeth-Ayerst) is the most decreased until recently (64). This is due in part to commonly used form of estrogen for HRT.1,2 the discovery that the cancer-causing effects of There are a number of other estrogens used for postmenopausal estrogen are related to its dose HRT.3 Table E-1 lists the estrogens either ap- and duration. proved for osteoporosis by the Food and Drug Ad- This appendix describes the range of choice re- ministration or accepted for this use by a garding the dose, routes of administration, and committee of the United States Pharmacopoeia. In combinations of hormones currently in use or un- addition to the estrogens listed in the table, the es- der study as treatment for postmenopausal HRT. trogens quinestrol (Estrovis tablets, Parke-Davis) This appendix also describes the appropriate fol- and chlorotrianisene (TACE capsules, Marion low-up of women on HRT. Finally, this appendix Merrel Dow) are approved by the FDA for treat- describes how dosing regimens may affect comment of menopausal symptoms (61). pliance with HRT and describes other factors that affect HRT compliance. ESTROGEN DOSING REGIMENS HRT involves the administration of estrogen A central question for clinical management of alone or in combination with progestins. In the postmenopausal HRT patients is how small a dose past, estrogen was typically administered without of estrogen may be administered without losing progestin (unopposed estrogen) in estrogen re- the beneficial effects of the therapy. The reduction placement therapy (ERT). Currently, the most in bone loss or menopausal symptoms must be commonly used regimens for a woman with a weighed against the adverse effects of estrogens. uterus include a progestin either in sequence with The higher the dose of estrogen, the more likely estrogen (e.g., 25 days of estrogens with a concur- are side effects, such as breast tenderness or fluid rent progestin administered during the last 12 to

I 49 50 Cost Effectiveness of Screening for Osteoporosis

retention (15, 27, 28).4 In addition, higher doses For women who have not had hysterectomies, may increase the risk of estrogen-related illness ACOG recommends the addition of a progestin to such as endometrial cancer or gallbladder disease. the estrogen. They recommend a dose of 10 mg/ Several studies have demonstrated that doses day for 12 days a month to reduce the incidence of of at least 0.625 mg per day of CEE or its equiva- hyperplasia and endometrial cancer. ACOG and lent are necessary to prevent or greatly reduce the American College of Physicians see no reason bone loss in the spine in peri- or postmenopausal to add progestin to estrogen for a woman without women (36,46,72). Lower doses offer only partial a uterus. protection against bone loss (46,47). The minimal dosage of estrogen adequate to prevent bone loss ❚ Routes of Administration in postmenopausal women is discussed in greater There are a number of routes for delivery of estro- detail in appendix C. gens other than by mouth. Intramuscular injec- The American College of Obstetricians and tions have been tested, but they are no longer used, Gynecologists recommends the following estro- not only because they are uncomfortable but also gen dosages for osteoporosis (l): because estrogen plasma concentrations are un- Estrogen Dose stable with this method of administration (68). conjugated estrogen 0.625 mg/day Vaginal rings and vaginal creams have also been transdermal estradiol 0.05 mg twice a week investigated (53,63), but plasma estrogen levels micronized estradiol 1.0 mg/day are unstable, probably because of irregular ab- estrone sulfate 1.25 mg/day

1 There are currently no generic forms of conjugated estrogens on the market, but there are generic forms of some of the other estrogens used for postmenopausal replacement therapy. In 1989, the Center for Drug Evaluation and Research of the FDA rejected applications for new drug approval submitted by Barr Laboratories for five dosage strengths of generic conjugated estrogen (7). Although the extent of absorption of the estrogen was the same as that of Premarin, the brand-name drug manufactured by Wyeth-Ayerst, the FDA ruled that the generic manufacturer must demonstrate that the rate of absorption must be the same in order for the generic products to be considered bioequivalent to the innovator drug. Barr Labs claimed that this was inconsistent with a January 1989 determination by the FDA’s Fertility and Maternal Health Drugs Advisory Committee that the rate of absorption was not relevant to bioequivalence (7). Although a few members acknowledged that there was no conclusive evidence that the rate of estrogen absorption is important in determining the safety and efficacy of conjugated estrogens, the FDA’s Generic Drugs Advisory Committee concluded in February 1991 that the rates of absorption must be the same to establish bioequivalence (16). The FDA contended that different absorption rates could make conjugated estrogens ineffective in treating osteoporosis. In addition, more rapid absorption of estrogen into the blood stream could lead to higher peak drug plasma concentrations which could increase the risk of endometrial cancer (30). At present, sponsors of generic conjugated estrogen products are required to perform studies of the blood concentration-time profiles of five of the predominant estrogens in Premarin brand of conjugated equine estrogen (4). As of 1995, there were no generic conjugated estrogens on the market, although the generic manufacturer Duramed had an ANDA pending for a 0.625 mg formulation (4). Wyeth-Ayerst, manufacturer of Premarin brand of conjugated estrogen, has argued that a conjugated estrogen product that does not also contain a sixth form of estrogen, del- ta(8,9)-dehydroestrone, is not substantially equivalent to Premarin. 2 Data from Wyeth-Ayerst, manufacturers of the most widely prescribed postmenopausal estrogen, show that three types of physician spe- cialists — obstetrician-gynecologists, internists, and family-practitioners — wrote 90 percent of estrogen prescriptions. Obstetrician-gynecologists prescribe the most, with 2,897,000 prescriptions, or 43 percent of prescriptions for postmenopausal estrogens. 3 The estrogens used in hormone replacement therapy are much less potent than the synthetic estrogens used in oral contraceptives. Because of this difference in potency, the side effect profile of estrogens used in hormonal replacement therapy differs from that of estrogens used for contraception.

4 Fibroid tumors and endometriosis may alSO be exacerbated by HRT. Fibroids and endometriosis are both estrogen-dependent conditions that regress at menopause. HRT in postmenopausal women who had significant problems from either of these diseases premenopausally requires careful surveillance: fibroids may enlarge, and endometriosis maybe reactivated. If HRT is subsequently discontinued, the fibroids will again shrink. However, the sequelae of endometriosis, such as chocolate cysts or adhesions, may persist even after estrogen has been withdrawn, and continue to cause symptoms (20). Recommended dosages for osteoporosis Generic name Brand name (t.m.) Manufacturer FDA-approved indications and/or menopausal symptomsa,b

Conjugated equine Premarin tablets Wyeth-Ayerst 1) Moderate to severe vasomotor Menopausal symptoms: 0,625 mg to 1.25 estrogen symptoms of menopause mg a day cyclically or continuously 2) Vaginal or urethral atrophy Osteoporosis: 0.3 mg to 1.25 mg a day, 3) Osteoporosis cyclically or continuously 4) Hypoestrogenism due to castration, hypogonadism, or primary ovarian failure 5) Breast cancer 6) Prostatic carcinoma

Diethylstilbestrol (DES)’ Diethylstilbestrol Lilly 1) Breast cancer Neither the USP DI nor the product labeling enteric-coated 2) Prostatic carcinoma includes dosage information for tablets d osteoporosis or menopausal symptoms. Diethylstilbestrol tablets d m Esterified estrogens Estratab tablets Solvay 1) Moderate to severe vasomotor Menopausal symptoms: 0,625 mg to 1.25 Menest tablet SmithKline symptoms of menopause mg, cyclically or continuously Beecham 2) Vulvar or vaginal atrophy 3) Hypoestrogenism due to hypogonadism, castration, or primary ovarian failure 4) Breast cancer 5) Prostatic carcinoma

Estradiol Estrace tablets Bristol-Myers 1) Osteoporosis Menopausal symptoms: 0.5 mg to 2 mg a Emcyt capsules Squibb 2) Moderate to severe vasomotor day, cyclically or continuously Pharmacia Adria symptoms of menopause Osteoporosis: 0.5 mg a day, cyclically or 3) Vulvar or vaginal atrophy continuously 4) Hypoestrogenism due to hypogonadism, castration, or primary ovarian failure 5) Breast cancer 6) Prostatic carcinoma —

Recommended dosages for osteoporosis Generic name Brand name (t.m.) Manufacturer FDA-approved indications and/or menopausal symptomsa’b

Estradiol transdermal Estraderm Ciba 1 ) Osteoporosis Osteoporosis or menopausal symptoms: system 2) Moderate to severe vasomotor One transdermal dosage system delivering symptoms of menopause 0.05 mg or 0.10 mg, per day worn 3) Vulvar or vaginal atrophy continuously and replaced twice a week 4) Hypoestrogenism due to hypogonadism, castration, or primary ovarian failure

Estropipate Ogen tablets Upjohn 1) Osteoporosise Menopausal symptoms: 0.75 mg to 5 mg Ortho-est tablets Ortho 2) Moderate to severe vasomotor estropipate per day, cyclically or symptoms of menopause continuously 3) Vaginal or vulvar atrophy Osteoporosis: 0.75 mg per day for 25 days 4) Hypoestrogenism due to hypogonadism, of a 31-day cycle castration, or primary ovarian failure

Ethinyl estradiolc Estinyl tablets Schering 1) Moderate to severe vasomotor Menopausal symptoms: 0.02 mg of 0.05 symptoms of menopause mg per day, cyclically or continuously 2) Hypogonadism 3) Prostatic carcinoma 4), Breast cancer a Dosages are for vasomotor symptoms of menopause. b Dosages were those recommended in the USP DI, See the United States Pharmacopeial Convention, USPDispensing Information (USP Dl): Volume 1—Drug Information for the Health Care Professionl (Taunton, MA Randy McNally, 1995). c A U.S. Pharmacopeia Advisory Committee has accepted osteoporosis as an unlabeled indication for this product. An indication for postmenopausal osteoporosis, however, is not included in the FDA-approved labeling for this product See The United States Pharmacopeial Convention,USP Dispensing Information (USP DI): Volume 1—Drug Information for the Health Care Profession/ (Taunton, MA Rand McNally, 1995). d Diethylstilbestrol is available only as a generic In the United States. e The FDA-approved labeling of Ortho-est (Ortho) brand of estropipate does not include an indication for osteoporosis Key: CEE = conjugated equine estrogen, HRT = hormone replacement therapy, t m = trademark SOURCE: Office of Technology Assessment, 1995 Appendix E Hormonal Replacement Therapy Regimens 53

sorption due to day-to-day changes in vaginal A number of progestins can be used for PERT.5 blood flow and secretion. The most commonly used progestin in the United Implantation of continuously released estrogen States is medroxyprogesterone (Provera, Up- under the skin (subcutaneous implantation) ap- john). In addition, the FDA has recently approved pears to result in stable estrogen levels (68). But a combination of medroxyprogesterone acetate once inserted, the implants are difficult to remove and conjugated equine estrogen for the prevention in case of overdose or intolerance (76). of osteoporosis. (See table E-2.) Administration of estrogen through a patch or Progestins produce progressive endometrial cream applied to the skin (percutaneous transder- atrophy, converting adenomatous hyperplasia to mal administration) has proved effective in treat- normal endometrium (85). Numerous studies ing postmenopausal symptoms (14,22) and in show that combined estrogen/progestin therapies reducing vertebral bone loss after menopause can return 98 to 99 percent of preexisting hyper- (72). plasia back to normal endometrium (32,73). (See Transdermal medication may increase com- appendix G.) Observational studies also show that pliance because it eliminates the need for multiple PERT users have a lower risk of endometrial can- dose scheduling, is easily administered, requires cer than ERT users (42). only twice weekly application, and is reversible An important unresolved issue regarding PERT (78). However, the gel is difficult to administer ac- is whether the benefits of progestins in protecting curately (68). Absorption is proportional to the the endometrium are outweighed by the effect of surface of application, and this surface cannot be progestins on the risk of coronary heart disease determined accurately. In addition, between 5 and (CHD). Epidemiologic studies of the relationship 20 percent of women may develop skin irritation of HRT to CHD have been largely limited to unop- (79). posed estrogens; the effect of progestin supplementation on heart disease risk has not been as THERAPY WITH PERT extensively evaluated, but recent evidence sug- The primary indication for adding progestins to gests that adding progestins to HRT may attenuate estrogen replacement therapy is to reduce the risk the beneficial effects of ERT on heart disease. (See of estrogen-induced irregular bleeding, endome- appendix I.) trial hyperplasia (abnormal overgrowth of the in- Progestins are more often responsible than es- ner lining of the uterus, or endometrium), and trogen for making hormonal replacement therapy endometrial cancer (26,87). (See appendix G). In unacceptable for some women, because adverse the Postmenopausal Estrogen/Progestin Interven- effects are common with progestins (71). Proges- tions (PEPI) study, almost half of the women tins can produce breast tenderness, bloatedness, assigned to unopposed estrogen therapy experi- edema, abdominal cramps, and an iatrogenic pre- enced endometrial hyperplasia over the two year menstrual-like syndrome (71,87). Patients also clinical trial (91); the PEPI trial protocol required commonly experience side effects such as anxiety, that these women be taken off of estrogen. Be- irritability, depressed mood, and drowsiness. cause of the unexpectedly large proportion of One of the primary reasons for stopping HRT is women on ERT who developed hyperplasia in discomfort with periodic bleeding (18). With se- PEPI, the directors of the Women’s Health Initia- quential therapies, regular bleeding occurs with tive long-term clinical trial of hormonal replace- 85 percent of patients (31,87). This proportion de- ment therapy decided to place all women assigned creases with time, and by age 65, 60 percent con- to unopposed estrogen on PERT. tinue to experience light bleeding (33). Patients

5 Synthetic Progesterone are often used in hormonal replacement therapy since natural progesterones cannot be absorbed orally. 54 Cost Effectiveness of Screening for Osteoporosis

Brand name Recommended dosage for Generic name (tin.) Manufacturer FDA-approved indications secondary amenorrheab

Medroxyprogesterone Amen Carnick 1 ) Secondary amenorrhea Secondary amenorrhea. May acetate Curretab Solvay 2) Abnormal uterine bleeding be given in dosages of 5 to 10 mg daily for from 5 to 10 Cycrin Esi Pharma days. Provera Upjohn

Norethindrone acetate Aygestin Esi Pharma 1) Secondary amenorrhea Secondary amenorrhea: May Norlutate Parke-Davis 2) Endometriosis be given in dosages of 2.5 to 3) Abnormal uterine bleeding 10 mg daily for 5 to 10 days during the second-half of the theoretical menstrual cycle.

a A U.S. Pharmacopeia Advisory Committee has determined that osteoporosisIS an accepted indication for these products. An indication for osteoporosis, however,IS not included on the FDA-approved labeling for these products See The United States Pharmacopeial Convention,USP Dispensing Information (USP DI): Volume 1—Drug Information for the Health Care Professional(Tauton, MA: Rand McNally, 1995) b The FDA-approved product Iabeling does not have dosage recommendations for use Of these products with estrogen in Osteoporosis KEY: t.m. = trademark SOURCE Off Ice of Technology Assessment, 1995 may be willing to tolerate bleeding for relief of Another approach to reduce the frequency of menopausal symptoms, but to an asymptomatic bleeding and improve compliance is to give se- woman in her 60s and 70s who is taking estrogens quential PERT with less than monthly progestin for the prevention of osteoporosis and cardiovas- therapy. For example, Williams and colleagues cular disease, the persistent bleeding and other found that that there was less vaginal bleeding side effects may be intolerable (26). when progestins were administered for 14 days Compliance may be improved with new con- every three months than when given for 14 days tinuous combined regimens of PERT that reduce every month (90). Menopausal women find less the frequency of menstrual bleeding. Continuous than monthly bleeding more acceptable than PERT involves daily administration of estrogen monthly bleeding (6). and a low dose of progestin (70). Several studies show that continuous PERT can relieve meno- ACUTE INDICATIONS FOR HRT pausal symptoms, eliminate periodic bleeding The most common indication for HRT is relief of within several months of initiation, and avoid en- menopausal symptoms (41).6 Hot flashes, or va- dometrial hyperplasia (5,28,37,40,48,70,84,91). somotor symptoms, are the most common symp- Most studies to date have found between one- tom of menopause that causes women to seek third and one-half of patients were bleeding after medical attention (81). Hot flashes occur in 60 to three months of continuous PERT, but most pa- 75 percent of women at the time of menopause tients were amenorrheic after 12 months. And (52,82). They are typically more severe in women most studies reported 90 percent or greater atro- with surgical menopause, because the severity of phic endometrium at 12 months.

6 The American College of Physicians distinguishes between diagnoses for short-term and long-term HRT (3). Short-term HRT is prescribed for women who suffer from postmenopausal symptoms, such as hot flashes and atrophic vaginitis. The American College of Physicians suggests therapy of one to five years for the treatment of symptoms associated with menopause. The goals of long-term HRT are the prevention of osteoporosis and decrease in the risk of heart disease. The American College of Obstetricians and Gynecologists recommends that for treatment of menopausal symptoms, the lowest dosage of estrogen that provides effective relief should be used (1). Appendix E Hormonal Replacement Therapy Regimens 55

vasomotor symptoms is thought to be related to in feelings of well being in women on HRT the rapidity in decline of estrogen (51). At first, (21,56,88,89). This effect may be independent flashes usually occur several times a day, and from its impact on menopausal symptoms (45).8 often interrupt sleep (25). Irritability, fatigue, and anxiety can result from sleep deprivation. For EVALUATION AND FOLLOWUP most patients, vasomotor symptoms are self-lim- OF WOMEN TAKING HRT iting, but for 25 to 50 percent of women, these flashes persist more than five years (77,82). Clini- Before hormonal replacement therapy is begun by cal trials have demonstrated that estrogen is effec- a postmenopausal woman, the American College tive in relieving these symptoms in about 95 of Physicians (ACP) recommends that her physi- percent of patients (19,75). cal condition should be assessed by a physician Estrogen has been found to relieve symptoms (3). The doctor should be aware of her medical of menopause that affect the vagina, uterus, ure- history and her current health in light of contrain- thra, and bladder.7 Estrogen replacement therapy dications to HRT. These contraindications include can prevent the vaginal atrophy associated with unexplained vaginal bleeding, acute liver disease, menopause and maintain the normal tone of sup- chronic impaired liver function, recent vascular porting ligaments and elastic tissues of the uterus thrombosis, breast cancer, and endometrial and vagina (9,80,89). Vaginal atrophy may result cancer. in vaginal dryness, itching, burning, and infec- The American College of Obstetricians and tion. Vaginal atrophy can also result in pain with Gynecologists (ACOG) recommends that women vaginal intercourse and resultant sexual dysfunc- taking hormonal replacement therapy be moni- tion (10). Estrogen can also prevent atrophy of the tored every year (1). At that time breast and pelvic bladder and urethra and the resultant painful examinations should be performed, a Pap smear urination, urgency, stress incontinence, frequency should be taken, and cholesterol level and blood of urination, urination at night, and dripping after pressure should be monitored. If the woman is on voiding (24,29,66,80). a regime that includes a progestin and there is no The absence of estrogen may cause skin to be- excess or prolonged bleeding, there is no need for come thinner, as the amount of collagen in the skin an annual endometrial biopsy. Mammograms decreases (10,80). Estrogen stimulates the synthe- should be performed annually on women over the sis of collagen, and in postmenopausal women re- age of 50. ceiving estrogen, the collagen of the skin is Endometrial biopsy is not deemed to be neces- maintained at premenopausal levels (11). sary in patients on sequential PERT because the Some investigators have shown reductions in onset of bleeding can be a useful predictor of en- anxiety and depressed mood, and improvements dometrial status. Patients with proliferation and

7 Alternatives to estrogen, such as progestogens, clonidine, or ergot alkaloids, have been used for symptomatic relief of vasomotor symptoms in women who are not candidates for estrogen replacement therapy (75,81). But none of these alternatives will prevent atrophy of the vagina (81).

8 A major difference of opinion persists as to whether estrogen therapy has any direct positive effect on mood, or whether the improved well-being reported by some women is simply due to their relief from vasomotor symptoms, and possibly due to a placebo effect (60). Some studies purport to show that a large portion of the influence of hormone replacement on affect is due to alleviation of hot flashes by hormone replacement (23). Yet other studies have demonstrated that estrogen therapy directly affects mood (12). Campbell and Whitehead, in a double- blind study of 64 women over four months, demonstrated significant improvements in certain psychological problems such as anxiety, irritability, worry about age, and optimism. They found that estrogen significantly improves anxiety and other psychological symptoms even in menopausal women who had had no vasomotor symptoms. However, a direct biological effect of loss of estrogen with menopause on depressed mood has not been demonstrated (35). Although estrogen may help alleviate depressed mood that accompanies menopause (44), major depression requires psychiatric treatment (35). 56 Cost Effectiveness of Screening for Osteoporosis

hyperplasia of the endometrium bleed earlier in Compliance with HRT is affected by various the cycle than is normal with sequential PERT factors, and knowledge of these factors suggests (59,87). ways of improving compliance. In addition, knowledge of these factors is important in under- COMPLIANCE WITH HRT standing how bias may affect our interpretation of A number of studies that have examined rates of observational studies of HRT’s risks and benefits. compliance with HRT have in general found long- (See appendices F and I for a discussion of bias.) term compliance rates to be low. One study of In general, compliance rates with drugs will be 1,586 women enrolled in the Harvard Community lower if the patient suffers no physical symptoms Health Plan who received new prescriptions for or if the symptoms disappear before the end of the HRT found that 27 percent stopped taking HRT treatment (62). Women who suffer from more se- within 100 days of receiving the prescription and vere menopausal symptoms, such as hot flashes, 40 percent had stopped after one year (50). are more likely to use hormone replacement thera- Some studies distinguish between commence- py. It has been found that women with surgical ment compliance (the proportion of women menopause are more likely to use and comply prescribed HRT who initiate therapy) and mainte- with long-term hormone replacement therapy, in nance compliance (proportion of women on HRT part because their menopausal symptoms tend to who continue to take it over a specified period of be more severe (13,43). Leaner women are more time). Speroff et al. (1991) estimated that the com- likely to use hormonal replacement therapy; be- mencement compliance rate for women with natu- cause body fat is an important nonovarian source ral menopause is between 21 and 60 percent (69). of estrogen production, thin women tend to have The five-year maintenance compliance rate is be- more severe menopausal symptoms than women tween 5 and 34 percent. For women with bilateral who are heavier (34). Women who smoke ciga- oophorectomies, the commencement compliance rettes are more likely to take HRT, possibly be- rate is between 31 and 89 percent. Their five-year cause of an antiestrogenic effect of smoking, maintenance rate is 13 to 71 percent (69). which intensifies menopausal symptoms (38). Compliance with HRT tends to decrease over But the proportion of women who use HRT for re- time. One study conducted for Wyeth Ayerst ex- lief of acute menopausal symptoms declines in the amined compliance with HRT in postmenopausal years following menopause as these symptoms di- women who were members of a prepaid group minish in frequency and severity. health benefit plan and who filled prescriptions Women typically do not suffer symptoms of os- for conjugated equine estrogens (Premarin, teoporosis, such as hip fractures or kyphosis, until Wyeth Ayerst) (83). Data on rates of compliance many years after menopause. Because low bone were gathered from pharmacy and medical re- mineral density (BMD) does not have obvious cords. Compliance rates were determined by symptoms unless fracture occurs, bone density comparing the number of days Premarin was pre- measurement may increase commencement and scribed to the number of days for which Premarin maintenance compliance with HRT. Some experts was dispensed. They found that compliance de- have suggested physicians could use densitome- clined from 62.7 percent over one year, to 56.1 try to help patients who are undecided about initi- percent over three years and to 46.8 percent over ating HRT to “visualize” their low bone mass seven years. (57).9 In addition, maintenance compliance might

9 Measurements of biochemical markers of bone resorption may also be used to improve compliance with HRT. These markers may allow the clinician to identify patients who are failing to respond to HRT, which may be because the patient is not complying with the prescribed regimen (41a). The use of biochemical markers of bone resorption as a tool to improve compliance with HRT has not been evaluated (Id.). Appendix E Hormonal Replacement Therapy Regimens 157

be improved by using densitometry to follow a pa- physicians in their prescribing of HRT, including tient’s bone density over time, although there is no estimates of benefits and risks that may not be evidence available to test this possibility. supported by scientific data (74). In addition, Women with below-normal BMD as deter- some physicians appear to use patterns of admin- mined by densitometry are more likely to take es- istration of HRT that may diminish the chance of trogen as a preventive measure for osteoporosis appropriate patient compliance and fail to adjust (65). One study surveyed 261 women in Califor- therapy when problems occur (74). nia who had had their BMD measured. Post- The clinical setting may also have an impact on menopausal women who had below-average compliance. Experimental clinical trials of HRT BMD for their age, sex, and race combination have generally shown better rates of compliance were five times as likely to begin taking estrogen than studies of HRT compliance outside of the as women with normal densitometry results (odds trial setting. For example, one clinical trial of HRT ratio 8.4, 95 percent confidence interval 3.4 to in women with hysterectomies showed perfect 20.9). While the study showed that more women compliance during the 18 months of the study with low BMD at densitometry initiated estrogen (39). A clinical trial of estrogen patches showed replacement therapy than women with normal perfect compliance over four months (58). Both of BMD, it did not report on the effect of densitome- these clinical trials involved small samples of try on long-term compliance with HRT. Another women (22 and 12, respectively.) In addition, study of compliance with HRT among 352 post- compliance may be better in clinical trials because menopausal women who had BMD measure- there are more intensive efforts at follow-up than ments, however, found that 40 percent of the generally occur in the normal clinical settings. women who were recommended HRT for low Compliance is also affected by the age at which bone density were not taking HRT eight months hormonal replacement therapy is initiated. Elder- after referral (67) ly postmenopausal women more frequently object Women’s attitudes and beliefs about HRT af- to the resumption of menstrual bleeding induced fect compliance with hormonal replacement ther- by PERT than perimenopausal and early postme- apy. In general, compliance with drugs will be nopausal women (6). Other factors also affect the lower if the patient is not convinced the medica- compliance of elderly patients with medication tion will help or if the patient is afraid of the devel- regimens. A patient is less likely to comply if she opment of side effects (62). Many women are has a poor understanding of the prescription resistant to taking HRT because it is “not natural” instructions, if the therapy is long term, or if the (54). Women may discontinue HRT because of prescription has complex instructions (62). An el- unacceptable side effects, such as resumed men- derly woman may suffer from impaired vision or struation, breast tenderness, weight gain, head- hearing that could impede her ability to read a aches, and abdominal bloating (55). In addition, drug label or hear instructions for its use. In addi- women may decide not to use HRT because it may tion, many elderly people live alone, and it has increase their risk of endometrial cancer and been shown that people who live alone are less breast cancer. likely to comply with a medication regimen than Greater patient education about the magnitude those who do not (62). In addition, the expense of and direction of effects and risks may improve certain medications may have an impact on com- compliance with HRT. In general, compliance is pliance by the elderly with limited fixed incomes. better with more patient education about the dis- Behavioral patterns of women who take estro- ease and the regimen (55). gen, such as regular physician visits (to refill pre- The uptake and compliance with HRT may also scriptions, for example), differentiate them from be affected by physicians’ beliefs and recommen- women who do not take estrogen. In order to ex- dations. A number of factors appear to influence amine the effect of health behaviors on HRT use, 58 Cost Effectiveness of Screening for Osteoporosis

Recommended dosages for Brand osteoporosis and Generic name name (t.m.) Manufacturer FDA-approved indications menopausal symptoms

Conjugated estrogens Prempro Wyeth-Ayerst 1) Moderate to severe Menopausal symptoms. and medroxyprogesterone vasomotor symptoms 0.625 mg CEE and 2.5 mg acetate for continuous of menopause medroxyprogesterone HRT 2) Vulvar and vaginal acetate daily atrophy Osteoporosis: 0.625 mg CEE 3) Osteoporosis (in and 2.5 mg women with an intact medroxyprogesterone uterus) acetate daily

Conjugated estrogens Premphase Wyeth-Ayerst 1) Moderate to severe Menopausal symptoms. and medroxyprogesterone vasomotor symptoms 0.625 mg CEE (cyclic acetate for sequential HRT of menopause administration) and 5 mg 2) Vulvar or vaginal medroxyprogesterone atrophy acetate for last two weeks of 3) Osteoporosis (in cycle women with an intact Osteoporosis: 0.625 mg CEE uterus) (cyclic administration) and 5 mg medroxyprogesterone acetate for last two weeks of cycle

KEY: CEE = conjugated equine estrogen, HRT = hormone replacement therapy, t m = trademark SOURCE: Office of Technology Assessment, 1995

Barrett-Connor et al. studied 1,057 postmenopau- their daily exercise over the past year, whereas sal women from the same socioeconomically up- only 29 percent of never users had increased their per-middle-class community in California who exercise. Women who never used HRT were less participated in a clinic evaluation of their estrogen likely to have implemented healthy behavior use patterns (8). After an average of 4.4 years later, changes, and were least likely to have had screen- 95 percent of these women completed a mailed ing evaluations than women who had used HRT. health survey questionnaire. This questionnaire Compliance may be affected by the type of asked them about recent changes in lifestyle be- packaging. One study of 177 patients compared a haviors that affect their health, such as consump- calendar-oriented system of HRT packaging to tion of dietary fat, salt use, and exercise habits, as conventional packaging of HRT (50). Compliance well as frequency of blood pressure checkups, rose from 23 percent when the pills were provided mammograms, and Pap smears. in conventional packaging to 82 percent when the The study found that women who were current- pills were provided in a prepackaged blister card ly using HRT were significantly more likely to system. Wyeth-Ayerst introduced a prepackaged have recently implemented new healthy lifestyle blister card system of packaging in 1995. (See behaviors than women who had never used HRT table E-3.) (8). For example, 70 percent of the women who Thus, a variety of factors affect compliance were currently using HRT had had a mammogram with HRT. Increased awareness of the factors af- in the last year, whereas only 45 percent of the fecting compliance with HRT suggests ways of women who had never used HRT had had one. improving long-term compliance. Thirty-eight percent of current users had increased Appendix E Hormonal Replacement Therapy Regimens 59

REFERENCES Thickness,” British Journal of Obstetrics and

10 American College of Obstetricians and Gy- Gynecology 92:256-259, 1985. necologists, “Hormone Replacement Thera- 12. Campbell, S., and Whitehead, M., “Oestro- py,” ACOG Technical Bulletin, No. 166 gen Therapy and the Menopause Syndrome,” (Washington, DC: American College of Ob- The Menopause: Clinics in Obstetrics and stetricians and Gynecologists, April 1992). Gynecology, Greenblatt and Studds (eds.) 2. American College of Obstetricians and Gy- (London, England: W.B. Saunders, 1977). necologists, “Osteoporosis,” ACOG Techni- 13. Cauley, J. A., Cummings, S.R., Black, D. M., cal Bulletin, No. 167 (Washington, DC: et al., “Prevalence and Determinants of Estro- American College of Obstetricians and Gy- gen Replacement Therapy in Elderly necologists, May 1992). Women,” American Journal of Obstetrics 3. American College of Physicians, “Preventive and Gynecology 163: 1438-1444, 1990. Care Guidelines: 1991,” Annals of Internal 14. Chetkowski, R., Meldrum, D., Steingold, K., Medicine 114(9):758-783, 1991. et al., “Biologic Effects of Transdermal Estra- 4. “APhA Weighs in on Oestrogens,” Scrip diol,” New England Journal of Medicine World Pharmaceutical News 2013:29, 1995. 314:1615-1620, 1986. 5. Archer, D. F., Pickar, J. H., and Bottiglioni, F., 15. Christensen, M. S., Hagen, C., Christiansen, “Bleeding Patterns in Postmenopausal C., et al., “Dose-Response Evaluation of Cy- Women Taking Continuous Combined or Se- clic Estrogen/Gestagen in Postmenopausal quential Regimens of Conjugated Estrogens Women: Placebo-Controlled Trial of Its Gy - with Medroxyprogesterone Acetate,” Obstet- necologic and Metabolic Actions,” American rics & Gynecology 83(5 Pt 1):686-692, 1994. Journal of Obstetrics and Gynecology 6. Barensten, R., Groenveld, F. P., Bareman, 144(8):873-879, 1982. F. P., et al., “Women’s Opinion on Withdrawal 16. “Conjugated Estrogens Rate of Absorption Bleeding with Hormone Replacement Thera- Should be Measured—FDA Advisory Com- py,” European Journal of Obstetrics & Gy- mittee; New Generic Formulations Could necology and Reproductive Biology 51(3); Take at Least Two Years to Develop,” F-D-C 203-207, 1993. Reports Mar. 4, 1991, pp. 9-10. 7. “Barr Labs Requests Conjugated Estrogens 17. Coope, J., “Menopause: Diagnosis and Treat- Hearing to Review FDA’s Denial of Blood ment,” British Medical Journal 289(6449): Level-Based ANDA Applications; Agency 888-890, 1984. Cites Difference in Absorption,” F-D-C Re- 18. Coope, J., and Marsh, J., “Can We Improve ports Dec. 4, 1989, pp. 8-10. Compliance with Long-Term HRT?” Maturi- 8. Barrett-Connor, E., “Postmenopausal Estro- tas 15(2):151-158, 1992. gen and Prevention Bias,” Annals of Internal 19. Coope, J., Thomson, J. M., and Poller, L., Medicine 115(6):455-456, 1991. “Effects of ‘Natural Oestrogen’ Replacement 9. Beard, M. K., “Atrophic Vaginitis. Can It be Therapy on Menopausal Symptoms and Prevented as Well as Treated?” Postgraduate Blood Clotting,” British Medical Journal Medicine 91(6):257-260, 1992. 4(5989):139-143, 1975. 10. Booher, D.L., “Estrogen Supplements in the 20. Cust, M., Gangar, K., Hillard, T., et al., “A Menopause,” Cleveland Clinic Journal of Risk-Benefit Assessment of Estrogen Thera- Medicine 57:154-160, 1990. py in Postmenopausal Women,” Drug Safety 11. Brincat, M., Moniz, C.F., Studd, J.W.W., et 5:345-358, 1990. al., “The Long Term Effects of the Meno- 21. Daly, E., Gray, A., Barlow, D., et al., “Mea- pause and of Administration of Sex Hor- suring the Impact of Menopausal Symptoms mones on Skin Collagen and Skin 60 Cost Effectiveness of Screening for Osteoporosis

on Quality of Life,” British Medical Journal 31. Fraser, I., “A Review of the Role of Proges- 307(6916):1420-1421 , 1993. tins in Hormonal Replacement Therapy: In- 22. De Ligni’eres, B., Basdevant, A., Thomas, fluence on Bleeding Patterns,” Maturitas G., et al., “Biological Effects of Estradiol 8:113-121, 1986. 17-beta in Post Menopausal Women: Oral 32. Gambrell, R., “Prevention of Endometrial Versus Percutaneous Administration,” Jour- Cancer with Progestogens,” Maturitas 8: nal of Clinical Endocrinology and Metabo- 159-168, 1986. lism 62:536-541, 1986. 33. Gambrell, R., Bagnell, C., and Greenblatt, 23. Dennerstein, L., Burrows, G. C., Hyman, R., “Role of Estrogens and Progesterone in G.J., et al., “Hormone Therapy and Affects,” the Etiology and Prevention of Endometrial Maturitas 1(4):247-259, 1979. Cancer” (review), American Journal of Ob- 24. Elia, G., and Bergamn, A., “Estrogen Effects stetrics and Gynecology 146:696-707, 1983. on the Urethra: Beneficial Effects in Women 34. Garcia Rodriguez, L. A., Pfaff, G. M., Schu- with Genuine Stress Incontinence,” Ob- macher, M. C., et al., “Replacement Estrogen stetrical and Gynecological Survey 48(7): Use and Body Mass Index,” Epidemiology 509-517, 1993. 1:219-223, 1990. 25. Erlick, Y., Tataryn, I. V., Meldrum, D. R., et 35. Gath, D., and Iles, S., “Depression and the al., “Association of Waking Episodes with Menopause,” British Medical Journal Menopausal Hot Flushes,” Journal of the 300(6735):1287-1288, 1990. American Medical Association 245(17): 36. Genant, H., Baylink, D., and Gallagher, J., 1741-1744, 1981. “Effect of Estrone Sulfate on Postmenopausal 26. Ernster, V., Bush, T., Huggins, G., et al., Bone Loss,” Obstetrics & Gynecology 76(4): “Benefits and Risks of Menopausal Estrogen 579-584, 1990. and/or Progestin Hormone Use,” Preventive 37. Gillet, J. Y., Andre, G., Faguer, B., et al., “In- Medicine 17:201-233, 1988. duction of Amenorrhea During Hormone Re- 27. Ettinger, B., Genant, H., and Cann, C., placement Therapy: Optimal Micronized “Long-Term Estrogen Replacement Therapy Progesterone Dose. A Multicenter Study,” Prevents Bone Loss and Fractures, ’’Annals of Maturitas 19(2): 103-115, 1994. Internal Medicine 102:319-324, 1985. 38. Greenberg, G., Thompson, S. G., and Meade, 28. Ettinger, B., Golditch, I., and Friedman, G., T.W., “Relation Between Cigarette Smoking “Gynecologic Consequences of Long-Term and Use of Hormonal Replacement Therapy Unopposed Estrogen Replacement Therapy,” for Menopausal Symptoms,” Journal of Epi- Maturitas 10:271-282, 1988. demiology and Community Health 41:26-29, 29. Fantl, J. A., Cardozo, L., and McClish, D. K., 1987. “Estrogen Therapy in the Management of 39. Hahn, R. G., Nachtigall, R. D., and Davies, Urinary Incontinence in Postmenopausal T. C., “Compliance Difficulties with Proges- Women: A Meta-Analysis. First Report of the tin-Supplemented Estrogen Replacement Hormones and Urogenital Therapy Commit- Therapy,” Journal of Family Practice 18(3): tee,” Obstetrics & Gynecology 83(1):12-18, 411-414, 1984. 1994. 40. Hazzard, W., “Estrogen Replacement and 30. “FDA Position on Conjugated Estrogens Cardiovascular Disease: Serum Lipids and Rate of Absorption in Determining Bioequi- Blood Pressure Effects,” American Journal valance Supported in Split Decision by FDA of Obstetrics and Gynecology 161 (6): Advisory Subcommittee,” F-D-C Reports, 1847-1853, 1989. May 7, 1990, pp. 7-10. Appendix E Hormonal Replacement Therapy Regimens 61

41. Hemminki, E., Kennedy, D. L., Baum, C., et 51. Maschak, C. A., et al., “The Relation of Phys- al., “Prescribing of Noncontraceptive Estro- iological Changes to Subjective Symptoms in gens and Progestins in the United States, Postmenopausal Women With and Without 1974-86,” American Journal of Public Hot Flashes,” Maturitas 6:301-308, 1985. Health 78(11): 1479-1481, 1988. 52. McKinlay, S. M., and Jeffries, M., “The 41a. Jenkins, D., Product Manager, Metra Biosys- Menopause Syndrome,” British Journal of tems, Mountainview, CA, personal commu- Social and Preventive Medicine 28: 108-115, nication, April, 1995. 1974. 42. Jick, S., Walker, A., and Jick, H., “Estrogens, 53. Mishell, D., Moore, D., Roy, S., et al., “Clini- Progesterone, and Endometrial Cancer,” Epi- cal Performance and Endocrine Profiles with demiology 4(1):20-24, 1993. Contraceptive Vaginal Rings Containing a 43. Johannes, C. B., Crawford, S. L., Posner, J. G., Combination of Estradiol and D-norgestrel,” et al., “Longitudinal Patterns and Correlates American Journal of Obstetrics and Gynecol- of Hormone Replacement Therapy Use in ogy 53:277-281, 1978. Middle-Aged Women,” American Journal of 54. Murkies, A. L., “Common Problems with Epidemiology 140(5):439-452, 1994. Hormone Replacement Therapy,” Australian 44. Lauritzen, C., “Clinical Use of Oestrogens Family Physician 21(3):217-218, 1992. and Progestogens,” Maturitus 12(3): 55. Nachtigall, L. E., “Enhancing Patient Com- 199-214, 1990. pliance with Hormone Replacement Therapy 45. Limouzin-Lamothe, M. A., Mairon, N., at Menopause,” Obstetrics & Gynecology Joyce, C. R., et al., “Quality of Life After the 74(4 Suppl.):77S-80S; discussion 81S-83S, Menopause: Influence of Hormonal Replace- 1990. ment Therapy,” American Journal of Obstet- 56. Nathorst-Boos, J., von Schultz, B., and Carls- rics and Gynecology 170(2):618-624, 1994. trom, K., “Elective Ovarian Removal and Es- 46. Lindsay, R., Hart, D., and Clark, D., “The trogen Replacement Therapy—Effects on Minimum Effective Dose of Estrogen for Pre- Sexual Life, Psychological Well-Being and vention of Postmenopausal Bone Loss,” Ob- Androgen Status,” Journal of Psychosomatic stetrics & Gynecology 63(6):759-763, 1984. Obstetrics and Gynecology 14(4):283-293, 47. Lindsay, R., Hart, D., Forrest, C., et al., “Pre- 1993. vention of Spinal Osteoporosis in Oophrecto- 57. Notelovitz, M., “Hormonal Therapy in Cli- mized Women,” Lancet 1151-1154, 1980. macteric Women: Compliance and Its Socio- 48. MacLennan, A. H., MacLennan, A., Wenzel, economic Impact,” Public Health Report S., et al., “Continuous Low-Dose Oestrogen 104( Suppl.):70-75, 1989. and Progestogen Hormone Replacement 58. Padwick, M. L., Endacott, M. B., and White- Therapy: A Randomised Trial,” Medical head, M. I., “Efficacy, Acceptability, and Met- Journal of Australia 159(2):102-106, 1993. abolic Effects of Transdermal Estradiol in the 49. Magos, A., Brewster, E., Singh, R., et al., Management of Postmenopausal Women,” “The Effects of Norethisterone in Postmeno- American Journal of Obstetrics and Gynecol- pausal Women on Oestrogen Replacement ogy 152: 1085-1091, 1985. Therapy: A Model for Premenstrual Syn- 59. Padwick, M., Pryse-Davies, J., and White- drome,” British Journal of Obstetrics and head, M., “A Simple Method for Determining Gynecology 93: 1290-1296, 1986. the Optimal Dose of Progestin in Postmeno- 50. Marwick, C., “Hormone Combination Treats pausal Women Receiving Estrogens,” New Women’s Bone Loss,” Journal of the Ameri- England Journal of Medicine 315:930-934, can Medical Association 272(19):1487, 1986. 1994. 62 Cost Effectiveness of Screening for Osteoporosis

60. Payer, L., “Do Hormones Have a Direct Ef- American Journal of Obstetrics and Gynecol- fect on Mood?” A Portrait of the Menopause, ogy 164(1 Pt 1): 165-174, 1991. H. Burger and M. Boulet (eds.) (Pearl River, 70. Sporrong, T., Hellgren, M., Samsioe, G., et NY: The Parthenon Publishing Group, 1991). al., “Comparison of Four Continuously Ad- 61. Physicians’ Desk Reference, 49th Ed. (Mont- ministered Progestogen Plus Estradiol Com- vale, NJ: Medical Economics Data Produc- binations for Climacteric Complaints,” tion Company, 1995). British Journal of Obstetrics and Gynecolo- 62. Public Health Service, Office of Inspector gy 95:1042-1048, 1988. General, Office of Evaluation and Inspec- 71. Stevenson, J. C., “Strategies to Increase Es- tions, Medication Regimens: Causes of Non - trogen Acceptance and Utilization and to De- compliance, OE1-04-89-89121 (Washington, crease Side Effects,” Postgraduate Medicine DC: March 1990). Spec. No.:15-17; discussion 33-43, April 63. Riggs, L., Hermann, H., and Yen, S., “Ab- 1989. sorption of Estrogens from Vaginal Creams,” 72. Stevenson, J., Cust, M., and Gangar, K., “Ef- New England Journal of Medicine fects of Transdermal Versus Oral Hormone 298:195-197, 1978. Replacement Therapy on Bone Density in 64. Ross, R. K., Paganini-Hill, A., Roy, S., et al., Spine and Proximal Femur in Postmenopau- “Past and Present Preferred Prescribing Prac- sal Women,” Lancet 336:265-269, 1990. tices of Hormone Replacement Therapy 73. Studd, J., Thorn, M., Paterson. M., et al., Among Los Angeles Gynecologists: Possible “The Prevention and Treatment of Endome- Implication for Public Health,” American trial Pathology in Postmenopausal Women Journal of Public Health 78:516-519, 1988. Receiving Estrogen,” The Menopause and 65. Rubin, S. M., and Cummings, S.R., “Results Postmenopause, N. Pasetto, R. Paoletti, and of Bone Densitometry Affect Women’s Deci- J.L. Ambrus (eds.) (Lancaster, England: MTP sions About Taking Measures to Prevent Press, 1980.) Fractures,” Annals of Internal Medicine 74. Stumpf, P. G., and Trolice, M. P., “Com- 116(12 pt 1):990-995, 1992. pliance Problems with Hormone Replace- 66. Rud, T., “The Effect of Estrogens and Gesta- ment Therapy,” Obstetrics and Gynecologic gens on the Urethral Pressure Profile in Uri- Clinics of North America 21(2):219-229, nary Continent and Stress Incontinent 1994. Women,” Acta Obstetric et Gynecological 75. Tataryn, I. V., Lomax, P., Meldrum, D. R.. et Scandinavia 59:265-270, 1980. al., “Objective Techniques for the Assess- 67. Ryan, P. J., Harrison, R., Blake, G. M., et al., ment of Hot Flashes,” Obstetrics & Gynecol- “Compliance with Hormone Replacement ogy 57:340, 1981. Therapy (HRT) After Screening for Post 76. Thorn, M., Collins, W., and Studd, J., “Hor- Menopausal Osteoporosis,” British Journal monal Profiles in Post-Menopausal Women of Obstetrics and Gynecology 99:325-328, After Therapy with Subcutaneous Implants,” 1992. British Journal of Obstetrics and Gynecolo- 68. Sitruk-Ware, R., “Estrogen Therapy During gy 88:426-433, 1981. Menopause: Practical Treatment Recommen- 77. Thompson, B., Hunt, S. A., Pumo, D., dations,” Drugs 39(2):203-217, 1990. “Menopausal Age and Symptomalogy in 69. Speroff, T., Dawson, N., Speroff, L., et al., ”A General Practice,” Journal of Biological Sci- Risk-Benefit Analysis of Elective Bilateral ence 5:71-82, 1973. Oophorectomy: Effect on Changes in Com- 78. Transdermal HRT Investigators Group, “A pliance with Estrogen Therapy on Outcome,” Randomized Study to Compare the Effective- Appendix E Hormonal Replacement Therapy Regimens 63

ness, Tolerability, and Acceptability of Two 2:362-367, 1974. Different Transdermal Estradiol Replace- 86. “WHI Trial Discontinues Use of Unopposed ment Therapies,” International Journal of Estrogen for Study Participants Due to Fertility and Menopausal Studies 38(1):5-11, NHLBI PEPI Findings,” The Blue Sheet Jan. 1993. 18, 1995, pp. 5-6. 79. Utian, W., “Transdermal Estradiol Overall 87. Whitehead, M., Hillard, B., and Crook, D., Safety Pro file,” American Journal of Obstet- “The Role and Use of Progestins,” Obstetrics rics and Gynecology 156: 1335-1338, 1987. & Gynecology 75(4) :59S-75S, 1990. 80. Utian, W. H., “Biosynthesis and Physiologic 88. Wiklund, I., Berg, G., Hammar, M., et al., Effects of Estrogen and Pathophysiologic Ef- “Long-Term Effect of Transdermal Hormon- fects on Estrogen Deficiency: A Review,” al Therapy on Aspects of Quality of Life in American Journal of Obstetrics and Gynecol- Postmenopausal Women,” Maturitas ogy 161: 1828-1831, 1989. 14(3):225-236, 1992. 81. Varner, R. E., “Acute Indication for Hormonal 89. Wiklund, I., Karlberg, J., and Mattson, L. A., Therapy in the Menopause,” Emergency “Quality of Life of Postmenopausal Women Medicine :93-100, Oct. 30, 1990. on a Regimen of Transdermal Estradiol Ther- 82. Walsh, B., and Shiff, I., “Vasomotor apy: A Double-Blind Placebo-Controlled Flushes, ’’Annals of the New York Academy of Study,” American Journal of Obstetrics and Science 592:346-356, 1990. Gynecology 168(3 Pt 1):824-830, 1993. 83. Walsh America/PDS, PDS Beta Premarin 90. Williams, D. B., Voigt, B.J., Fu, Y. S., et al., Patient Study, Wyeth-Ayerst, 0690-9001- “Assessment of Less than Monthly Progestin 4060 (Phoenix, AZ: January 1991). Therapy in Postmenopausal Women Given 84. Weinstein, L., Bewtra, C., and Gallagher, J., Estrogen Replacement,” Obstetrics & Gy- “Evaluation of a Continuous Combined Low- necology 84(5):787-793, 1994. Dose Regimen of Estrogen-Progestin for 91. Writing Group for the PEPI Trial, “Effects of Treatment of the Menopausal Patient, ’’Amer- Estrogen or Estrogen/Progestin Regimens on ican Journal of Obstetrics and Gynecology Heart Disease Risk Factor in Postmenopausal 162:1534-1539, 1990. Women,” Journal of the American Medical 85. Wentz, W., “Progestin Therapy in Endome- Association 273(3):199-208, 1995. trial Hyperplasia,” Gynecologic Oncology Appendix F: Evidence on Breast Cancer and Hormonal Replacement Therapy F

reast cancer, after lung cancer, is the sec- suggested that estrogen could increase the inci- ond leading cause of death from cancer. dence of breast cancer, by examining the role of The American Cancer Society estimates estrogens in the development of mammary tumors that one in nine American women will de- in mice (3). Subsequently, Moolgivkar and Knud- velopB breast cancer during her lifetime (65). The son proposed that estrogen could increase the risk impact of hormone replacement therapy (HRT) on of breast cancer by increasing the rates of division the risk of breast cancer, even if small, would be and numbers of breast cells, which increases the substantial given the high baseline risk, as well as likelihood that an initiating factor (such as ioniz- the societal cost, of this illness. For this reason, ing radiation, chemicals, or viruses) will damage this question has been one of the most widely stu- cellular DNA (51). Such DNA damage, in turn, died with modern epidemiologic techniques. Un- leads to a series of errors in cell division, produc- fortunately, given the complexity of the issues ing so-called “intermediate” cells, which finally involved, no clear-cut answer is available at this results in transformed, or malignant cells. time. The hypothesis that HRT increases breast can- This appendix reviews the evidence linking cer risk is further supported by observations that HRT to an increase in the risk of breast cancer. factors that increase a woman’s exposure to estro- First, the biological plausibility of a link between gen and progestin increase her risk of breast can- HRT and breast cancer risk will be reviewed. Se- cer. Thus, early menarche (age of onset of cond, the epidemiological evidence of HRT and menstruation) and late menopause are associated breast cancer risk will be reviewed. Virtually all of with an increased breast cancer risk (75). Also, the epidemiological evidence is observational, women who have had surgical removal of the ova- consisting of case-control studies and cohort stud- ries have a lower breast cancer risk (77). There is ies. The findings, and discussion of the strengths also strong evidence that obese postmenopausal and weaknesses of the studies on which they are women are at an increased risk of breast cancer based, follow. (19). This may be because the chief source of estrogen after menopause is the conversion in fat tis- BIOLOGICAL PLAUSIBILITY sue of the hormone androstenedione, made in the The relationship of HRT and breast cancer is con- adrenal gland, to the estrogen estrone (46). sistent with a number of observations. Bittner first 165 66 Cost Effectiveness of Screening for Osteoporosis

It is uncertain whether the addition of proges- CASE-CONTROL STUDIES tins would increase the risk of breast cancer above Tables F-1 and F-2 at the end of this appendix estrogen alone. Key and Pike have reviewed the present the results of 30 case-control studies of the experimental evidence bearing on the hormonal risk of breast cancer in users of hormonal replace- control of breast cell division (42). They noted ment therapy. The fourth column of the table that breast cell division peaks during the later compares the risk of breast cancer among never phase of the menstrual cycle, corresponding to a users of hormonal replacement therapy with those progesterone peak. They concluded that, although who have ever used hormonal replacement thera- knowledge of the hormonal control of division py. Of these 30 case-control studies, five showed rates was incomplete, the available data could an increased risk of breast cancer among ever us- support two possible interpretations. ers (30,33,37,44,83). Nineteen studies demon- The first model suggests that women receiving strated no increased risk of breast cancer in ever a combination of estrogen and progestin will have users of hormonal replacement therapy. However, an increased risk of developing breast cancer over most of these latter studies found increased risks those receiving estrogen alone. This “estrogen among certain subgroups of users. The other 6 plus progesterone” model posits that estradiol, the studies either did not compare ever users to never major ovarian estrogen, itself may induce breast users (2,23,35) or did not provide statistical analy- cell division in the early phase of the menstrual sis of results (19,48,60). cycle. However, the addition of progesterone, produced in the later phase of the menstrual cycle, in- ❚ Duration duces much more cell division, perhaps because The fifth column of tables F-1 and F-2 describe the estrogen produced in the early phase of the relationship of breast cancer risk to duration of es- menstrual cycle has stimulated the formation of trogen use. Most of the studies finding no increase progesterone receptors on breast cells (42). This in the risk of breast cancer among ever users also increased cellular proliferation then places the found no correlation of breast cancer risk with breast tissue at risk for malignant change. duration of use. However, several studies, includ- The alternative model suggests that the addi- ing most studies which have found the risk of tion of a progestin will have little effect on the risk breast cancer to increase among ever users have of breast cancer associated with estrogen. This found that the risk of breast cancer increases with “estrogen alone” hypothesis is supported by ex- longer durations of use (2,6,18,23,27,28,44,63, perimental data demonstrating that progesterone 84). In addition, two studies found increased risks shows little significant cell division-stimulating among users with the greatest cumulative dose, effect. These results suggest that cell division is which is based on average daily dose multiplied induced by estradiol alone, with little contribution by the duration of use (63,82). However, Jick by progesterone (42). Such an explanation, the au- found an increased risk among ever users, but did thors note, requires a dose-response relationship not find a correlation with duration of use (80). between the plasma concentration of estradiol, Three studies found increased risk only among which peaks at the end of the early phase of the women with shorter durations of use (33,35,53). menstrual cycle, and the amount of breast cell di- In these studies, increased risk of breast cancer vision. Furthermore, such a model must account among the groups of users of the longest duration for the 4-to 5-day lag between these changes in es- may have been difficult to detect because of the tradiol concentration and the subsequent changes relatively smaller number of women in these in rates of cell division observed in breast tissue. groups. Appendix F Evidence on Breast Cancer and Hormonal Replacement Therapy 67

❚ Dose when adjusted for a number of confounding fac- The sixth column of tables F-1 and F-2 describe tors (44). Nomura found a significantly increased the relationship of breast cancer to the dose of es- risk of breast cancer among women of Japanese trogen. Bergkvist found a significantly increased ancestry whose last dose was eight or more years risk of breast cancer among users of the potent es- ago when compared with community controls but trogen diethylstilbesterol and among users of not when compared with hospital controls (53). “other” estrogens, which included users of rela- No correlation of risk with recency of use was tively high dose injectable forms of estrogen (2). found among white women (53). However, the study found no correlation of risk with the doses of oral conjugated estrogens (CE) ❚ Time Since First Use that are commonly used in hormonal replacement The sixth column of tables F-1 and F-2 present therapy (2). Hoover found a trend toward in- data on the relationship of breast cancer to the time creased risk among users of high doses of estro- since first use of HRT, or latency. Eleven of the gens (greater than 1.25 mg CE per day or the case control studies address this issue. Eight of the equivalent) (28). Hulka found an increased risk of case control studies show no correlation of risk breast cancer among users of injectable estrogens, with time of first HRT use. Ewertz found an in- but no significant increased risk among users of creased risk among women with natural meno- the highest doses of oral estrogens (greater than pause more than five years prior to breast cancer 1.25 mg CE per day or the equivalent) (33). Four diagnosis, and whose first dose of hormonal re- studies found no correlation between risk of breast placement therapy was more that 12 years ago. No cancer and dose of estrogen (37,39,40,49). similar increase in risk was found in women with natural menopause within five years of breast can- ❚ Recency cer diagnosis or women with surgical menopause The sixth column of tables F-1 and F-2 describe (18). Hulka found an increased risk among the relationship of the recency of estrogen use, or women whose first dose of hormonal replacement the time since last use of estrogen, to the risk of therapy was five to nine years ago, but no signifi- breast cancer. Thirteen case-control studies have cant increase in risk was detected in users whose examined this issue. Of those, seven found no first dose was 10 or more years ago (33). Weins- relationship between recency of estrogen use and tein found on increased risk of breast cancer only breast cancer risk. Hulka found an increased risk in women 10 to 19 years since first use (80). of breast cancer among users whose last dose was two to five years past, but no increase in risk COHORT STUDIES among users whose last dose was within the past Cohort studies of the relationship of breast cancer year or among those whose last dose was six or to use of hormonal replacement therapy are pre- more years ago (33). Kaufman found a reduced sented in tables F-3 and F-4 at the end of this ap- risk of breast cancer among women with a surgical pendix. Of the 18 studies identified by OTA, menopause whose last dose was 10 or more years seven demonstrated a statistically significant in- ago (39). The author explains that this low relative creased risk of breast cancer among users of hor- risk may be due either to chance or the fact that monal replacement therapy. Six studies did not women who have had their ovaries removed and show an increased risk of breast cancer that was are more likely to be prescribed estrogen generally statistically significant. One study found a de- for a short period of time also have a lower risk of creased risk of breast cancer among users of hor- breast cancer (39). La Vecchia found a significant- monal replacement therapy (78). Three studies ly increased risk of breast cancer among users of provided no statistical analysis of results. One estrogens whose last dose was 10 or more years study demonstrated a decreased risk of breast can- ago, but this risk was only marginally significant cer among users of estrogen with progesterone, 68 Cost Effectiveness of Screening for Osteoporosis

but this study did not control for confounding received placebo. Double-blinded randomization variables (20). A decreased risk of breast cancer was discontinued after 10 years. In the subsequent among users of estrogen and progesterone was 12 years, women were offered the choice of start- also found in the only clinical trial to examine this ing, stopping, or continuing hormone replace- issue (52) (described below); however, a lower ment therapy. During the 10-year clinical trial, risk of breast cancer in users of estrogen and pro- there were no significant differences in breast can- gesterone has not been confirmed by other studies cer incidence between the treated and the placebo (2,67). group. After 22 years of follow-up, there was a statistically significant increase in breast cancer ❚ Duration risk in never users of hormonal replacement thera- Tables F-3 and F-4 also show the effect of the py versus ever users. However, the size of this duration of use of hormonal replacement therapy study was quite small, involving 89 pairs of on breast cancer risk. Some studies were able to women, and the results are unstable. demonstrate an increase in risk of breast cancer with increasing duration of use (2,29,61). Howev- COMBINED ESTROGEN-PROGESTIN er, five studies were not able to detect an increase THERAPY AND BREAST CANCER RISK in risk with increased duration of use. Colditz It is uncertain whether the addition of progestins found an increased risk among current users of to estrogen replacement therapy would alter HRT five to 10 years, but not among users of shorter or users risk of breast cancer, as few studies have ex- longer durations (13). Schairer found an increased amined this issue. Bergkvist and colleagues ex- risk only of preinvasive (in situ) cancers with amined this issue in a study of breast cancer in a duration of ERT use (67). cohort of 23,000 women from the Uppsala Health ❚ Dose Care Region of Sweden. They found a significant increase of breast cancer in users of estrogen The few cohort studies that have looked at the alone; they also found a similar increase in risk of relationship of dose to risk of breast cancer have breast cancer among users of combined estrogen not consistently demonstrated an increased risk and progestin. The increase in risk among com- with increasing dose of estrogen (13,29,61). bined estrogen-progestin users, however, did not reach statistical significance, in part due to the rel- ❚ Recency and Time Since First Use atively small number of users of combined estro- Some studies have demonstrated an increased risk gen-progestin in the cohort. The investigators with current users of estrogen, but not with past concluded that progestins offered no protection users (13, 14,89). Other studies have found that the against the development of breast cancer (2). risk of breast cancer increases with time since first A recent cohort study by Schairer and col- use (34,35). leagues found- that users of estrogen-progestin combinations may have a higher risk of breast CLINICAL TRIALS cancer than users of estrogen alone (67). The study Only one clinical trial has examined the relation- examined the incidence of breast cancer among ship of hormonal replacement therapy to breast 49,017 postmenopausal women who had partici- cancer risk (52). Subjects were continuously hos- pated in the Breast Cancer Detection Demonstra- pitalized postmenopausal women. Treated tion Project (BCDDP). For ever users of estrogen women and control group members were matched alone, there was no increased risk of breast cancer. for age, smoking history, and medical diagnosis. For users of estrogen and progestin combinations, The treatment group received estrogen-progestin however, there was an increased risk of breast can- hormone replacement therapy. The control group cer that was of marginal statistical significance Appendix F Evidence on Breast Cancer and Hormonal Replacement Therapy 69

(relative risk 1.2 (95 percent confidence interval Women who take hormonal replacement thera- 1.0 to 1.6)). py are more likely to engage in other healthy be- All of the studies of hormone replacement and haviors. And women who are willing to take breast cancer risk, except one, are observational, hormonal replacement therapy long-term are, by so the possible impact of selection bias cannot be definition, more compliant. As has been discussed entirely ruled out. Barrett-Connor explained that in detail in Appendix I, compliant women are less it is uncertain how selection bias may affect there- likely to get heart disease, cancers, and other dis- sults of studies of HRT use and breast cancer ( 1 a). eases. Although epidemiological studies have at- Some biases may result in an exaggerated esti- tempted to statistically control for many of these mate of breast cancer risk in HRT users. For exam- sources of bias, it has not been possible to com- ple, women who take hormonal replacement pletely control for so-called compliance bias be- therapy tend to be more educated and of higher cause of its ill-defined nature. socioeconomic status than other women (la). The uncertainty about the relation between Studies have shown that women of higher socio- breast cancer risk and hormone replacement thera- economic class are at higher risk of breast cancer. py will not be resolved until we have the results of Therefore, epidemiological studies that fail to ac- a randomized clinical trial of HRT in postmeno- count for differences in socioeconomic status be- pausal women (32). Because the increase in risk of tween HRT users and nonusers may overestimate breast cancer in HRT users appears to be small, a the risk of breast cancer in HRT users. large study would be required to have sufficient Women on HRT have been found to be more statistical power to detect this small increase in likely to have mammograms (2a). Breast tumors risk. Given that the risk of breast cancer increases in HRT users are therefore more likely to be de- with duration of use, the controlled clinical trial tected. This bias may explain for the lower stage would take 10 or more years to complete. and grade of tumors detected in HRT users, and The Women’s Health Initiative, sponsored by the improved prognosis of breast cancers in HRT the National Institutes of Health, is a large long- users (la,9). (See discussion below.) term randomized clinical trial examining the ef- Other biases may result in an underestimate of fect of hormone replacement therapy on heart breast cancer risk in HRT users (la). Women who disease and osteoporosis in postmenopausal have an early menopause or surgical removal of women. (See description in Appendix I.) This trial the ovaries (oophorectomy) are more likely to be will also help to resolve many of the questions treated by their physicians with HRT. Breast can- about the relationship between hormone replace- cer risk in these women may be underestimated ment therapy and breast cancer risk and other dis- because both early menopause and oophorectomy eases affected by hormone replacement therapy. are associated with decreased risks of breast can- Problems with conducting such a study arc the cer. Women are more likely to be prescribed estro- expense of the trial and the practical problems in gen if they have menopausal symptoms, and thin conducting a clinical trial long-term. Also, be- women tend to have more severe menopausal cause sequential and continuous hormonal re- symptoms. Thin women are also at decreased risk placement therapy causes bleeding and other of breast cancer, so this is another source of bias. symptoms, both the investigator and the subject Physicians may be reluctant to prescribe HRT will become aware of their assignment, introduc- to women with benign breast disease or a family ing a source of bias. Finally, by the time the trial is history of breast cancer, another source of de- completed, new HRT regimens may be available, creased estimate of risk (la). And some physi- raising the question of whether the results of the cians will not prescribe HRT until their patient has Women’s Health Initiative apply to these new reg- had a mammogram, and if the mammogram is ab- imens. normal, will not prescribe HRT. 70 Cost Effectiveness of Screening for Osteoporosis

STAGE OF BREAST CANCER improved prognosis (70). Breast cancers from 35 AT DIAGNOSIS IN HRT USERS women who had taken HRT (mostly estrogen and VERSUS NONUSERS progestin combinations) were compared to breast There is some evidence that estrogen users devel- cancers from postmenopausal women who had op breast cancer of lower stage and grade than never taken hormones, matched for age and type breast cancers in nonusers. This maybe an artifact of breast cancer to HRT users. They found that of surveillance bias or may be because estrogen HRT users had smaller tumors, significantly less induces a less malignant form of breast cancer. In spread to lymph nodes, and had significantly low- a population-based case control study of breast er S-phase fractions (a measure of the rate of can- cancer in postmenopausal women, Brinton and cer cell division). The investigators concluded colleagues found that there was a significant trend that the small tumor size, low S-phases, and lim- of greater risk of breast cancer with increased ited nodal involvement of HRT users suggests duration of HRT use, and that this increase in risk that, despite a possibly increased risk of breast was greatest for the lowest stage tumors (6). After cancer, the mortality rate for breast cancer in HRT 10 or more years of estrogen use, the increase in users will not be increased in comparison with risk of large (greater than 1 cm) invasive breast nonusers. The investigators could not rule out the cancers was 1.29 (p less than 0.05), but the in- possibility, however, that the results may have crease in risk of small (1 cm or less) tumors and been due to better surveillance and earlier diagno- carcinoma in situ was 1.51 (p less than 0.05) and sis of breast cancer in HRT users. 1.90 (p less than 0.05), respectively. Bonnier and colleagues concluded that the low- Hunt and colleagues, in a study of a cohort of er stage of breast cancers in HRT users was not due 4544 British women receiving HRT at menopause to surveillance bias (4). The investigators com- clinics, found that, of the 40 breast cancers that de- pared 68 postmenopausal women who were re- veloped among the cohort that were identified by ceiving HRT at the time of diagnosis of breast stage, 27 (68 percent) were classified as Stage I cancer with 282 breast cancer patients who had (nonmetastatic tumors 2 cm or less) at diagnosis, not received prior HRT, and whose date and age of which is a higher proportion of early stage tumors onset of breast cancer were similar to that of the at diagnosis than expected based on comparison breast cancer patients that had received HRT. Pa- with stage at breast cancer diagnosis in the general tients who developed breast cancer during HRT population (34). The lower than expected stage of had fewer locally advanced cancers (tumors that breast cancer at diagnosis in cohort members, had extended into lymph nodes) and more well- however, could be explained by the fact that 1) the differentiated cancers. In addition, the probability average member of the cohort had been followed of metastasis-free survival tended to be better in for less than 5 years, and 2) cohort members, all of HRT users. The investigators found that the favor- whom were on HRT at recruitment, presumably able prognosis in HRT users was not likely to be did not have any previous diagnosis of breast can- due to better cancer surveillance among HRT cer at that time (57).10 users, because x-ray detection was not more fre- Squiteri and colleagues found that hormone us- quent among patients undergoing HRT. In addi- ers present with slower growing breast tumors of tion, the delay between first symptoms and earlier stage than nonusers, possibly resulting in

1 Hunt and colleagues also found that short term users of HRT had a significantly lower death rate from breast cancer than would be expected by comparison with population age-specific breast cancer death rates (observed to expected ratio = 0.55 (0.28-0.96)) (34). As Pike and colleagues explained, however, for a member of the cohort to die during the five year follow up, she had to first be diagnosed with breast cancer and then die of that disease (57). The expected number of such deaths cannot be derived straightforwardly from population age-specific death rates. Appendix F Evidence on Breast Cancer and Hormonal Replacement Therapy 71

diagnosis was slightly but not significantly short- BREAST CANCER MORTALITY IN HRT er in HRT users. USERS VERSUS NONUSERS Additional information is needed on whether There is conflicting evidence about whether an in- the addition of progestin has an impact on the creased incidence of breast cancer among HRT us- stage and grade of breast cancer related to estro- ers results in an increased rate of breast cancer gen. Schairer and colleagues, reporting on the re- deaths. A number of studies have found that estro- sults from the BCDDP cohort (described above) gen users do not have an increase in deaths from found that estrogen-progestin combinations were breast cancer. Petitti and colleagues analyzed the related to a larger risk of preinvasive (in situ) can- 26 breast cancer deaths that occurred during 13 cers (relative risk 2.3 (95 percent confidence inter- years followup of the 6,093 women in the Walnut val 1.3 to 3.9)) than estrogen alone (relative risk Creek cohort (56). The relative risk of death from 2.3 (95 percent confidence interval 1.3 to 3.9)), breast cancer for women who used HRT but not but neither estrogen or estrogen-progestin com- oral contraceptives was 0.8 (0.4 to 1.8) compared binations were related to an increased risk of inva- to women who used neither HRT nor oral contra- sive cancers (67). ceptives. Jones and colleagues found evidence that tu- Vakil and colleagues also found reduced breast mors induced by estrogen-progestin combina- cancer mortality among postmenopausal estrogen tions may have a better prognosis than tumors users in a cohort of 1,483 postmenopausal women induced by estrogen alone (38). The investigators from Ontario and Saskatchewan (78). The ratio of identified 460 perimenopausal and postmenopau- observed to expected mortality from breast cancer sal breast cancer patients hospitalized in Perth, among HRT users was 0.48 (p less than 0.01) for Western Australia, between January 1990 and De- the Ontario women and 0.45 (p less than 0.01) for cember 1991. They questioned each of the pa- the Saskatchewan women. tients about HRT use, and reviewed medical In a cohort study of 8,881 postmenopausal resi- records and pathology reports for data related to dents of Leisure World Retirement Community in breast cancer prognosis. They found that the mean Los Angeles, Henderson and colleagues found a level of estrogen and progestin receptors was low- reduction in breast cancer mortality among estro- est in users of estrogen alone highest in users of es- gen users of 0.81 (no confidence interval pro- trogen-progestin combinations, consistent with a vided) (26). Although the investigators did not better prognosis for estrogen-progestin users. have information about breast cancer stage at Levels of Cathepsin D, which is inversely related diagnosis, they suggested that estrogen users may to breast cancer risk, were highest in users of es- have less extensive cancers at diagnosis than non- trogen alone, and lowest in nonusers. The tumors users because of increased breast cancer surveil- were smallest in estrogen-progestin users, and lance among estrogen users and better health largest in users of estrogen alone, although the dif- awareness of women who use estrogens. ference was not statistically significant. There was Bergkvist and colleagues, in an analysis of sur- no significant difference in lymph node involve- vival rates in women with breast cancer in the ment of cancer between estrogen-progestin users Uppsala Health Care Region of Sweden, found and users of estrogen alone. The percentage of all that ever users of HRT had significantly greater HRT users with involved lymph nodes (23 per- survival rates than never users (2). The investiga- cent), however, was significantly lower than the tors compared survival rates in 261 breast cancer percentage of nonusers (44 percent). The authors patients who used HRT prior to diagnosis with stated that they could not rule out that this last 6,617 breast cancer patients from the same geo- finding could have been due to differences in surveillance. 72 Cost Effectiveness of Screening for Osteoporosis

graphic region who did not have any recorded use cancer mortality, either overall or in subgroups, of HRT. despite increased incidence. Information on estrogen use was obtained from Results of a study by Strickland and colleagues a regional prescription database and from a mailed suggest that the favorable survival of breast cancer questionnaire, and information on breast cancer patients who used HRT is due to surveillance bias survival was obtained from the Swedish National (73). The investigators compared the survival Cancer Registry (2). (The registry did not, howev- time between diagnosis and death of 256 postme- er, have information about tumor stage and grade.) nopausal women with breast cancer, 174 of whom The investigators found that the relative 8 year were never users of estrogens, 21 of whom were survival rate of women diagnosed with breast can- past users of estrogens, and 61 of whom were cur- cer who used hormonal replacement therapy was rently using estrogens at the time of diagnosis. In- 10 percent higher than those who had not taken formation on survival time, as well as stage of hormonal replacement therapy, which corre- breast cancer at diagnosis, was obtained from the sponded to a 40 percent reduction in excess Southwestern Oncology Group Tumor registry. mortality (2). Separate analysis of relative surviv- They found that the median time between breast al by age at diagnosis showed a significant surviv- cancer diagnosis and death was less than 84 al advantage for estrogen-treated women at each months for never users and past users of estrogens, age over 50, and was greatest for estrogen-treated and was 143 months for current users of estrogens. women 60 years old and older at diagnosis, with After controlling for stage of breast cancer at diag- an approximately 40 percent lower mortality rate nosis, however, the survival time for never users than never users with breast cancer. and past users of HRT was not significantly differ- The relative survival rates were highest for ent from current users. women who were current users of HRT at diagnosis, and the survival advantage of estrogen users was decreased with longer time between cessation CONCLUSIONS of estrogen and diagnosis, so that the survival Although the evidence on the link between estro- rates of estrogen users who had stopped taking es- gen therapy and the risk of breast cancer is based trogens more than 12 months before diagnosis almost entirely on case-control and cohort studies, was close to that of never users of estrogens (l). which cannot entirely control for biases and con- Also, the relative survival rates were best among founding factors (64), the inconsistency in results women treated with progestins combined withes- among both kinds of studies suggests that the ef- trogen during part or all of the course of HRT. fect of estrogens on breast cancer is likely to be There were several possible alternative ex- small. Indeed, when they were found, such planations of these results. First, a favorable im- associations were generally weak. Discrepancies pact of estrogens on forces of mortality other than in the results among studies are not readily ex- breast cancer, most notably heart disease, may plained by study design or implementation and have accounted for the favorable survival rates of may likely be due to chance. HRT users. Second, women who are prescribed For purposes of this model, we assumed in the HRT represent a healthy selection of the general base case that the relative risk of breast cancer population. Third, the favorable survival rates of with HRT would be a modest 1.35 times the base- HRT users maybe due to surveillance bias (2). line rate in the population of women of a certain A subsequent study of breast cancer mortality age, but the higher risk would not occur until the by the same group attempted to correct for the duration of use had exceeded 9 years. This in- “healthy user” effect (85). Despite these correc- crease in risk is consistent with the range of esti- tions, the investigators found no increase in breast mates of breast cancer risk with long-term use Appendix F Evidence on Breast Cancer and Hormonal Replacement Therapy 73

from several recent metaanalyses and epidemio- This estimate is also within the range of estimates logical reviews (Grady (relative risk 1.25 (95 per- from epidemiological reviews by Persson and col- cent confidence interval 1.04 to 1.51) for eight or leagues (relative risk 1.5 to 3.0 with 10 to 15 years more years of ERT use) (21); Steinberg (relative of use) (55), Pike (relative risk 1.75 after 20 years risk 1.3 (1.2 to 1.6 after 15 years of use) (72); Col- of ERT use) (57) and Henderson and colleagues ditz (relative risk 1.23 (95 percent confidence in- (relative risk 1.5 to 2.0 if moderate doses of CEE terval 1.08 to 1.40) for 10 or more years of are used for 10 to 20 years) (26). We have also as- estrogen use) (12); Sillero-Arenas (relative risk sumed that there was no difference in stage dis- 1.23 (1.07 to 1.42) after more than 12 years use) tribution or mortality from breast cancer in (69); Hulka (relative risk approximately 1.3 to 1.5 estrogen users. Observational studies that have with long-term use) (31); Steinberg (relative risk found better stage and grade breast cancers in HRT 1.15 to 1.29 after 10 years of CEE use) (71); Mack users have inherent risks of surveillance bias. (relative risk 1.2 at 5 years of use, increasing to 1.4 Finally, we have assumed that, once diagnosed at 10 years of use) (47); Prentice (relative risk 1.3 with breast cancer, women would be taken off for ever use, and possibly larger risks with long- HRT. There is, however, a debate in the literature term use) (59). over whether women previously treated for breast Once duration exceeds nine years, the relative cancer may start or resume HRT ( 11, 15,36,45,74). risk of breast cancer is assumed to remain elevated Proponents argue that there is little direct evidence for the rest of the woman’s lifetime. This assump- that HRT has an adverse effect on women pre- tion is consistent with the observation that breast viously treated for breast cancer who subsequent- cancer risk remains elevated in women with late ly received HRT (81). The National Cancer menopause and the hypothesis by Pike that HRT Institute recently announced the initiation of a induces a hormonal milieu similar to late meno- randomized clinical trial to determine the influ- pause (57,58). ence, if any, of HRT on the clinical course of breast Because of the great uncertainty about the mag- cancer (79). nitude and exposure pattern of risk elevation, the best case assumption was that there would be no REFERENCES increased risk of breast cancer among users of 1. Armstrong, B.K., “Oestrogen Therapy After HRT. This estimate is consistent with the metaan- the Menopause—Boon or Bane?” Medical alysis by Dupont and Page (16), who limited their Journal of Australia 148:213-214, 1988. analyses to studies of conjugated estrogens, and la. Barrett-Connor, E., “Risks and Benefits of excluded European studies where use of stronger Replacement Estrogen,” Annual Review of synthetic estrogens is common. This estimate is Medicine 43:239-251, 1992. also consistent with the metaanalyses of Khoo and 2. Bergkvist, L., Adami, H., Persson, I., et al., Chick (43) (no increase in breast cancer risk), “The Risk of Breast Cancer After Estrogen Henrich, (24) (no increased risk of breast cancer and Estrogen-Progestin Replacement,” New among ever-users of estrogens) and Armstrong England Journal of Medicine 321(5): (summary relative risk 0.96 (0.89 to 1.05) after ad- 293-297, 1989. justment for menopausal status; no effect of dura- 2a. Bergkvist, L., Tabar, L., Adami, H. O., et al., tion of use) (l). “Mammographic Parenchymal Patterns in Under the worst case, we assumed a relative Women Receiving Noncontraceptive Estro- risk of 2.0 after 9 years of therapy. This worst-case gen Treatment,” American Journal of Epide- estimate is consistent with the largest relative miology 130(3):503-510, 1989. risks of breast cancer found in cohort studies of HRT users (2,35,50,76); these large increases in risk were generally associated with long-term use. 74 Cost Effectiveness of Screening for Osteoporosis

3. Bittner, J.J., “The Causes and Control of Obstetrics and Gynecology 168:1472-1480, Mammary Cancer in Mice,” Harvey Lecture 1993. 42:221-246, 1948. 13. Colditz, G. A., Stampfer, M.J., Willett, W. C., 4. Bonnier, P., Remain, S., Giacalone, P. L., et et al., “Prospective Study of Estrogen Re- al., “Clinical and Biologic Prognostic Factors placement Therapy and Risk of Breast Cancer in Breast Cancer Diagnosed During Postme- in Postmenopausal Women,” Journal of the nopausal Hormone Replacement Therapy,” American Medical Association 264:2 Obstetrics & Gynecology 85(1):11-17, 1995. 648-2653, 1990. 5. Boston Collaborative Drug Surveillance Pro- 14. Colditz, G. A., Stampfer, M.J., Willett, W. C., gram, “Surgically Confirmed Gallbladder et al., “Type of Postmenopausal Hormone Disease, Venous Thromboembolism, and Use and Risk of Breast Cancer: 12-Year Fol- Breast Tumors in Relation to Postmenopausal low-Up from the Nurses’ Health Study,” Can- Estrogen Therapy,” New England Journal of cer Causes and Control 3:433-439, 1992. Medicine 290(1):15-19, 1974. 15. Creasman, W. T., “Estrogen Replacement 6. Brinton, L. A., Hoover, R., and Fraumeni, Therapy: Is Previously Treated Cancer a Con- J. F., Jr., “Menopausal Oestrogens and Breast traindication?” Obstetrics & Gynecology Cancer Risk: An Expanded Case-Control 77(2):308-312, 1991. Study,” British Journal of Cancer 54(5): 16. Dupont, W. D., and Page, D.L., “Menopausal 825-832, 1986. Estrogen Replacement Therapy and Breast 7. Burch, J., Byrd, B., Jr., and Vaughn, W., “The Cancer,” Archives of Internal Medicine Effects of Long-term Estrogen on Hysterecto- 151:67-72, 1991. mized Women,” American Journal of Obstet- 17. Dupont, W. D., Page, D. L., Rogers, L. W., et rics and Gynecology 188:778-782, 1974. al., “Influence of Exogenous Estrogens, Pro- 8. Bush, T. L., Cowan, L. D., Barrett-Connor, E., liferative Breast Disease, and Other Variables et al., “Estrogen Use and All-Cause Mortal- on Breast Cancer Risk,” Cancer 63(5): ity: Preliminary Results From the Lipid Re- 948-957, 1989. search Clinics Program Follow-Up Study,” 18. Ewertz, M., ’’Influence of Non-Contraceptive Journal of the American Medical Association Exogenous and Endogenous Sex Hormones 249(7):903-906, 1983. on Breast Cancer Risk in Denmark,” Interna- 9. Bush, T. L., Helzlsouer, K., “Estrogen Re- tional Journal of Cancer 42(6): 832-838, placement Therapy and Risk of Breast Can- 1988. cer” (letter), Journal of the American Medical 19. Folsom, A. R., Kaye, S. A., and Prineas, R. J., Association 266(10):1357, 1991. “Increased Incidence of Carcinoma of the 10. Casagrande, J., Gerkins, V., Henderson, B.E., Breast Associated with Abdominal Adiposity et al., “Exogenous Estrogens and Breast Can- in Postmenopausal Women,” American Jour- cer in Women with Natural Menopause,” nal of Epidemiology 131(5):794-803, 1990. Journal of the National Cancer Institute 20. Gambrell, R., Bagnell, C., and Greenblatt, 56(4):839-841, 1976. R., “Role of Estrogens and Progesterone in 11. Cobleigh, M. A., Berris, R. F., Bush, T., et al., the Etiology and Prevention of Endometrial “Estrogen Replacement Therapy in Breast Cancer” (review), American Journal of Ob- Cancer Survivors,” Journal of the American stetrics and Gynecology 146:696-707, 1983. Medical Association 272(7):540-545, 1994. 21. Grady, D., Rubin, S. M., Petitti, D. B., et al., 12. Colditz, G.A., Egan, K. M., and Stampfer, “Hormone Therapy to Prevent Disease and M.J., “Hormone Replacement Therapy and Prolong Life in Postmenopausal Women,” Risk of Breast Cancer: Results from Epide- Annals of Internal Medicine 117(12): miologic Studies,” American Journal of 1016-1037.1992. Appendix F Evidence on Breast Cancer and Hormonal Replacement Therapy 75

22. Hammond, C. B., Jelovsek, F. K., Lee, K.L., et of the American Medical Association 252(1): al., “Effects of Long-Term Estrogen Replace- 81-82, 1984. ment Therapy. II. Neoplasia,” American Jour- 33. Hulka, B. S., Chambless, L.E., Deubner, nal of Obstetrics and Gynecology 133(5): D. C., et al., “Breast Cancer and Estrogen Re- 537-547, 1979. placement Therapy,” American Journal of 23. Harris, R.E., Namboodiri, K. K., Wynder, Obstetrics and Gynecology 143:638-644, E. L., “Breast Cancer Risk: Effects of Estro- 1982. gen Replacement Therapy and Body Mass,” 34. Hunt, K., Vessey, M., and McPherson, K., Journal of the National Cancer Institute “Mortality in Cohort of Long-Term Users of 84(20):1575-1582, 1992. Hormone Replacement Therapy: An Updated 24. Heinrich, J. B., “The Postmenopausal Estro- Analysis,” British Journal of Obstetrics and gen/Breast Cancer Controversy” (review), Gynecology 97(12): 1080-1086, 1990. Journal of the American Medical Association 35. Hunt, K., Vessey, M., McPherson, K., et al., 268(14): 1900-1902, 1992. “Long-Term Surveillance of Mortality and 25. Henderson, B., Paganini-Hill, A., and Ross, Cancer Incidence in Women Receiving Hor- R., “Decreased Mortality in Users of Estrogen mone Replacement Therapy,” British Journal Replacement Therapy,” Archives of Internal of Obstetrics and Gynecology 94:620-635, Medicine 151:75-78, 1991. 1987. 26. Henderson, B. E., “The Cancer Question: An 36. Isaacs, C.J., and Swain, S. M., “Hormone Re- Overview of Recent Epidemiologic and Ret- placement Therapy in Women with a History rospective Data,” American Journal of Ob- of Breast Carcinoma,” Hematology and stetrics and Gynecology 161(6 pt 2): Oncology Clinics of North America 8(1): 1859-1864, 1989. 179-195, 1994. 27. Hiatt, R. A., Bawol, R., Friedman, G. D., et 37. Jick, H., Walker, A. M., Watkins, R. N., et al., al., “Exogenous Estrogens and Breast Cancer “Replacement Estrogens and Breast Cancer,” After Oophorectomy,” Cancer 54(l): American Journal of Epidemiology 112(5): 139-144, 1984. 586-594, 1980. 28. Hoover, R., Glass, A., Finkle, W.D., et al., 38. Jones, C., Ingram, D., Mattes, E., et al., “The “Conjugated Estrogens and Breast Cancer Effect of Hormone Replacement Therapy on Risk in Women,” Journal of the National Prognostic Indices in Women with Breast Cancer Institute 67(4):815-820, 1981. Cancer,” Medical Journal of Australia 29. Hoover, R., Gray, L.A., Cole, P., et al., 161(2):106-110, 1994. “Menopausal Estrogens and Breast Cancer,” 39. Kaufman, D.W., Miller, D. R., Rosenberg, L., New England Journal of Medicine 295(8): et al., “Noncontraceptive Estrogen Use and 401-405, 1976. the Risk of Breast Cancer,” Journal of the 30. Horowitz, M., Need, A.J., Philcox, J.C., et American Medical Association 252:63-67, al., “Effect of Calcium Supplementation on 1984. Urinary Hydroxyproline in Osteoporotic 40. Kaufman, D.W., Palmer, J.R., de Mouzon, J., Postmenopausal Women,” American Journal et al., “Estrogen Replacement Therapy and of Clinical Nutrition 39:857-859, 1984. the Risk of Breast Cancer: Results from the 31. Hulka, B. S., “Links Between Hormone Re- Case-Control Surveillance Study,” Ameri- placement Therapy and Neoplasia,” Fertility can Journal of Epidemiology 134(12): and Sterility 62(6 Suppl. 2): 168S-175S, 1375-1385, 1991. 1994. 41. Kelsey, J.L., Fischer, D. B., Holford, T. R., et 32. Hulka, B. S., “When is the Evidence for ‘No al., “Exogenous Estrogens and Other Factors Association’ Sufficient” (editorial), Journal in the Epidemiology of Breast Cancer,” Jour- 76 Cost Effectiveness of Screening for Osteoporosis

nal of the National Cancer Institute 67(2): Carcinogenesis,” Journal of the National 327-333, 1981. Cancer Institute 66: 1037-1052, 1981. 42. Key, T.J.A., and Pike, M. C., “The Role of 52. Nachtigall, M.J., Smilen, S. W., Nachtigall, Oestrogens and Progestagens in the Epide- R.D., et al., “Incidence of Breast Cancer in a miology and Prevention of Breast Cancer,” 22-Year Study of Women Receiving Estro- European Journal of Cancer and Clinical gen-Progestin Replacement Therapy,” Ob- Oncology 24(1):29-43, 1988. stetrics & Gynecology 80(5):827-830, 1992. 43. Khoo, S.K., and Chick, P., “Sex Steroid Hor- 53. Nomura, A. M., Kolonel, L. N., Hirohata, T., mones and Breast Cancer: Is There a Link et al., “The Association of Replacement Es- with Oral Contraceptives and Hormone Re- trogens with Breast Cancer,” International placement Therapy?” Medical Journal of Journal of Cancer 37(1):49-53, 1986. Australia 156(2):124-132, 1992. 54. Palmer, J. R., Rosenberg, L., Clarke, E. A., et 44. La Vecchia, C., Negri, E., Franceschi, S., et al., “Breast Cancer Risk after Estrogen Re- al., “Non-Contraceptive Oestrogens and placement Therapy: Results from the Toronto Breast Cancer: An Update” (letter), Interna- Breast Cancer Study,” American Journal of tional Journal of Cancer 50:161-162, 1992. Epidemiology 134(12): 1386-1395, 1991. 45. Lobo, R. A., “Hormone Replacement Thera- 55. Persson, L, Adami, H. O., and Bergkvist, L., py: Oestrogen Replacement After Treatment “Hormone Replacement Therapy and the for Breast Cancer?” Lancet 341:1313-1314, Risk of Cancer in the Breast and Reproductive 1993. Organs: A Review of Epidemiological Data,” 46. Longcope, C., Pratt, J. H., Schneider, S. H., et HRT and Osteoporosis, J.O. Drife and J.W. al., “Aromatization of Androgens by Muscle Studd (eds.) (New York, NY: Springer-Ver- and Adipose Tissue in Vivo,” Journal of Clin- lag, 1990). ical Endocrinology and Metabolism 46: 56. Petitti, D., Perlman, J., and Sidney, S., “Non- 146-152, 1978. contraceptive Estrogens and Mortality: Long- 47. Mack, T. M., and Ross, R.K., “A Current Per- Term Follow-Up of Women in the Walnut ception of HRT Risks and Benefits,” Osteo- Creek Study,” Obstetrics and Gynecology porosis: Physiological Basis, Assessment, 70(3, part 1):289-292, 1987. and Treatment, H.F. DeLuca and R. Mazess 57. Pike, M., Berstein, L., and Ross, R., “Breast (eds.) (Shannon, Ireland: Elsevier Science Cancer and Hormone Replacement Therapy,” Publishing, Co., Inc., 1990). Lancet 335:297-298, 1990. 48. Mack, T. M., Henderson, B.E., Gerkins, V.R., 58. Pike, M. C., “Reducing Cancer Risk in Wom- et al., “Reserpine and Breast Cancer in a Re- en Through Lifestyle-Mediated Changes in tirement Community,” New England Journal Hormone Levels,” Cancer Detection and Pre- of Medicine 292(26):1366-1371, 1975. vention 14(6):595-607, 1990. 49. McDonald, J.A., Weiss, N. S., Daling, J.R., et 59. Prentice, R. L., “Epidemiologic Data on al., “Menopausal Estrogen Use and the Risk Exogenous Hormones and Hepatocellular of Breast Cancer,” Breast Cancer Research Carcinoma and Selected Other Cancers,” Pre- Treatment 7(3):193-199, 1986. ventive Medicine 20:38-46, 1991. 50. Mills, P. K., Beeson, W.L., Phillips, R.L., et 60. Ravnihar, B., Seigel, D. G., and Lindtner, J., al., “Prospective Study of Exogenous Hor- “An Epidemiologic Study of Breast Cancer mone Use and Breast Cancer in Seventh-Day and Benign Breast Neoplasias in Relation to Adventists,” Cancer 64(3):591-597, 1989. the Oral Contraceptive and Estrogen Use,” 51. Moolgavkar, S. H., and Knudson, A.G., Jr., European Journal of Cancer 15(4):395-405, “Mutation and Cancer: A Model for Human 1979. Appendix F Evidence on Breast Cancer and Hormonal Replacement Therapy 77

61. Risch, H.A., and Howe, G. R., “Menopausal 71. Steinberg, K. K., Smith, S. J., Thacker, S. B., Hormone Usage and Breast Cancer in Sas- et al., “Breast Cancer Risk and Duration of katchewan: A Record-Linkage Cohort Estrogen Use: The Role of Study Design Study,” American Journal of Epidemiology in Meta-Analysis,” Epidemiology 5(4): 139(7):670-681, 1994. 415-421, 1994. 62. Rohan, T. E., and McMichael, A. J., “Non- 72. Steinberg, K. K., Thacker, S. B., Smith, S. J., ContraceptiveExogenous Oestrogen Therapy et al.. “A Meta-Analysis of the Effect of Estro- and Breast Cancer,” Medical Journal of Aus- gen Replacement Therapy on the Risk of tralia 148(5):217-221, 1988. Breast Cancer,” Journal of the American 63. Ross, R. H., Paganini-Hill, A., Gerkins, V. R., Medical Association 265(15): 1985- 1990, et al., “A Case-Control Study of Menopausal 1991. Estrogen Therapy and Breast Cancer,” Jour- 73. Strickland, D. M., Gambrel], R. D., Butzin, nal of the American Medical Association C. A., et al., “The Relationship Between 243(16):1635-1639, 1980. Breast Cancer Survival and Prior Postmeno- 64. Rossouw, J. E., and Harlan, W. R., “Postme- pausal Estrogen Use,” Obstetrics & Gynecol- nopausal Estrogen and the Risk of Breast ogy 80(3, part 1):400-404, 1992. Cancer: The Need for Randomized Trials” 74. Theriault, R. L., and Sellin, R. V., “A Clinical (editorial), Annals of Epidemiology 4(3): Dilemma: Estrogen Replacement Therapy in 255-256, 1994. Postmenopausal Women with a Background 65. Rovner, S., “Risk of Breast Cancer Rises to of Primary Breast Cancer” (review), Annals 1 in 9 American Women,” The Washington of Oncology 2:709-717, 1991. Post, Health Section, p.5, Feb. 5, 1991. 75. Thomas, D. B., “Do Hormones Cause Breast 66. Sartwell, P.E., Arthes, F. G., and Tonascia, Cancer?” Cancer 53:595-604, 1984. J.A., “Exogenous Hormones, Reproductive 76. Thomas, D. B., Persing. J. P., and Hutchinson, History, and Breast Cancer,” Journal of the W. B., “Exogenous Estrogens and Other Risk National Cancer Institute 59(6):1589-1592, Factors for Breast Cancer in Women with Be- 1977. nign Breast Diseases,” Journal of the Nation- 67. Schairer, C., Byrne, C., Keyl, P., et al., al Cancer Institute 69(5): 1017-1025, 1982. “Menopausal Estrogen and Estrogen-Proges- 77. Ttichopoulos, D., MacMahon, B., and Cole, tin Replacement Therapy and Risk of Breast P., “The Menopause and Breast Cancer,” Cancer (United States),” Cancer Causes and Journal of the National Cancer Institute Control 5:491-500, 1994. 48:605-613, 1972. 68. Sherman, B., Wallace, R., and Bean, J., “Es- 78. Vakil, D. V., Morgan, R.W., and Halliday, M., trogen Use and Breast Cancer: Interaction “Exogenous Estrogens and Development of with Body Mass,” Cancer 51(8): 1527-1531, Breast and Endometrial Cancer,” Cancer De- 1983. tection and Prevention 6(4-5):415-424, 1983. 69. Sillero-Arenas, M., Delgado-Rodriguez, M., 79. Vassilopoulou-Sellin, R., and Theriault, Rodigues-Canteras, R., et al., “Menopausal R. L., “Randomized Prospective Trial of Es- Hormone Replacement Therapy and Breast trogen-Replacement Therapy in Women with Cancer: A Meta-Analysis,” Obstetrics & Gy- a History of Breast Cancer.” Monograph of necology 79(2):286-294, 1992. the National Cancer Institute 16: 153-159, 70. Squitieri, R., Tartter, P. I., Ahmed, S., et al., 1994. “Carcinoma of the Breast in Postmenopausal 80. Weinstein, A.L., Mahoney, M. C., Nasca, Hormone User and Nonuser Control Groups,” P. C., et al., “Oestrogen Replacement Therapy Journal of the American College of Surgeons and Breast Cancer Risk: A Case-Control 178(2):167-170, 1994. 78 Cost Effectiveness of Screening for Osteoporosis

Study,” International Journal of Epidemiolo- lman, S. D., “The Epidemiology of Breast gy 22(5):781-789, 1993. Cancer in 785 United States Caucasian 81. Wile, A.G., Opfell, R.W., and Margileth, Women,” Cancer 41(6):2341-2354, 1978. D. A., “Hormone Replacement Therapy in 84. Yang, C.P., Daling, J.R., Band, P. R., et al., Previously Treated Breast Cancer Patients,” “Noncontraceptive Hormone Use and Risk of American Journal of Surgery 165:372-375, Breast Cancer,” Cancer Causes and Control 1993. 3:475-479, 1992. 82. Wingo, P.A., Layde, P. M., Lee, N. C., et al., 85. Yuen, J., Persson, I., Bergkvist, L., et al., “The Risk of Breast Cancer in Postmenopau- “Hormone Replacement Therapy and Breast sal Women Who Have Used Estrogen Re- Cancer Mortality in Swedish Women: Results placement Therapy,” Journal of the American After Adjustment for ‘Healthy Drug-User Ef- Medical Association 257(2):209-215, 1987. fect’,” Cancer Causes and Control 83. Wynder, E.L., MacCornack, F.A., and Stel- 4(4):369-374, 1993. Relationship of breast Relationship of Relationship of breast cancer to dose, Number of oases breast canoer to cancer to duration of recency, and latency a,b a,b a,b Author Description of cases and controls and controls estrogen USe estrogen USe of estrogen USe

Boston Cases and controls were consecutive 51 breast cancer 9% of cases were estrogen “Duration of use in the cases Collaborative postmenopausal patients, ages 45 to cases; 774 controls users; 8% of controls were of breast cancer . . was Drug 64 years, admitted to the general estrogen users; the similar to that of control Surveillance medicine and surgical wards of 24 difference was not users. ” Program hospitals in the Greater Boston area in statistically significant (1974) 1972. Cases had surgically confirmed breast cancer. Controls were postmenopausal women who were admitted to these hospitals with acute illnesses, elective surgery, or o orthopedic treatment. Patients were interviewed during admission. Sartwell Cases were women 20 to 74 years of 284 cases (65,8% Adjusted RR: 0.82 (0.6-1 .2)* <6 mo.: 0.87 (1977) age with carcinoma of the breast post menopausal) 6-11 mo.: 0.61 admitted to Johns Hopkins Hospital (1 9.7% 1-1.9 yrs.: 1.40 ● adjusted for age, race, marital between 1969 and 1972. Controls noncontraceptive 2-4.9 yrs.: 0.70 status, menopausal history, and were chosen from among other estrogen users); 367 pregnancy history. >5 yrs.: 0.62 patients except those from the controls (76.8% None of the adjusted relative obstetric or gynecology services. All postmenopausal) risks were significantly subjects were given a questionnaire by (26.7% different from unity. an interviewer. noncontraceptive estrogen users) Wynder (1 978) Cases were pre- and postmenopausal 785 cases (267 34.1 % of postmenopausal white women selected from seven postmenopausal); cases and 36.8% of hospitals in New York City, with 2,231 controls (630 postmenopausal controls diagnosis of breast cancer between postmenopausal) used estrogen 1969 and 1975. Controls were white (nonsignificant difference). women admitted to the surgical services of these same hospitals during the same period. All subjects were interviewed. Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency Description of cases and controls and controls estrogen us&b estrogen use a,b of estrogen use a,b Author —.——— Ravnihar Cases and controls were women ages 374 breast cancer Ages 50-64. cases 11.4% Ages 50 to 64: Current users (ages 50 to (1979) 15 to 64 years selected from patients cases (184 were controls 11.7% <24 mo.: 64). 1/1 84 breast cancer admitted to a Slovenian hospital. ages 50 to 64), 748 No tests of statistical 11/1 84 breast cancer cases cases Cases were women admitted for breast cancer significance were 30/368 controls 5/368 controls aspiration or biopsy of malignant or controls (368 were performed. >24 mo.: Past use: benign breast diseases. Two controls ages 50 to 64) 3/1 84 breast cancer cases 20/1 84 breast cancer cases from other hospital services were 8/368 controls 38/368 controls selected for each case and matched unknown duration 7 cases, for age and date of admission. 5 controls Interviews were conducted between 1972 and 1974.

Jick (1980) Cases were postmenopausal women, 97cases (39% Natural menopause: 3.4 Duration had no effect on Dose had no effect on risk of ages 45 to 64, identified from a current estrogen (90% 2.1 -5.6) for current risk of breast cancer. breast cancer. prepaid health care organization’s users); 139 controls users (last use within 12 (Group Health Cooperative of Puget (37% current users) months of date of diagnosis) Sound) records as having the versus nonusers, diagnosis of breast cancer between Hysterectomized women. 1975 and 1978. Controls were 1,1 (90% 0.7-1 .9) postmenopausal women ages 45 to 64 years matched for age with cases and hospitalized about the same time. Information on cases and controls was obtained from interviews and medical and pharmacy records. Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency Author Description of cases and controls and controls estrogen use a,b estrogen use a,b of estrogen use a,b

Kelsey (1 981 ) Cases were women ages 45 to 74 330 cases (9% One or both ovaries intact At least one ovary intact years admitted to Connecticut users); 1,348 controls (pre- and postmenopausal). (pre- and postmenopausal). hospitals between 1977 and 1979 with (10% users) O.R. 0.9 (0.6-1 .2) 1-49 mg - months. O.R. 0.9 newly diagnosed breast cancer. Both ovaries removed: O.R. (no c.i.) Controls were women of the same age 0.9 (0.5-1 .5) >50 mg - months: O.R. 0.6 span admitted to other surgical (test for trend: p= 0.08) services (excluding gynecology) Both ovaries removed: between 1977 and 1979. All cases and 1-49 mg - months: O.R. 0.7 controls were interviewed. (no c.i.) >50 mg - months: O.R. 1.0 (test for trend: P= O.88) “For estrogen-replacement therapy, there is a nonsignificant decrease of less than 5 percent in risk for breast cancer with each year of use. ” Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency a,b a,b a,b Author Description of cases and controls and controls estrogen USe estrogen USe of estrogen use 0 Hulka (1 982) Cases were postmenopausal women 163 cases (52 users), Estrogen use was defined Natural menopause. Natural menopause. admitted to two North Carolina 372 hospital controls as use greater than 6 0,5-3 yrs.: 2.1 (p < 0.05) <0,625 mg conjugated hospitals between 1977 and 1978 with (90 users), 737 months. (comm. controls); 2.6 (p < estrogen (or equivalent). 1.9 a diagnosis of breast cancer. Hospital community controls Ever use: 0.05) (hosp. controls) (NS) (comm. controls), 1.8 controls were postmenopausal women (171 users) natural menopause 1.8 (p < 4-9 yrs. :1.5 (NS) (comm. (NS) (hosp. controls) admitted to these same hospitals with 0.05) (comm. controls); 1.7 controls); 1.6 (NS) (hosp. >0,625 mg: 1,0 (NS) problems that were not gynecologic or (p < 0.05) (hosp. controls) controls) (comm. controls), 0.8 (NS) referable to the breasts. Controls were 10+ yrs.: 1.7 (NS) (comm. Ever use (oral estrogens (hosp. controls) matched to cases by age, race, date controls); 0.7 (NS) (hosp. only, excluding users of Injectable, 4,4 (p < 0.05) of admission, and hospital. controls) injectable estrogens): 1.3 (comm. controls) 4.0 (p < Postmenopausal community controls (NS) (comm. controls); 1.2 0.05) (hosp. controls) were obtained from hospital referral regions. All study subjects were (NS) (hosp. controls) Recency (time since last interviewed. Surgical menopause. 1.3 use): Natural menopause. (NS) (comm. controls); 1.2 O-1 yr.: 1.6 (NS) (comm.); (NS) (hosp. controls) 1,3 (NS) (hosp.) 2-5 yrs., 2,2 (NS) (comm.), 3.2 (p < 0.05) (hosp.) 6+ yrs.: 1.8 (NS) (comm.); 1.8 (NS) (hosp.) Latency (time since first use): Natural menopause: 0.5-4 yrs.: 1.2 (NS) (comm.); 1,7 (NS) (hosp.) 5-9 yrs.: 2.4 (p < 0.05) (comm.), 3.1 (p< 0.05) (hosp.) 10-14 yrs.: 2.1 (NS) (comm.): 1.3 (NS) (hosp.) 15+ yrs.: 1,5 (NS) (comm.); 1.4 (NS) (hosp.)

I Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency Author Description of cases and controls and controls estrogen use a,b estrogen use a,b of estrogen use a,b Sherman Cases were white patients seen for 113 cases (32% Estrogen use was defined (1983) breast cancer surgery at the University users), 113 controls as use for more than one of lowa Hospitals between 1974 and (45% users) month, 1978. Controls were patients without Unadjusted RR 0.71 history of cancer from the general (0.34-0.1 1) medicine and surgery wards, matched Adjusted RR* 0.55 for age and hospital payment category. (p= 0.029) A trained interviewer administered a

questionnaire to all subjects. *adjusted for weight and height Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency a,b a,b a,b Author Description of cases and controls and controls estrogen use estrogen USe of estrogen use

Horowitz Cases and controls were 257 breast cancer Estrogen use was defined (1984) postmenopausal women, age 45 or cases, including 150 as at least 0.3 mg/day of older, evaluated at Yale New Haven breast cancer cases estrogen for at least three Hospital, Connecticut, between 1976 diagnosed by months. and 1979. Patients with clinical mammography, and Group 1: O.R. 0.4 (0.3-0.7) conditions making them unlikely to 107 breast cancer have received postmenopausal cases diagnosed by Group 2: O.R. 0.5 (0.3-0.8) estrogens were excluded from control biopsy. Group 3: O.R. 0.8 (0.5-1.4) groups chosen to reduce the likelihood Control group 1: 150 Group 4: O.R. = 0.9 (0.5-1 .7) of ascertainment bias and detection (normal by when only those medical bias. Four case control groups were mammography) records which had compared. Control group 2: 150 specific notations about Group 1.150 breast cancer patients (benign breast use or nonuse of initially diagnosed by mammography disease by estrogens were used; O.R. were compared to 150 women with mammography) = 3.3 (2.2-5.0) when those mammographically normal breasts. medical records with no Control group 3: 107 Group 2: same 150 breast cancer specific notations about (histologically normal patients were matched with 150 estrogen use were biopsy) women with benign breast disease by classified as nonusers. mammography. Control group 4: 257 (hospitalized patients Group 3:107 breast cancer patients with initial diagnosis by breast biopsy with other diagnoses) were matched with 107 control patients (conventional control group) with histologically benign disease. Group 4:257 breast cancer patients were matched to 257 control patients chosen from the medical or surgical wards of the hospital (conventional control group). Data were obtained from hospital and physician office records.

Kaufman Cases and controls were from several 1,610 cases Estrogen use was defined Duration of conjugated Natural menopause: (1984) hospitals in the United States and (925 as use at least 18 months estrogen use: <1.25 mg: 1.2 (0.5-2.5) Canada. Cases were pre- and post- postmenopausal), prior to admission, Natural menopause: >1.25 mg: 0.7 (0.3-1.5) menopausal women younger than 70 1,606 controls Pre- and postmenopausal <1 year: 0.9 (0.5-1 .5) Hysterectomy: years of age (median age 51 ) (1 ,127 ever use of 1-4 years: 0.9 (0.5-1 .5) <1.25 mg: 0.7 (0.2-3.3) admitted to these hospitals between postmenopausal) noncontraceptive estrogens: 5-9 years: 0.7 (0.4-1 .5) >1.25 mg: 0.4 (0.2-1 .0) 1976 and 1981 with the diagnosis of conjugated estrogens: 0.9 >10 years: 1.3 (0.6-2.8) Hysterectomy and breast cancer made no more than six (0.7-1.1) Hysterectomy only: oophorectomy: months prior to admission. Controls nonconjugated estrogens: <1 year: 1.3 (0.5-3.3) <1.25 mg: 2.0 (0.6-6.4) were women less than 70 years of age 0.8 (0.6-1 .1) 1-4 years: 1.2 (0.5-2.8) >1.25 mg: 0.5 (0.3-1 .0) (median age 51) who were admitted to all estrogens: 0.8 (0.5-1 .2) 5-9 years: 0.7 (0.2-1 .7) these hospitals for malignant Natural menopause: Premenopausal use of >10 years: 0.3 (0.1-1 .0) conditions judged to be unrelated to —all use within 10 yrs. 0 conjugated estrogen: 1.3 noncontraceptive estrogen, and with Hysterectomy and before admission: 1.0 age within one decade of control (0.6-2.9) oophorectomy: (0.6-1 .5) subjects. Postmenopausal use of <1 year: 0.4 (0,1 -1 .0) —all use ending >=10 yrs. conjugated estrogen: 0.8 1-4 years: 0.8 (0.4-1 .6) before admission: 0.5 (0.7-1 .1) 5-9 years: 1.1 (0.5-2.3) (0.3-1.1) >10 years: 0.5 (0.2-1 .0) —use spanning 10 yrs. before admission: 1.4 (0.8-2.4) —last use within 10 yrs. plus current use (use within past year): 0.6 (0.3-1 .2) Hysterectomy and oophorectomy: —use within 10 yrs. before admission: 1.0 (0.5-1 .8) —use ending >= 10 yrs. before admission: 0.3 (0.1 -0.8)C —use spanning 10 yrs. bef ore admission: 0.5 (0.3-1 .0) —use within 10 yrs. plus current use (use within past year): 1.2 (0.5-2.6) Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency a,b a,b a,b of estrogen USe Author Description of cases and controls and controls estrogen use estrogen USe Whites: Recency (time since last Nomura Cases were white women or women of 161 white cases; 161 Japanese: use), Whites: Japanese ancestry, ages 45 to 74 hospital controls, 159 1.1 (0.7-1.6) compared with 1-12 me.: 0.9 (0.4-2.0) (1986) <8 yrs.: 1.3 (0,6-2.8) (average age 57 for Japanese, 61 for neighborhood neighborhood controls, community controls; 0.5 community; 1.2 (0.6-2.6) white), diagnosed with breast cancer controls 1.0 (0.6-1 .4) compared with (0.2-1.0) hospital controls hospital controls hospital between 1975 and 1980 in one of 181 Japanese cases; 13-72 me.: 0.7 (0.4-1.5) seven hospitals in Oahu, Hawaii. One 183 hospital controls, Whites: community controls; 1.4 8-16 yrs.: 0.7 (0.3-1.4) hospital control was selected for each 181 neighborhood 0.9 (0.5-1.3) compared with (0.6-2.9) hospital controls community, 0.5 (0.2-1 .0) hospital case, matched for sex, race, age, controls neighborhood controls; 73+ mo.: 1.3 (0.7-2.6) Oahu residency, time of 0.7 (0.4-1 .1) compared with community controls; 0.8 16 + yrs.: 1.1 (0.6-2.2) hospitalization, and hospital. Controls hospital controls (0.4-1 .6) hospital controls community; 0.8 (0.4-1.5) with a diagnosis of cancer were Japanese. hospital excluded. One neighborhood control Japanese: was selected for each case, matched 1-12 me.: 2.4 (1 .3-4.7) community controls; 1.0 <8 yrs.: 1,0 (0.5-1 .9) for sex, race, age and Oahu residency. (0.6-1.8) hospital controls community; 0.8 (0.4-1.7) Patients were interviewed. hospital 13-72 mo.: 0.7 (0.3-1.5) community controls; 0.6 8-16 yrs.: 2.3 (1 .1-4.7) (0.3-1 .2) hospital controls community; 1.1 (0.6-1 .9) hospital 73+ mo.: 1.9 (0.8-4.4) community controls; 1.2 16 + yrs.: 2.6 (1 .1-6.1) (0.6-2.4) hospital controls community, 1.0 (0.5-2.0) hospital

< 1 year: 1,3 (1 ,0-1.8) <1.25 mg: 0.8 (0.4-1 .5) Kaufman Followup on cohort described in 1,686 cases (18Y0 RR 1.2 (1 .0-1.4) 1-4 years: 1.2 (0.9-1.6) > 1.25 mg: 1,2 (0.7-2.0) (1991) Kaufman (1987). Cases were users); 2,077 controls Type of estrogen: (17% users) 5-9 years: 1.4 (0.9-2.2) mixed mg: 1.6 (0.6-4.0) postmenopausal women ages 40 to 69 unopposed: 1.2 (1 .0-1.4) 10-14 years: 1.0 (0.6-1.6) Recency (time since last years (median age 59 years) conjugated: 1.3 (1 .0-1.6) >15 years. 0.9 (0.4-1 .9) use): diagnosed with breast cancer between other estrogen. 1.3 (0.6-2.8) <12 mo.: 1.1 (0,7-1.6) 1980 and 1986 and hospitalized in one opposed by progestin: 1.7 12-35 mo.: 1.3 (0.8-2.4) of several hospitals in seven U.S. and (0.9-3.3) Canadian cities. Controls were 36-59 mo.: 0.8 (0.4-1 .4) postmenopausal women ages 40 to 69 60-119 mo.: 1.5 (1 ,0-2.2) years (median age 59 years) >120 mo.: 1.2 (0,9-1 .6) hospitalized with malignant and Number of years since 5 nonmalignant nongynecological years of use. conditions judged to be unrelated to <5 yrs.: 2.0 (0.9-4.2) estrogen use. Data were obtained form 5-9 yrs.: 1.3 (0.9-2.1) interviews and hospital records. 10-14 yrs.: 0.5 (0.2-1 .0) >15 yrs.: 1.3 (0.7-2.4)

II Appendix F Evidence on Breast Cancer and Hormonal Replacement Therapy 187 Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency Author Description of cases and controls and controls estrogen usea,b estrogen use a,b of estrogen use a,b

La Vecchia European multicenter study (1986 3,037 cases (4.9% RR 1.4 (1.1-1.8) adjusted for <3 years, RR 1,2 (0.9-1 ,7) Recency (time since last (1992) study updated to Dec. 1990); cases users); 2,569 controls age adjusted for age use). were ages 26 to 74 years (median age (3.5% users) Adjusted RR * 1.3 (1 .0-1 .8) Adjusted RR 1.2 (0.9-1 .7) <10 yrs.: RR 1,3 (0.9-1 .5) 56 years) with histologically confirmed (This represents a lower risk >3 years, RR 1,5 (0,9-2.5) adjusted for age breast cancer admitted to one of estimate than the 1986 adjusted for age Adjusted RR: 1.2 (0.8-1 .8) >10 yrs.: RR 1.5 (1.1 -2,3) several hospitals in Northern Italy. study, with regression Adjusted RR 1.5 (0,9-2.6) Controls were 25 to 74 years (median towards the mean overall (test for trend, p < 0.05) adjusted for age age 56 years) admitted to hospitals in results from other studies.) Adjusted RR: 1.5 (1 .0-2.3) Northern Italy for acute conditions that Latency (time since first were not hormonal, gynecological, or Updated results may be overestimated by using use). malignant. <10 yrs.: RR 1,5 (1 ,0-2.3) orthopedic controls, which Cases and controls were questioned may be inversely related to adjusted for age by trained interviewers. Results estrogen use (but separate Adjusted RR 1.3 (O 8-2.0) include both pre- and postmenopausal analyses by diagnostic >10 yrs.: RR 1.4 (1 ,0-2.0) women, Subjects were followed from subcategories did not lead adjusted for age 1983 to 1990 to any appreciable Adjusted RR 1.4 (1 .0-2.0) difference in risk). Risk estimates may be affected by higher socioeconomic status of users but risk estimates were not modified by alliance of indicators of socioeconomic status.

*Adjusted for age, geographic area, marital status, education, benign breast disease, family history of breast cancer, nulliparity, age at first birth, age at menarche, type of menopause, age at menopause, body mass index, and oral contraceptive use a Unless otherwise specified, measured relationshipS i relative risk of breast cancer in HRT b 95% confidence interval is given in parentheses C This low relative risk may be due to chance or due to the fact that women who have ovaries removed at a young age’1 ) have lower risk of breast cancer, (2) are more likely to be prescribed estrogen, generally for a short period of time KEY: C.l. = confidence interval; NS = not statistically significant; O R = odds ratio; RR = relative risk SOURCE: Office of Technology Assessment, 1995 Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency Author Description of cases and controls and controls estrogen use a,b estrogen use a,b of estrogen use a,b

Mack (1 975) Cases and controls were white female 99 cases breast Ever use 1.6 (no c.i.) residents of a retirement community in cancer, 396 controls Use at least 5 years before Southern Los Angeles, median age 71. (26% ever users of diagnosis. 1.7 (no c.i.) Cases were diagnosed with breast estrogen among cancer between 1971 and 1975. controls) Controls were selected from a roster of all women in the community, matched with cases for age and date of entry into the community. Information was gathered from questionnaires and medical records.

Casangrande Two groups of subjects were selected. Group 1: For women with natural (1976) Group I was composed of case-control 60 cases; 53 controls menopause. pairs who were white residents of Los Group II Group 1: unadjusted RR 0.47 Angeles County Cases were between 33 cases, 27 controls (no c.i.); 50 and 64 years of age at diagnosis of adjusted RR* 0.75 breast cancer, diagnosed between Group 11: unadjusted RR 1969 and 1972. A control, matched for 2,15 (no c.i.); adjusted RR* age and socioeconomic status, was 3.1 Pooled estimate, 1.2 (p - selected from the outpatient rosters of each index cases’ referring physician. 0.40) Group II cases were white patients whose breast cancers were diagnosed *adjusted for age at menopause between 1972 and 73, and who were between the ages of 50 and 59 at disease diagnosis, lived in SIX middle class white health districts of eastern Los Angeles Cases were matched with healthy control neighbors ages 50 to 59 years Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency a,b Author Description of cases and controls and controls estrogen use a,b estrogen use a,b of estrogen USe

ROSS (1980) Cases were white women diagnosed 138 cases of breast Estrogen use was defined > 7 yrs.: 1.8 Total mg dose (TMD) (= with breast cancer between 1971 and cancer, 281 controls as use beginning more than (test for trend. p = 0.02) daily dose x duration) 1977, between 50 and 74 years of 4 months preceding No exposure (O TMD). age, from two Los Angeles retirement diagnosis. ovaries intact. 1.0 communities. Two postmenopausal All: 1.1 (0.8-1.9) ovaries removed: 1.0 controls were selected for each case Ovaries intact: 1.4 (0.7-2.4) all. 1.0 from the same community, matched for Ovaries removed. 0.8 Low exposure (< 1.500 age, race, move-in date, and marital (0.5-3.5) TMD): status. Estrogen use was ascertained ovaries intact: 0.9 (0.4-1.7) from interviews, medical records, and ovaries removed: 0.9 pharmacy records. (0.2-3.2) all: 0.8 (0.5-1 .5) High exposure (>= 1,500 TMD) (3 yrs. x 1.25 mg/d): ovaries intact: 2.5 (1 .2-5.6) ovaries removed: 0.7 (0.2-2.4) all: 1.9 (1 .0-3.3)

Hoover (1981 ) Cases were all women with breast 345 cases; Ever use: 1.4 (1 .0-2.0) Number of prescriptions Usual daily dose: cancer identified from the tumor 611 controls Natural menopause: 1.3 noted, nonuser: 1.00 registry of Kaiser Foundation Health (69% estrogen users) (0.8-2.1) 0.1.00 < 1 .25mg, 1.4 Plan of Portland, Oregon occurring Oophorectomized women. 1 : 1.1 > 1.25mg, 1.8 from January 1969 to December 1975. 1.5 (0.3-6.6) 2-4: 1.3 (test for trend. Controls were drawn from 5% of a 5-9. 1.8 p = 0.005) random sample of all members of the > 10: 1.8 Kaiser Foundation Health Plan. (test for trend: p = 0.013) Information was gathered from Years between first and last medical records. Average age of prescription. cases and controls was 57. none 1.00 <4, 1,4 >5, 1.7 (test for trend. D = 0.022)

I Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency a,b a,b Author Description of cases and controls and controls estrogen usea,b estrogen use of estrogen USe

Hiatt (1 984) Study subjects were identified from list 119 cases, 119 RR 0.7 (0.3-1 .6) Chart of notations of Three or more years since of operations performed in all Northern controls (90% estrogen use >=5 yrs.: 2.1 first use, 0.8 (0.4-1.9) California Kaiser Foundation Health estrogen users) (1 .2-3,6) Plan hospitals between 1953 and Duration >= 3 yrs.: 1.8 1979. Cases were identified by (0.9-3.6) hospital discharge records. Controls were chosen from women with same age, year of oophorectomy, and date of entry into health plan membership. Information was gathered from medical records.

Brinton (1986) Subjects for the study were from a 1,960 cases; 2,258 1.03 (0.9-1.2) <5 yrs.: 0.89 (0,8-1.0) Premarin 0.3 mg: multicenter breast cancer screening controls 5-9 yrs.: 1.09 (0.9-1.3) <10 yrs. use: 1.04(NS) program. Cases were white women 10-14 yrs.: 1.28 (0.9-1 .6) 10+ yrs. use: 0.76(NS) who underwent natural or surgical 15-19 yrs.: 1.24 (0.9-1 .8) total: 0.99 (0.7-1 .4) menopause at least three months prior 20+ yrs.: 1.47 (0.9-2.3) Premarin 0.625 mg: to the diagnosis of breast cancer; (test for trend: p < 0.01) <10 yrs. use: 0.90(NS) cases were diagnosed between 1973 10+ yrs. use: 1.94 (p < 0.05) and 1980. Controls were chosen from total: 1.05 (0.8-1 .3) women who did not have biopsy Premarin 1.25 mg: during course of screening and were <10 yrs. use: 0.94(NS) matched to the cases for race, age, 10+ yrs. use: 1.13(NS) time of entry, medical center and total. 1.02 (0.9-1.2) length of continuation in the program. Information was gathered through Premarin 2.5 mg: home interviews. <10 yrs. use: 0.77(NS) 10+ yrs. use: 1.00(NS) total: 0.84 (0.5-1.4) Years since initial use: <1 o: 1.03 (0.9-1 .2) 10-14: 1,15 (0.9-1.4) 15-19: 0.95 (0.7-1.2) 20+: 0,97 (0.7-1 .2) (test for trend. p = 0.45) Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency a,b a,b Author Description of cases and controls and controls estrogen use a,b estrogen use of estrogen USe

McDonald Cases were white female residents of 183 cases, 531 Estrogen use was defined 1-5 yrs.: O 83 Never: 1.00 (1986) King County, Washington, ages 50 to controls as at least 1 yr. of estrogen >6 yrs.: 0,68 0.2-1.0 mg: 0.55 74, in whom breast cancer was use (test for trend p = 0.06) > 1.0 mg: 0,81 diagnosed from July 1977 through Overall. 0.74 (0.51 -1 .08) (test for trend. p = 0.22) August 1978, cases were identified Natural menopause O 76 Recency (time since last from a cancer reporting system. (0.46-1 .26) use): Never 1,00 Controls were white female residents of Hysterectomy with current user or <= 5 yrs.: 0.75 King County, 50 to 74 years old, oophorectomy: 1.28 >6 yrs.: 0.76 without breast cancer. All cases and (0.43-3.80) (test for trend. p = 0.14) controls were interviewed. “[S]ome variation In Latency (time since first proportions was present use). Never. 1.00 between different <10 yrs.: 0,73 hysterectomy-oophorectomy > 10 yrs.: 0.74 subgroups. However, each (test for trend: p=O.11) of these differences could easily have been due to chance “

Hunt (1987) Subjects were women, ages 45 to 54 53 breast cancer Adjusted’ RR 12-30 receiving hormonal replacement cases, 106 controls months 1.0 (no c.i.) therapy, recruited from 21 menopause 31-48 months 48 (1 .5-156) clinics around Britain. Recruitment was 49-72 months 5.3 (1 .4-202) both retrospective and prospective. >73 months, 36 (0,9-1 5,0) Subjects were followed from 1978 to 1982. Two controls were selected for *adjusted for uterine and ovarian each case from cohort. status

Wingo (1 987) CASH study, all subjects were 1,369 cases, 1,645 Adjusted RR for users of All women Dose (ever users compared postmenopausal women enrolled from controls more than 3 months versus < 1 year 1,0 (0,7-1 .3) with never users) eight different geographic areas in the nonusers. 1,0 (0,9-1.2) 1-4 yrs.: 1.1 (0.8-1.3) (milligram-months): United States. Cases were women 25 All women. 1.0 (0.9-1.2) ever 5-9 yrs.: 1.1 (0.8- 1.5) <25 1 1 (0,8-1 .5) to 54 years old with cancer diagnosed users versus nonusers 10-14 yrs.: 0.8 (0.5-1.3) 25-499. 1.3 (0.9-1.9) between 1981 and 1982 and Identified 15-19 yrs.: 1.3 (0.6-2.6) 50-749. 1.1 (0.7-1.8) Hysterectomy and bilateral through the SEER cancer registry. >20 yrs.: 1.8 (0.6-5.8) 75-99 .9.1.9 (1.1 -3.3) oophorectomy: ever users Controls were selected from the same (test for trend. p = 0.7) > 100: 0.8 (0.6-1 .2) versus nonusers: 1.3 geographical area by random digit (test for trend p = 0.03) (0.9-1.9) Hysterectomy with bilateral dialing of residential telephone oophorectomy. Recency (time since last numbers. Information was gathered Hysterectomy only ever <1 year: 1,6 (0.9-2,8) use): through interviews. users versus nonusers: 1.1 1-4 yrs.: 1.3 (0.9-2.0) <1 yr.: 1.0 (0.8-1.2) (0.8-1 .5) 5-9 yrs.: 1.1 (0.7-1 .8) 1-4 yrs.: 1.2 (0.9-1 .6) Natural menopause: ever 10-14 yrs.: 1.5 (0.8-2.9) 5-9 yrs.: 1.1 (0.8-1 .6) users versus nonusers: 0.8 >15 yrs.: 1.7 (0,7-4,4) >10 yrs.: 1.0 (0.5-1 .8) (0.6-1 .1) (test for trend: p = 0.9) (test for trend: p = 0.06) Hysterectomy only. Latency (time since first <1 year: 0.9 (0.6-1 .5) use): All women: 1-4 yrs.: 1.1 (0.8-1 .6) < 1 year: 0.9 (0.5-1.5) 5-9 yrs.: 1.6 (1 .0-2.8) 1-4 yrs.: 1.1 (0.8-1.4) 10-14 yrs.: 0.6 (0.3-1 .2) 5-9 yrs.: 1.1 (0.8-1.4) >15 yrs.: 2.0 (O. 7-5.5) 10-14 yrs.: 0.9 (0.7-1 .3) (test for trend: p = 0.7) 15-19 yrs.: 1.1 (0.6-2.1) Natural menopause. >20 yrs.: 1.7 (0,8-3,7) < 1 year: 0.8 (0.4-1 .4) (test for trend: p = 0.8) 1-4 yrs.: 0.9 (0.6-1 .3) >5 yrs.: 0.7 (0.3-1.4) (test for trend. p = 0.6) Relationship of breast Relationship of Relationship of breast cancer to dose, recency, and latency Number of cases breast cancer to cancer to duration of o a,b a,b a,b Author Description of cases and controls and controls estrogen USe estrogen use of estrogen USe Recency (time since last Ewertz (1 988) Cases were pre- and postmenopausal 1,484 cases (56.2% Menopausal.’ Menopausal: women below 70 years of age postmenopausal), 1,16 (0.64-2.1 1) <3 years, 1.08 (0,51 -2.27) use). Menopausal: <3 yrs.: 1,59 (0,80-3,16) diagnosed with breast cancer between 1,334 controls (58.9% Post-menopausal” 3-5 yrs.: 1.10 (0.38-3.21) 1983 and 1984, Identified from the files postmenopausal) 1,28 (0.96-1 .71) 6+ yrs.: 1.57 (0.55-4.44) 3+ yrs.: 0.53 (0.19-1 .45) of a Danish clinical trial of breast Artificial menopause. (test for trend: p = 0.44) Postmenopausal: cancer therapy and from a Danish 1.04 (0.69-1 .57) Postmenopausal. <3 yrs.: 1.48 (1 .01 -2.15) cancer registry. Controls were a <3 yrs.: 0,89 (0,56-1 .41) 3-5 yrs.: 1.13 (0.56-2.30) random sample of women from the ● “menopausal” defined as natural 3-5 yrs.: 0.93 (0.52-1 .68) 6-8 yrs.: 0.99 (0.52-1 .88) general population, stratified for age, menopause within 5 years of 6-8 yrs.: 1.82 (0.98-3.37) 9+ yrs.: 1.18 (0.68-2.02) diagnosis identified from a Danish population 9-11 yrs.: 1.34 (0.70-2.54) Artificial menopause: registry. Data were collected by mailed **’’postmenopausal” defined as natural menopause more than 5 124 yrs.: 2.32 (1 .31 -4.12) <3 yrs.: 1.09 (0.70-1 .71) questionnaire. years before diagnosis (test for trend: p = 0.002) 3-5 yrs.: 0.57 (0.21-1 .51) “Exposure to estrogen or Artificial menopause: 6-8 yrs.: 2.76 (0.73-10.5) progestagen, alone or in <3 yrs.: 1,01 (0.55-1 .85) 9+ yrs.: 0.91 (0.39-2.11) combination-type therapy, 3-5 yrs.: 0.81 (0.39-1 .70) Latency (time since first did not affect the breast 6-8 yrs.: 1.52 (0.65-3.53) uses): cancer risk. Sequential 9-11 yrs.: 1.44 (0.70-2.96) Menopausal: therapy with oestrogen and 12+ yrs.: 0.88 (0.48-1 ,64) <6 yrs.: 1,84 (0,84-4,04) progestagen ., was (test for trend p > 0.5) 6+ yrs.: 0.57 (0.25-1 .30) associated with an Postmenopausal: increased risk of <6 yrs.: 0,95 (0.45-2,02) borderline statistical 6-8 yrs.: 1.10 (0.60-2.04) significance (RR=1.36 9-11 yrs.: 103 (0.62-1 .73) (0.98-1 .87)).” 12+ yrs.: 174 (1.19-2 55) Artificial menopause. <6 yrs.: 1,07 (0,53-2.16) 6-8 yrs.: 1,48 (0.71-3.09) 9-11 yrs.: 127 (0.64-2 49) 12+ yrs.: 0.89 (0.52-1 .50) Relationship of breast Relationship of Relationship of breast cancer to dose, Number of cases breast cancer to cancer to duration of recency, and latency a,b a,b Author Description of cases and controls and controls estrogen usea,b estrogen USe of estrogen use

Rohan (1988) Cases were women from Adelaide, 281 cases; 288 Unadjusted RR 0.88 <24 me,, unadjusted, 1.02 Recency (time since last South Australia, with breast cancer controls (0.57-1 .37) (0,61-1 .72) use): <= 2 yrs.: unadjusted. reported to a cancer registry between Adjusted RR ● 1.03 adjusted: 0.99 (0.56-1 .76) 0.85 (0.32-2.25) 1982 and 1984, and who were 20 to 74 (0.62-1 .69) >24 me,, unadjusted: 0.61 adjusted. 1.25 (0.44-3.58) >2 yrs.: unadjusted, 0.90 years old at the time of diagnosis. Women with bilateral (0.29-1 .31) (0.54-1 .48) Controls were women from Adelaide oophorectomy: 0.30 adjusted: 0.94 (0.40-2.21 ) adjusted: 0.88 (0.51 -1 .54) with no history of breast cancer, (0.09-0.94) adjusted for age A relatively small number of identified from electoral rolls. Latency (time since first Natural menopause: 1.01 women used exogenous Information was obtained through use):<= 15 years since first (0.69-1 ,47) adjusted for age estrogens and only a questionnaires. Reported here is minority reported relatively use: unadjusted: 0.80 information only on those women who long durations of use (0.44-1 .45) were postmenopausal (no menses * adjusted for years of educalion, adjusted: 0.79 (0.41-1 ,55) within 12 months or surgical practice of breast self-examination, >15 years, unadjusted: 1.10 history of breast cancer in first menopause). degree relative, age at last (0.58-2.08) menstrual period, and history of adjusted: 1.27 (0.63-2.54) bilateral oophorectomy > 15 years since first and estrogen therapy >=24 me.: 1.54 (0.43-5.45) Age at first use: <45 y.o.: unadjusted: 0.64 (0.31-1 .33) adjusted. 0.79 (0.35-1 .80) >45 y.o.: unadjusted: 1.12 (0.66-1 .92) adjusted. 1 12 (0.61-2.04) 96 Cost Effectiveness of Screening for Osteoporosis Relationship of breast Relationship of Relationship of breast cancer to dose, Number of oases breast cancer to cancer to duration of recency, and latency a,b a,b Author Description of cases and controls and controls estrogen use a,b estrogen use of estrogen use

Palmer (1991 ) Cases were women under age 70 who 607 breast cancer Estrogens alone: 1.0 Conjugated estrogens. Recency (time since last had breast cancer diagnosed more cases; 1,214 controls (0.7-1.3) <1 yr.: 0.9 (0,5-1 .7) use). Never: 1.0 1-4 yrs.: 0.7 (0.4-1 .3) <1 yr.: 0.4 (0.2-0,9) than six months before interview, Estrogen plus progesterone: 1-2 yrs.: 5.2 (2.0-13) identified at a major cancer treatment 0.6 (0.2-2.0) 5-9 yrs.: 0.8 (0.4-1 .6) hospital in metropolitan Toronto 10-14 yrs.: 0.6 (0.2-1 .8) 3-4 yrs.: 1.0 (0.3-3.1) between 1982 and 1986. Controls >15 yrs.: 1.4 (0,6-3,3) 5-9 yrs.: 0.4 (0.2-1 .1) were identified from tax assessment >10 yrs.: 0.8 (0.4-1.4) rolls of all residents of Ontario. Two Latency (time since first controls were matched to each case use): Never: 1.0 o for age and neighborhood; 41 percent Less than 5 yrs. total use of cases and 42 percent of controls and < 10 yrs. since first use. were pre- or postmenopausal. Cases 0.8 (0.4-1 .5) and controls were interviewed in their 10-19 yrs.: 0.9 (0.5-1 .8) homes. >20 yrs.: 0.5 (0.2-1 .6) Five or more years total use and < 10 yrs. since first use: 0.9 (0.3-2.5) 10-19 yrs.: 0.5 (0.2-1 .0) >20 yrs.: 2.1 (0,9-5.0)

Yang (1992) Cases were all British Columbia 669 cases; O.R. 1.0 (0.8-1 .3) Long-term use (>= 10 years): Current use: women under 75 years of age who 685 controls for ever use of unopposed O.R. 1.6 (1 .1-2.5) O.R. 1.4 (1 .0-2.0) were diagnosed with breast cancer estrogen during 1988 and 1989. Controls were O.R. 1.2 (0.6-2.2) drawn from voter registration lists from for ever use of estrogen and the same province, and were matched progesterone with cases on the basis of age. Analysis included only postmenopausal women. Information was gathered by mailed questionnaire.

Weinstein Cases were female residents of Long 1,436 cases; There was no significant There was no significant There was no significant (1993) Island, NY, aged 20 to 79, who were 1,419 controls association between association of risk with association of HRT with diagnosed with breast cancer from ever-use of HRT and breast duration of use. recency of estrogen use. January 1984 to December 1986. Age- cancer risk. There was a significant and county-matched controls were increased risk of breast selected from driver’s license files. cancer in women with 10 to 19 years since first exposure. a Unless otherwise specified, measured relationship is relative risk of breast cancer in HRT. b 95% confidence interval is given in parentheses. KEY: Cl, = confidence interval; NS = not statistically significant; O R = odds ratio; RR = relativerisk. SOURCE: Office of Technology Assessment, 1995. Relationship of dose, Relationship of Relationship of duration recency, and latency of Number of study estrogen use to of estrogen use to breast estrogen to breast a,b a,b Author Description of study subjects breast cancer risk cancer risk cancer risk

Thomas White women who were initially treated 1,439 women (66 cases Unadjusted RR 1.80 No evidence was seen of an (1982) for biopsy-proven benign breast breast cancer) (504 (1 ,04-3.10) increased relative risk of diseases from 1942 to 1975 in a single estrogen users) adjusted RR 1.84 breast cancer with private surgery practice were followed (1 ,05-3,23) increased duration of through 1976 for development of There was no variance estrogen use. breast cancer. Patients were followed for age or year of first up through letters, phone calls, clinic use. records, and death certificates. Average follow-up was 12,9 years.

Bush (1983) Participants were white women, aged 2,270 white women (593 Breast cancer deaths: 40 to 69 years at baseline, and users, 1,677 nonusers) users: O followed for an average of 5.5 years. nonusers. 12 All women in the cohort were No statistical analysis of participants in the Lipid Research breast cancer deaths Clinics Program Follow-up Study, was provided. conducted in 10 North American Clinics between 1972 and 1976. All subjects were examined at initiation, and were followed with clinic visits and by review of death certificates. Information on decedents was gathered from medical records and family members. Petitti (1987) Walnut Creek Contraceptive Drug 3,437 women who never Risk of breast cancer Study; subjects were women aged 18 used oral contraceptives death in users 0.8 to 54 recruited from December 1968 to or estrogen; 2,656 (O 4-1.8) adjusted for February 1972. All subjects received a women who had used age history and physical exam at initiation estrogens, but not oral and were followed by subsequent contraceptives exam or questionnaire through 1977 Until the end of 1983, deaths were Identified through the California Death Index Users of oral contraceptives were excluded from this analysis......

Relationship of dose, o Relationship of Relationship of duration recency, and latency of Number of study estrogen use to of estrogen use to breast estrogen to breast Author Description of study subjects breast cancer risk cancer riska,b cancer riska,b

Bergkvist Women who had been prescribed 23,224 women age 35 RR 11 (1 .0-1 .3) All HRT users. <6 mos.: 0,7 (0,4-1 .0) (1989) estrogens for conditions related to the and older who had filled Study suggested there menopause were identified through at least one prescription was no protection from 7-36 mos.: 1.1 (0.9-1 .4) records of the pharmacies in the for estrogen; 253 breast the addition of 37-72 mos.: 1.0 (0,8-1 .4) health care region around Uppsala, cancer cases progesterone. 73-108 mos: 1.3 (0.9-1 .9) Sweden. Recruitment began in April > 109 mos,: 1.7 (1,1 -2,7) 1977 and ended in March 1980 and estrogen only. subjects were followed for an average <6 mos,: 0.8 (0,5-1 ,4) of 6 years. Expected numbers of 7-36 mos.: 1.1 (0.8-1 .5) breast cancer cases were estimated 37-72 mos.: 0.9 (0.6-1 .3) according to incidence rates of breast 73-108 mos.: 0.9 (0.5-1 .6) cancer in the region. Median age of >109 mos,. 1,8 (1 ,0-3.1) women in the cohort was 53.7 at time estrogen plus of inclusion into study. Information was progesterone.’ gathered by mailed questionnaire from <6 mos.. 0.5 (0,2-1 .8) a subcohort of 1 in 30 women 7-36 mos.: 0.7 (0.3-1 .3) randomly chosen from the cohort. 37-72 mos.: 0.9 (0.3-2.6) 73-108 mos,: 4.4 (0,9-22.4) >109 mos,: (no estimate)

*Only a small number of women received combination therapy, so confidence Intervals are wide Relationship of dose, Relationship of Relationship of duration recency, and latency of Number of study estrogen use to of estrogen use to breast estrogen to breast Author Description of study subjects breast cancer risk cancer riska,b cancer riska,b

Mills (1989) Subjects were white Seventh-Day 20,341 women in cohort Unadjusted RR 1.67 <1 yr.: 2,28 (1 .38-3.79) <1 yr,: 2.28 (1 .38-3.79) Adventist women, residing in (80% postmenopausal at (1.1 7-2.39) 1-5 yrs.: 1.56 (0.95-2.56) 1-5 yrs.. 1.56 (0.95-2.56) California, who completed a study initiation) (66% adjusted RR’ 1.39 6-10 yrs.: 2.75 (1.64-4.64) 6-10 yrs.. 2.75 questionnaire and who were followed ever users of HRT); 215 (1 ,00-1 .94) 10+ yrs.: 1.53 (0.92-2.54) (1 .64-4.64) for 6 years, from 1976 and 1982. Mean cases of breast cancer There was no strong 10+ yrs.. 1.53 age of cohort in 1976 was 55,4 years. increase in risk with duration (0.92-2.54) *adjusted for age, ages at Information was gathered from annual of HRT. There was no strong questionnaires, tumor registries and menarche, first birth, and menopause, Quetelet’s Index, increase in risk with for cases, from medical records. maternal breast cancer, and a duration of HRT. history of previous benign breast disease natural menopause: 1.74 (1 .10-2,74) hysterectomy, 1.30 (0.78-2.18) .

Relationship of dose, Relationship of Relationship of duration recency, and latency of Number of study estrogen use to of estrogen use to breast estrogen to breast a,b Author Description of study subjects breast cancer risk cancer riska,b cancer risk

Colditz (1 990) Female registered nurses 30 to 55 23,607 postmenopausal Current use: 1.40 Current users: Current users years of age completed a mailed female registered (1 ,16-1 .67) 1-11 mos.: 1,28 (0.8-2.1) 0.3 mg/d:1 .55 (1 .0-2.5) questionnaire. Follow-up nurses; 722 cases of 12-23 mos.: 1,32 (0.8-2.2) 0.625 mg/d: 1.42 questionnaires were mailed every 2 breast cancer 24-35 mos.: 1.44 (0.9-2.2) (1 .0-1 .9) years. Data were gathered between 36-59 mos.. 1.26 (0.9-1 .9) 1,25 mg/d:1.48 (1 .0-2.2) 1976 and 1986. Only those RNs that 60-119 mos.. 1.62 (1 .2-2,1) <1,25 mg/d: 2.27 were postmenopausal are included in 120-179 mos.: 1.28 (0.8-2.0) (1 .0-5,3) these results. >180 mos,: 1,19 (0,6-2.2) Trend with increasing past users: dosage was not 1-11 mos.: 1.00 (0.7-1 .4) significant (test for trend: 12-23 mos.: 1.05 (0.7-1 .5) p = 0.56). 24-35 mos.: 0.65 (0.4-1 .1) current use. 1.40 36-59 mos.. 1.02 (0.7-1 .5) (1 .16-1 .67) 60-119 mos.: 1.05 (0.7-1 .5) past use. 0.99 120-179 mos.: 0.92 (0.5-1.7) (0.82-1.19) >180 mos.: 0,79 (0,3-2.5) time since last use: current 1.36 (1.1 1-1 ,67) 1-11 mos.. 1.62 (0.98-2.67) 12-35 mos.. 1.09 (0.79-1 .50) 36-59 mos.: 0.89 (0.60-1 .31) 60-119 mos.. 0.93 (079-1 .47) > 120 mos. 0,70 (0.45-1 .10) .

Relationship of Relationship of duration recency, and latency of Number of study estrogen use to of estrogen use to breast estrogen to breast a,b a,b Author Description of study subjects breast cancer risk cancer risk cancer risk —. — Breast cancer incidence Henderson Prospective study of postmenopausal 8,881 postmenopausal RR 0,81 (c. I. not After adjusting for age, there current use. RR 1.33 (1991) female residents of Leisure World women (the number of reported) for breast was no evidence of (1,1 2-1 ,57) adjusted Retirement Community in Southern deaths from breast cancer deaths in ever increased risk with California. Residents are cancer was not specified users versus never users increasing duration of use for age predominantly white, moderately in report) (57% ever of estrogen among current users (test past use: RR 0.90 (0.77-1 .04) adjusted affluent, and well-educated. Median estrogen users) for trend: p =0.41) or past age of cohort was 73 at study users (test for trend p = for age initiation. Study was initiated in 1981, 0.46). average follow-up is 7.5 years. Information was gathered through mailed questionnaires and death registries. Colditz (1 992) Subjects were female registered O 480,665 person-years of Ever use 1.08 nurses 30 to 55 years of age 1976. follow-up, 1,050 incident (0.96-1 .22) adjusted for Reported here are results of 12 years cases of breast cancer age of follow-up. Data was obtained by current use of hormones: questionnaires mailed every two years, 1.33 (1 .12-1 .57) cases of breast cancer were adjusted for age confirmed by review of pathology current use of reports and hospital records. unopposed estrogen: 1,42 (1.1 9-1 .70) current use of estrogen and progesterone: 1.54 (0.99-2.39) current use of progesterone alone: 2.52 (0.66-9.63) current use of conjugated estrogens 1,42 (1 .19-1 .20) current use of estrogen/ progestin 1.54 (0.99-2.39) current use of estrogen/ testosterone 2.45 — (0.95-6.35) Relationship of dose, Relationship of Relationship of duration recency, and latency of Number of study estrogen use to of estrogen use to breast estrogen to breast Author Description of study subjects breast cancer risk cancer riska,b cancer riska,b

Schairer Subjects were participants in the 49,017 Current ERT users: 1.3 (1994) Breast Cancer Detection (1,185 breast cancer (1 .1-1 .5) Demonstration Project, a breast cases) (46.2 of past users: 0.9 (0,8-1 .1) cancer screening program conducted person-years in study Current PERT users: 1.2 0 between 1973 and 1980. (The analysis involved ERT, 6% with (0.9-1 .6) reported here included all women who combined PERT) past users: 1.4 (1 .0-2.0) did not have a menses for at least 3 estrogen alone: months prior to an interview. Reported 1.0 (0.9-1 .2) here is followup through 1989). estrogen and progestin. 1.2 Information was collected by (1 ,0-1 .6) telephone interviews mailed in situ tumors only: estrogen questionnaires, and pathology reports. alone: Average age at start of followup was 1.4 (1 ,0-2.0) 57.4 years. Mean duration of followup estrogen and progestin. 2.3 was 6,2 years. (1.3-3.9) no significant association of ERT or PERT with invasive tumors duration of use: There was no significant association of use of ERT with duration of use. However, risk of in situ breast cancer rose with increasing duration of use, with users of 10 years or more having about twice the risk as non users (test for trend, p= 0.02). There was no clear pattern of risk associated with duration of use for PERT users, either for all cancers, in situ cancers or invasive tumors. a Unless otherwise specified, measured relationshipIS relative risk of breast cancer in HRT b 95 % confidenceInterval is given in parentheses. KEY: c.i. = confidence interval; O.E. ratio = observed to expected ratio: OR = odds ratio; NS = not statistically significant; RR = relative risk SOURCE: Office of Technology Assessment, 1995

I Relationship of dose, Relationship of Relationship of duration recency, and latency of Number of study estrogen use to of estrogen use to breast estrogen to breast a,b Author Description of study subjects breast cancer risk cancer riska,b cancer risk Burch (1 974) Subjects were hysterectomized women 1,000 hysterectomized Observed breast cancer on estrogen replacement therapy, women cases. 33, expected. followed for an average of 14.32 years. 23.7 Expected number of deaths from U.S. observed breast cancer Public Health Service cancer morbidity deaths: 6: expected: statistics. 7,85 No statistical analysis of the data was presented. Hoover (1976) The medical records of all white 1,891 women in cohort; RR 1.3 (1 .0-1 .7) <5 yrs.: 0.9 (0,5-1 .5) “The increased risk women seen in one private practice in 49 cases of breast 5-9 yrs.: 1.2 (0.6-2.0) associated with stronger Louisville, KY, from 1939 to 1972, were cancer developed 10-14 yrs.: 1.3 (0.6-2.4) medication and reviewed. Expected rates for the 15+ yrs,: 2.0 (1,1 -3.4) non-daily regimens are general population were obtained from Trend of greater risk with based on small numbers Second and Third National Cancer increased duration is but are statistically Surveys. Average age of women in the statistically significant significant. ” cohort was 49 years. Mean follow-up (p= 0.02). 10+ years follow-up. was 12 years. A finer breakdown of the 0.3 mg: 1.6 (0.9-2.7) follow-up duration after 10 0.625 mg. 1.1 (0.5-2.0) years indicated that the >0,625 mg: 2.7(1 ,2-5.3) excess becomes manifest There was no statistically after about 12 years of significant increase in estrogen use. breast cancer risk with 10-12 yrs.: 1.2 (no c.i.) less than 10 years 13-16 yrs.: 1.9 followup. 17-24 yrs.: 2.0 ......

— Relationship of dose, Relationship of Relationship of duration recency, and latency of Number of study estrogen use to of estrogen use to breast estrogen to breast a,b Author Description of study subjects breast cancer risk cancer risk a,b cancer risk

Hammond Subject had been followed at least 5 301 patients treated with Estrogen users. O.E. (1979) years at Duke University Medical estrogen and 309 ratio 1.06 (O 3-2.7) for untreated patients whites Center, Durham, NC, with diagnoses o associated with a hypo-estrogenic No breast cancers state (e.g., premature ovarian failure or occurred in nonwhite pituitary tumor). Information was estrogen users. gathered retrospectively from medical nonusers of estrogen records, and in some cases from O.E. ratio 0.5 (O.1-1.5) referring physicians, patients, or death for whites certificates. Subjects were divided into O.E. ratio 0.5 (0.0-2.9) two groups. those who never received for nonwhites estrogen and those who received estrogen for longer than 5 years. (Those estrogen users for 5 years or less were excluded.) The observed incidence of breast cancer was compared to age and race-specific incidence rates from the Third National Cancer Survey (southeast United States). Gambrell Subjects were women from Wilford Hall 5,563 women; 53 cases Estrogen plus (1983) USAF Medical Center in San Antonio, of breast cancer progesterone O 3 Texas who received various forms of (0.1-0.8) hormonal therapy. Patients with a estrogen only. 0.7 diagnosis of breast cancer between (0.5-1.1) 1975 through 1981 were identified estrogen vaginal cream: from a tumor registry. Expected values 0.4 (0.2-1 .6) were obtained from the Third National Cancer Survey (1975) and the National progesterone or Cancer Institute Surveillance, androgen users. 0.7 Epidemiology, End Result (SEER) data (0.3-1.5) (1980). Information was gathered from untreated women 1,4 mailed questionnaires, clinic and (1 .1-1 9) hospital records, and registries. This study was criticized for falling to control for confounding functions, including age,

I Relationship of dose, Relationship of Relationship of duration recency, and latency of Number of study estrogen use to of estrogen use to breast estrogen to breast a,b a b Author Description of study subjects breast cancer risk cancer risk cancer risk ’

Vakil (1983) Incidence of breast cancer in a cohort 1,483 menopausal Standard mortality ratio of women, 32 to 62 years of age, women for breast cancer. 0.48 receiving estrogen treatment for (p < 0,01) compared menopausal symptoms among the with Ontario controls, patients of 20 gynecologists in the 0.45 (p< 0.01) metropolitan Toronto area was compared with compared to two control groups: the Saskatchewan controls age-specific breast cancer incidence standard incidence ratio rates of the female populations of of breast cancer. 0.62 Ontario and of Saskatchewan. (p < 0,01) compared Estrogen therapy was begun between with Ontario controls; 1960 and 1970 and subjects were 0.70 (p < 0.01) followed up to 17 years. Information compared with was gathered from gynecologists and Saskatchewan controls cancer and death registries. Hunt (1987) Subjects were women receiving 4,544 women; 503 cases O.E. ratio breast cancer There was no significant Interval since first use of hormonal replacement therapy, of breast cancer incidence 1.59 increase in incidence with HRT: recruited from 21 menopause clinics (1 18-2,10) increasing duration of O-4 years. O.E. ratio 1.40 around Britain. Subjects were followed hysterectomy only O.E. estrogen use (0.85-2.46) from 1978 to 1982. Most women were ratio 3.08 5-9 years. O.E. ratio 1.45 (0.88-2.24) 45 to 54 years of age at time of hysterectomy and recruitment. Subjects were recruited 10+ years O.E. ratio oophorectomy both retrospectively and prospectively. 3.07 (1 .47-5.64) O.E. ratio 166 Expected numbers were obtained from There was evidence of a cancer registry roles. All patients were natural menopause. trend in ratio with interval interviewed at study initiation. Deaths O.E ratio 1.19 since first use (test for were reported from central registries. trend. p = 0.08). 108 Cost Effectiveness of Screening for Osteoporosis Relationship of dose, Relationship of Relationship of duration recency, and latency of Number of study estrogen use to of estrogen use to breast estrogen to breast Author Description of study subjects breast cancer risk cancer riska,b cancer riska,b

Risch (1 994) Subjects were women ages 43 to 49 32,790 women (742 Unopposed estrogen For unopposed estrogen, Unopposed estrogens, years of age in 1976, resident in breast cancer cases) 1,33 (1.1 1-1 ,59) risk increased by 7 percent 1 to 126 tablets/yr. 1,039 Saskatchewan, Canada, who were both opposed and for each 252 tablets (O 78-1 38) identified for the master file of the unopposed estrogen prescribed (approximately I 127-378 tablets/yr government health Insurance plan that 1.10 (O 35-O 42) year of use) (RR 1072 1,161 (0,83-1 63) covers virtually all residents of the No breast cancer cases (1 02-1.13) 379-756 tablets/yr province. These women’s health plan occurred among the 171 For opposed estrogens, 1,041 (0,66-1 63) registration number was used to obtain subjects who used there was no significant >757 tablets/yr 1.498 their prescription records form the opposed estrogens increase in risk for each 252 (1 ,05-2,13) plan’s pharmacy database for the progestins (both alone or tablets prescribed (RR 1,211 period from January 1976 to June (0.72-2.05). 1987, The women’s health plan combined with estrogen) registration number allowed the 0,93 (0.51 -1 68) For unopposed progestins, investigators to Iink the women’s there was no significant pharmacy records to the increase in risk for each 84 Saskatchewan Provincial Cancer tablets prescribed (equal to Registry. Thirty-one percent of the seven tablets per month for cohort used in opposed estrogens 12 months) (RR 1,0003 (mostly conjugated estrogens), 2,07 (O 80-1 25)). percent used opposed estrogens, Estrogen use was defined as use of 3.5 years or more..— a Unless otherwise specified, measured relationshipIS relative risk of breast cancer in HRT b 95% confidence Interval is given in parentheses KEY: c.i. = confidence internal; NS = not statistically significant; O.E. ratio = observed to expected ratio; O.R. = odd ratio; RR = relative risk SOURCE: Office of Technology Assessment, 1995 Appendix G: Evidence on HRT and Endometrial Cancer G

ndometrial cancer, the most common gy - trogen replacement therapy have been on the re- necologic cancer, occurs in about one bound as physicians have been prescribing pro- woman out of 1,000 in the population each gestins in sequence with estrogens to prevent year (15). An average 50-year-old white estrogen from inducing endometrial hyperplasia womanE has a 2.6 percent lifetime risk of endome- (19,47). trial cancer (l). And about eight out of every 100 Obesity and other conditions associated with a women diagnosed with endometrial cancer die of high level of endogenous estrogens are associated this disease (l). Evidence that estrogen replace- with an increased risk of endometrial cancer, so it ment therapy increases the risk of endometrial is not surprising that estrogen replacement thera- cancer is well established and is consistent with a py also increases the risk of endometrial cancer variety of observations. (7). The relationship of endometrial cancer with use The increase in endometrial cancer with estro- of estrogen replacement therapy is consistent with gen replacement therapy is also physiologically trends in the incidence of endometrial cancer. In plausible, and is consistent with observations the United States, there was a dramatic increase in about the relationship of estrogen to the endome- prescriptions for estrogen replacement therapy trium. Estrogen is a growth hormone for the endo- between the mid- 1960s and the early 1970s (47). metrial tissue lining the inside of the uterus. In Estrogen was usually prescribed alone, without a premenopausal women, estrogen levels begin to progestin, and was given for three weeks out of a rise at the beginning of the monthly menstrual four-week cycle. A rise in incidence of endome- cycle, and progesterone levels increase near the trial cancer coincided with this increase in pre- end of the cycle, causing the endometrial tissue to scriptions for estrogen. By 1976, the first case- mature. In the absence of implantation of a fertil- control studies were published that revealed ized egg into the endometrium, estrogen and pro- significant increases in risk of endometrial cancer gesterone levels fall and the endometrial tissue is in estrogen users compared with nonusers (57,73, sloughed off, resulting in menstruation. 92). After these reports, sales of estrogen replace- If estrogen stimulation continues unopposed ment therapy began to drop, as did endometrial by progesterone, the endometrium continues to cancer rates (47). Since 1980, prescriptions fores- grow, producing hyperplasia, or overgrowth of the

I 111 112 Cost Effectiveness of Screening for Osteoporosis

endometrium (19). Hyperplasia has been shown of use, the increase in risk has been estimated to be to advance to carcinoma in situ, and eventually to as little as 2.6 to as great as 63. endometrial cancer (31,52,63). This progression The risk of endometrial cancer has been shown has been observed in patients with diseases char- to be related to dose of estrogen. (See tables G-1 to acterized by excessive unopposed estrogen secre- G-4.) Hence, the minimum effective dose to main- tion, such as Stein-Leventhal Syndrome (74), es- tain bone mineral density and to relieve postme- trogen-producing tumors (32), and certain types nopausal symptoms is commonly prescribed. of infertility (69). Progestins have been shown to (See appendix E.) produce maturation of estrogen-primed endometrium and regression of hyperplastic tissue to nor- RECENCY OF USE OF ESTROGEN mal endometrium (79). It has even led to regres- The risk of endometrial cancer decreases after sion of some well-differentiated carcinomas in cessation of therapy. Some studies have reported some patients (24,67). that risks of endometrial cancer returned to levels Numerous case-control and cohort studies have of nonusers after only six months to two years (40, documented an increase in endometrial cancer 57), while others have found the increase in risk to with use of estrogens. These are presented in persist for up to 15 years after estrogen replace- tables G-1 to G-4 at the end of this appendix. ment therapy is stopped (8,62,70,72). The data Up to 20-fold increases in risk of endometrial comparing the trends of estrogen prescription vol- cancer have been detected in case-control studies ume with endometrial cancer incidence are more of estrogen replacement therapy. (See tables G-1 consistent with a short time interval between and G-2.) Among case-control studies, relative cessation of estrogen replacement therapy and de- risks are generally lower in hospital-based case- cline in endometrial cancer risk (5). control studies that use as controls women with gynecologic problems, probably because uterine STAGE AND GRADE OF bleeding is one of the most common gynecologic ENDOMETRIAL CANCER problems and estrogen commonly causes this Endometrial cancer arising in estrogen users is of symptom (28). Relative risks are generally higher lower stage and grade and much less likely to re- in population-based case-control studies and hospital-based case-control studies that use as con- sult in death than endometrial cancer arising in nonusers of estrogen. A number of case-control trols women without gynecologic problems, in studies have consistently found a lower stage and part because surveillance for endometrial cancer grade of endometrial cancer in estrogen users. is increased among women taking estrogen (28). (See table G-3.) Virtually all endometrial cancers in estrogen users are diagnosed before they have DURATION AND DOSE OF ESTROGEN spread beyond the uterus. In cases where endome- Studies of the relationship of endometrial cancer trial cancer has not spread beyond the uterus, hys- to duration of estrogen replacement therapy indi- terectomy is usually curative. The survival among cate that significant increases in risk of endome- estrogen users diagnosed with endometrial cancer trial cancer can be detected in as little as six is favorable (12). Barrett-Connor reported that months to one year after initiation of estrogen re- women not on estrogen survive less well than placement therapy (4,58,72,75,92). Epidemiolog- women with endometrial cancer taking estrogen ic studies have shown that the risk of endometrial (6). Furthermore, there was little evidence that cancer increases with increased duration of use. mortality from endometrial cancer increased dur- (See tables G-1 to G-4.) For 10 or fewer years of ing the period of rising incidence of the disease use, the risk ranges from no significant increase to from estrogen use in the population (47). a 36-fold increase in risk. For more than 10 years Appendix G Evidence on HRT and Endometrial Cancer 113

However, some users of estrogen replacement age less aggressive than those that arise in women therapy do develop cancers that have spread be- who have not taken estrogen replacement (82). yond the uterus (Stage III and Stage IV) (23, 70, The author reviewed five case-control studies ex- 72), and some estrogen users die of this complica- amining the association between prior postmeno- tion (18,57,62). pausal estrogen use and endometrial cancer prog- There are several factors that may account for nosis. They found that, although estrogen use is this relatively favorable prognosis. First, the low- associated with an increased risk of endometrial er stage and grade of endometrial cancer in estro- cancer, that association tended to weaken when gen users may be due to detection bias. Estrogen only invasive and high-grade tumors are consid- users are closely monitored with endometrial ered. The authors explained that one possible rea- biopsies annually and at times of irregular bleed- son for this finding was that tumors that arise in ing. Vaginal bleeding is an early symptom of en- the presence of exogenous estrogens are on aver- dometrial cancer, and women taking estrogen re- age less aggressive than those that arise in their ab- placement therapy may bleed earlier and be sence. Another possible explanation, they noted, biopsied earlier than women receiving less regular was detection bias, that endometrial cancer in es- medical care (7,56). The favorable stage and grade trogen users may be detected earlier than in nonus- may also be due in part to case ascertainment — ers of estrogen. This may be because estrogen us- the detection of occult cancers in the endometrium ers may tend to seek care more promptly than of users who bleed because they are taking estro- nonusers, their access to medical care may be gen (38). The apparently favorable survival expe- greater, or the physicians of estrogens may detect rience of user cases is also likely due in part to pa- endometrial cancers early because they are partic- tients with estrogen-induced benign hyperplasia ularly wary of the development of these cancers in mislabeled as cases (56). Bias may also be their patients on estrogen. introduced by a greater likelihood of estrogen A third possible explanation, according to the treatment in women who have menopausal prob- authors, is overdiagnosis of endometrial cancer in lems associated with unsuspected cancer or a estrogen users. Because the histological criteria greater likelihood of cancer (56). for separating the more advanced cases of endo- The lower stage and grade of estrogen-induced metrial hyperplasia are ambiguous, some cases of tumors may be because these tumors are more be- estrogen-related advanced hyperplasia are being nign than tumors that arise in the absence of estro- incorrectly labeled as early endometrial cancer, gen. Estrogen-induced endometrial cancers may giving rise to a false association of estrogen use be better differentiated and slower growing than with low-grade, low-stage cancers. endometrial cancers that arise in the absence of in- Deligdisch and Holinka have provided addi- ducement by exogenous estrogen. tional evidence that patients known to be at in- Estrogen-induced irregular bleeding, hyperpla- creased risk of endometrial cancer due to exposure sia, and localized cancers of the endometrium re- to estrogen are likely to develop better differen- sult in an increased prevalence of hysterectomy tiated and less aggressive forms of cancer (16). among estrogen users (20). Thus, even though the The researchers examined the cellular characteris- endometrial bleeding, hyperplasia, and cancers tics of the tumors of 95 patients with Stage I endo- associated with estrogen use do not substantially metrial cancer. Noting that endometrial hyperpla- increase mortality, they do contribute to medical sia is excessive growth of endometrial tissue costs associated with estrogen replacement thera- caused by estrogen stimulation, they found that py (15). endometrial cancers with hyperplasia were better Weiss and colleagues were among the first to differentiated and less invasive than endometrial suggest that endometrial cancers that arise in cancers without hyperplasia. women taking estrogen replacement are on aver- 114 Cost Effectiveness of Screening for Osteoporosis

ESTROGEN USE AND SURVIVAL FROM history and the response of the family physician to ENDOMETRIAL CANCER a letter requesting more detailed information. The Epidemiologic studies have consistently found investigators compared the stage and grade of en- that, among postmenopausal women diagnosed dometrial cancer in ever users of CEE to never us- with endometrial cancer, estrogen users have ers of postmenopausal estrogens. Only 8 percent markedly better survival than never users of estro- of CEE users had Stage H or III cancers at diagno- gen. sis, compared with 16 percent of nonusers. And 43 Robboy et al. concluded that survival differ- percent of tumors in CEE users were well differen- ences between estrogen users and nonusers was tiated, compared with 29 percent of nonusers. due to differences in grade of tumor at diagnosis The 5-year survival rate, after adjustment for (68). The authors identified 274 women treated age, was 94.2 percent in ever users of CEE and for endometrial cancer at the Massachusetts Gen- was 81.3 percent in nonusers, a difference that was eral Hospital between 1940 and 1971. Pathologi- highly significant (p = 0.001). When differences cal specimens for each woman were examined to in stage were taken into account, survival was not confirm the diagnosis of endometrial cancer. Hos- significantly different between the two groups. pital and clinic records were available for 190 of Collins et al. studied endometrial cancer stage, these women, and were reviewed for a history of grade, and survival in 860 women referred to a postmenopausal estrogen use. They found that 85 London, Ontario cancer clinic between 1967 and percent of the 274 patients with endometrial can- 1976 (13). Information on prior estrogen use was cer were stage I at diagnosis, and 7 percent were obtained through a questionnaire. About one third stage II, with no significant difference in stage at of the patients had a history of estrogen use, de- diagnosis between estrogen users and nonusers. fined as use of estrogen for 6 months or more be- However, the tumors that developed in estrogen fore diagnosis. users were significantly more differentiated than At all stages of endometrial cancer, estrogen those that developed in nonusers (p less than users had a significantly greater 5-year survival 0.05). Five- and 10-year survival was also signifi- than nonusers. The researchers found that, after cantly better in users than in nonusers, but surviv- adjusting for a number of risk factors for mortal- al in users and nonusers was not significantly dif- ity, endometrial cancer patients with no history of ferent once adjusted for differences in grade of prior estrogen use had a 5.4 times greater risk of tumor. death from cancer than endometrial cancer pa- The authors did not rule out that their findings tients with a history of prior estrogen use. could be explained by earlier detection in estrogen The authors posited that endometrial cancer pa- users because of better endometrial cancer sur- tients with a history of estrogen use had higher veillance. This explanation was supported by the survival rates because cancers associated with fact that the average age of estrogen users at diag- prior estrogen use are less aggressive tumors. The nosis was four years less than nonusers (56 versus authors, however, did not rule out the possibility 60 years of age, p less than 0.02). that selection or surveillance bias may have con- Elwood et al. concluded that survival differ- founded their findings. ences between estrogen users and nonusers is al- In a study of 379 white women ages 50 to 74 most entirely due to differences in the stage and from King County, Washington, with newly diag- grade of endometrial cancers at diagnosis (1 8). El- nosed endometrial cancer, Chu and colleagues wood et al. studied 494 women seen at a Vancouv- concluded that although the use of postmenopau- er clinic between 1968 and 1972 for treatment of sal estrogen leads to an increased risk of endome- newly diagnosed endometrial cancer. All patients trial cancer, there is no increased risk of endome- were followed until death or to 1975. Information trial cancer death in postmenopausal estrogen on estrogen use was based on both the patient’s users (12). The authors obtained information on Appendix G Evidence on HRT and Endometrial Cancer 115

cases of endometrial cancer diagnosed between (16). There is substantial evidence that women January 1975 and April 1976 from the Cancer Sur- who take progestins with estrogen are at no in- veillance Center, a population-based registry creased risk of developing endometrial cancer serving western Washington State. Additional in- compared with postmenopausal women who do formation was obtained from interviews of the pa- not take estrogen. tient’s physician. Information on estrogen use, Until recently, only large-scale cross-sectional medical and reproductive history, and risk factors studies were available on the effect of combined for endometrial cancer was obtained by interview- estrogen and progestin therapy on endometrial ing the patient; for the 12 percent of study partici- cancer risk, and these studies showed that the pants who could not be interviewed, this informa- combination reduced the incidence of endometrial tion was obtained by reviewing the medical cancer to below that of an untreated population records of primary care physicians. Fully 98 per- (25). A number of prospective studies have shown cent of estrogen users (defined as use of estrogen that the incidence of endometrial cancer is in- for one or more years after menopause) had tu- creased with unopposed estrogen replacement mors stage O or I at diagnosis, compared with 88 therapy, but not with combined estrogen and pro- percent of nonusers. Only 2 percent of estrogen gestin therapy (25, 64). users had stage II or III cancers at diagnosis, Persson et al. examined the incidence of endo- compared to 12 percent in nonusers, a difference metrial cancer in hormone replacement therapy that was statistically significant. (HRT) in the Uppsala health care region, which Estrogen users with endometrial cancer had a serves one-sixth of the population of Sweden (64). small but significantly better four-year survival Using the region’s prescription database, he was rate than women of the same age in the general able to identify 23,244 women over age 35 who population, as calculated from Washington state filled one or more prescriptions for HRT between life tables for white women (relative survival ratio April 1977 and March 1980. Women from the co- 1.05 (1.04-1.06)). Estrogen users with endome- hort who developed endometrial cancer were trial cancer also had a significantly better four- identified from the region’s cancer registry. In- year survival than nonusers with endometrial can- formation on lifetime exposures to estrogen and cer, the latter group having a relative survival ratio progestin, compliance, and sociodemographic of 0.89 (0.80-0.99) compared with women of the data were obtained on 735 randomly selected same age. members of the cohort. Comparison was made to The authors stated that the possibility that these women in the general background population. results were due to self selection or detection bias A relatively high proportion of the HRT users could not be ruled out. They also mentioned that in this cohort were receiving progestin and estro- other factors that may confound the interpretation gen replacement therapy (PERT), allowing com- of their results include differences in follow-up parison to be made with estrogen replacement between estrogen users and nonusers, differences therapy (ERT) (64). The investigators found that, in cancer therapy between estrogen users and non- while users of ERT has a significantly increased users. They also noted that the interpretation of re- risk of endometrial cancer (relative risk 1.8 (95 sults may be limited by the relatively short (four- percent confidence interval 1.1 to 3.2) after expo- year) follow-up period. sure to any estrogen for 6 years), users of PERT were at no increased risk (relative risk 0.9 (95 per- ESTROGEN/PROGESTIN cent confidence interval 0.4 to 2.0)). REPLACEMENT THERAPY However, for some of the less androgenic pro- Because even a relatively benign cancer is an un- gestins (such as medroxyprogesterone acetate acceptable complication, most physicians add a (Provera), the most commonly used progestin in progestin to suppress endometrial hyperplasia the United States), and in the regimens and lower 116 Cost Effectiveness of Screening for Osteoporosis

doses commonly used today, there are insufficient medroxyprogesterone acetate, 10 mg per day, studies with endometrial cancer as an endpoint; most for 10 days each month. most studies of efficacy look at an intermediate Current users of estrogen alone had a relative endpoint, such as reversal of endometrial hyper- risk of endometrial cancer of 6.5 (95 percent con- plasia. Medroxyprogesterone acetate, in a dose of fidence interval 3.1 to 13.3), whereas current users 10 mg for 12 days, is the least androgenic regimen of estrogen and progesterone had a relative risk of that has been best documented to prevent hyper- 1.9 (95 percent confidence interval 0.4 to 8.7). plasia (63,85). Past users of estrogen alone or estrogen and pro- Although courses of medroxyprogesterone gestin had no increased risk of endometrial cancer. acetate of fewer than 12 days have been shown to The study found that users of estrogens for three to reduce the incidence of estrogen-induced endo- four years had a relative risk of 1.9 (95 percent metrial hyperplasia (86), the minimum duration to confidence interval 0.4 to 8.7), and that users of reduce the incidence to zero is 12 days per month five years or more had a relative risk of 22 (95 per- (64,78,87). cent confidence interval 1.5 to 24.1). Users of es- Some clinicians prefer a lower dose, 2.5 or 5 trogen and progestin for more than three years had mg, of medroxyprogesterone acetate. These a relative risk of 1.3 (95 percent confidence inter- smaller doses are often given concurrently with val 0.5 to 3.4). The researchers concluded that estrogen throughout the month (7,84). A continu- there does not appear to be any substantial in- ous low-dose regimen avoids the withdrawal crease in risk associated with combined use with bleeding of cyclic progestin, which may lead to increasing duration of therapy. The researchers poor compliance. In addition, these lower doses cautioned, however, that there were relatively few are less likely to induce premenstrual-type symp- women who used combined therapy for more than toms associated with progestins Long-term data five years. on the ability of continuous low-dose progestin to A number of clinical trials have demonstrated protect the endometrium overtime is limited (85). that the sequential or continuous addition of a pro- Additional data is also needed on the effects of gestin reduces the incidence of or eliminates endo- these treatments on lipids, lipoproteins, and other metrial hyperplasia, thought to be a precursor to metabolic parameters (11). endometrial cancer. Woodruff and Pikar examined One recent case-control study provides evi- the incidence of hyperplasia in a one-year, ran- dence that menopausal women taking estrogen re- domized clinical trial of conjugated estrogens placement therapy can significantly reduce their (Premarin) and medroxyprogesterone acetate risk of endometrial cancer if they also take me- (Provera) in 1,724 postmenopausal women (91). droxyprogesterone (45). The study examined The subjects were divided into five groups: two women between the ages of 50 and 64 who were groups received continuous estrogen/progestin treated from 1979 to 1989 at Group Health, a regimens, two groups received sequential estro- Seattle, Washington health maintenance organiza- gen/progestin regimens, and one group received tion. Researchers identified 172 cases of endome- unopposed estrogen regimen. 1 They found that, trial cancer and compared use of hormones in while endometrial hyperplasia developed in 20 these women with that of 1,720 women who did percent of women on Premarin alone, hyperplasia not have cancer. Users of combined therapy used

1 The regimens examined were as follows: (1)0.625 mg Premarin plus 2.5 mg Provera daily; (2) 0.625 mg Premarin plus 5 mg Provera daily; (3) 0.625 mg Premarin daily plus 5 mg Provera for 14 days per month; (4) 0.625 mg Premarin plus 10 mg Provera for 14 days per month; and (5) 0.625 mg Premarin daily unopposed by progestin (Woodruff, 1994). Appendix G Evidence on HRT and Endometrial Cancer 117

developed in one percent or less of women in the In a recent metaanalysis, Grady et al. estimated four Premarin/Provera groups (91).2 a risk of endometrial cancer in ERT users that was Other trials of sequential or continuous regi- intermediate between OTA’s base case and worst mens using other estrogens and progestins have case estimates (28). They concluded that the risk demonstrated less hyperplasia in PERT users than of endometrial cancer increased with prolonged in users of estrogen alone (21,90). duration of ERT use, from a relative risk of 1.4(95 Although the incidence of endometrial cancer percent confidence interval 1.0 to 1.8) for less than is reduced in estrogen-progestin users, the risk of one year of use, 2.8 (95 percent confidence inter- endometrial cancer is not eliminated completely. val 2.3 to 3.5) for two to five years of use, 5.9 (95 One group of investigators reported on 25 postme- percent confidence interval 4.7 to 7.5) for six to 10 nopausal women who developed endometrial can- years of use, and 9.5 (95 percent confidence inter- cers while taking PERT for one or more years (59). val 7.4 to 12.3) for more than 10 years of use (28).3 Twenty-three (98 percent) of the women had can- For PERT users, OTA assumed that there cers limited to the uterus, but two had disease ex- would be no increase in endometrial cancer risk tending beyond the uterus. All of the women were over that of the baseline population. This is con- alive and disease free after a median follow-up of sistent with the estimates of endometrial cancer 26 months. The endometrial cancers that did occur risk from the metaanalysis by Grady and col- among PERT users were usually associated with leagues, who found that case-control studies esti- regimens that had inadequate doses of progestins. mated a slightly increased risk of endometrial cancer in PERT users (relative risk 1.8), whereas the IMPLICATIONS FOR OTA’S COST few cohort studies of PERT users have estimated a EFFECTIVENESS MODEL slightly decreased risk of endometrial cancer (rel- The evidence is strong that endometrial cancer ative risk 0.4) (28). risks begin to rise soon after the initiation of ERT. In modeling the impact of HRT on endometrial Following the weight of the evidence presented in cancer, OTA made a number of simplifying as- tables G-1 and G-2, OTA assumed that the relative sumptions. In the case of ERT, OTA assumed that risk of endometrial cancer during the first nine the relative risk of endometrial cancer would sub- years of ERT would be 2.5 and in subsequent side to that of the baseline population in the year years would rise to 7.0. The sensitivity of results following cessation of HRT. This assumption is to changes in these assumptions was also tested. consistent with observations that the risk of endo- For the case most favorable to ERT, OTA assumed metrial cancer drops rapidly after discontinuing that relative risk of endometrial cancer is 1 for the estrogen use. There are, however, a number of first nine years of therapy and rises to 2.0 during studies that have been able to detect relatively the 10th and subsequent years of ERT. This best small elevations in risk of endometrial cancer that case is based on the assumption that the apparent persist several years after cessation of therapy. increased risk of endometrial cancer in ERT users Grady et al. estimated a relative risk of endome- is largely due to surveillance bias. In the worst trial cancer of 2.3 (95 percent confidence intervals case, the relative risk would be 7.5 in the first nine 1.8 to 3.1 ) five or more years after discontinuation years of ERT and 15.0 thereafter. This estimate is of long-term ERT use (28). based on epidemiological studies that detected the OTA assumed that endometrial cancers in HRT highest risks of endometrial cancer in HRT users. users would be early stage and grade, and would

2 Although the incidence in endometrial hyperplasia did not differ significantly among the Premarin/Provera groups, none of the women who received the sequential or continuous regimens with the highest dosages of progestins developed endometrial hyperplasia (91). 3 An earlier metaanalysis by Gradyand colleagues estimated a relative risk of 8 in long-term estrogen users (28). 118 Cost Effectiveness of Screening for Osteoporosis

be cured by hysterectomy. OTA also assumed, for REFERENCES simplicity, that there would be no endometrial 1. American Cancer Society, Clinical Oncology cancer deaths in HRT users. The metaanalysis by for Medical Students and Physicians: A Mul- Grady et al. estimated that ERT is related to a large tidisciplinary Approach, 6th cd., P. Rubin increase in risk of early stage cancers (relative risk (cd.) (American Cancer Society: Atlanta, GA, 4.2 (95 percent confidence interval 3.1 to 5.7) for 1983). Stage O and 1 cancers) (28). They found a trend to- 2. American College of Obstetricians and Gy- ward later stage endometrial cancers in ERT users necologists, “Endometrial Cancer and Estro- that did not reach statistical significance (relative gen Therapy,” American Family Physician risk 1.4 (95 percent confidence interval 0.8 to 2.4) 49(4):972, 1994. for Stage 2 to 4 cancers). 3. American College of Obstetricians and Gy- Observational studies have been unable to de- necologists, Estrogen Replacement Therapy tect a significantly increased risk of endometrial and Endometrial Cancer, ACOG Committee cancer death in ERT users (20,49,62,65). This Opinion 126 (Washington, DC: American may be due in part to the small number of endome- College of Obstetricians and Gynecologists, trial cancer deaths in these studies. In a metaanaly - August 1993). sis, Grady and colleagues were able to use pooled 4. Antunes, C. M., Strolley, P. P., Rosenshein, data from these studies to detect a trend toward in- N. B., et al., “Endometrial Cancer and Estro- creased endometrial cancer deaths in ERT users gen Use: Report of a Large Case-Control that failed to reach statistical significance (relative Study,” New England Journal of Medicine risk 2.7 (95 percent confidence interval 0.9 to 300(1):9-13, 1979. 8.0)) (28). Because endometrial cancer is less 5. Austin, D., and Roe, K., “The Decreasing In- common than breast cancer, hip fracture, or heart cidence of Endometrial Cancer: Public Health disease, and because there are relatively small Implications,” American Journal of Public numbers of invasive endometrial cancers and Health 72:65-68, 1982. deaths due to HRT-induced endometrial cancer, 6. Barrett-Connor, E., “Estrogen Replacement OTA’s simplifying assumptions about endome- Therapy After Age 65—Who, When, Where trial cancer stage and endometrial cancer deaths in and Why?” unpublished paper presented at ERT users should not have a substantial impact on AGS/AFAR Annual Meeting, Fifty Years of the results of OTA’s analysis. Caring for Older Adults, conference spon- OTA also assumed that women diagnosed with sored by the American Geriatrics Society, endometrial cancer would remain off hormonal Washington, DC, Nov. 14-18, 1992. replacement therapy. It was previously thought 7. Barrett-Connor, E., “Risks and Benefits of that HRT could induce the growth of any residual Replacement Estrogen,” Annual Review of endometrial cancer cells, and thereby increase the Medicine 43:239-251, 1992. risk of recurrence. There is a growing consensus, 8. Brinton, L., and Hoover, R., “Estrogen Re- however, that a history of endometrial cancer is placement Therapy and Endometrial Cancer not a contraindication to continuing HRT, at least Risk: Unresolved Issues,” Obstetrics & Gy- with respect to women who have had hysterecto- necology 81(2):265-271, 1993. mies for tumors that have not spread beyond the 9. Buring, J., Bain, C., and Ehrmann, R., “Con- uterus (3,14,60). It is doubtful, however, that most jugated Estrogen Use and Risk of Endome- women would be willing to resume HRT after trial Cancer,” American Journal of Epide- having had endometrial cancer. miology 124(3):434-441, 1986. 10. Bush, T., Cowan, L., Barrett-Connor, E., et al., “Estrogen Use and All-Cause Mortality,” Appendix G Evidence on HRT and Endometrial Cancer 119

Journal of the American Medical Association Unopposed Estrogen Replacement Therapy,” 249(7):903-906, 1983. Maturitas 10:271-282, 1988. 11. Christianson, C., and Riis, B., “Five Years 21. Ettinger, B., Selby, J., Citron, J.T., et al., “Cy- with Continuous Combined Oestrogen/Pro- clic Hormone Replacement Therapy Using gestin Therapy. Effects on Calcium Metabo- Quarterly Progestin,” Obstetrics & Gynecol- lism, Lipids and Lipo-Proteins and Bleeding ogy 83(5 pt 1):693-700, 1994. Pattern,” British Journal of Obstetrics and 22. Ettinger, B., Selby, J., Citron, J. T., et al., “Gy- Gynecology 97:1087-1092, 1990. necologic Complications of Cyclic Estrogen 12. Chu, J., Schweid, A. I., and Weiss, N., “Sur- Progestin Therapy,” Maturitas 17(3): vival Among Women with Endometrial Can- 197-124, 1993. cer: A Comparison of Estrogen Users and 23. Ewertz, M., Schou, G., and Boice, J., “The Nonusers,” American Journal of Obstetrics Joint Effect of Risk Factors on Endometrial and Gynecology 143:569-573, 1982. Cancer,” European Journal of Cancer and 13. Collins, J., Dormer, A., Allen, L., et al., “Oes- Clinical Oncology 24(2):189-194, 1988. trogen Use and Survival in Endometrial Can- 24. Ferenczy, A., “How Progestogens Effect En- cer,” Lancet 961-963, 1980. dometrial Hyperplasia and Neoplasia,” Con- 14. Creasman, W.T., “Recommendations Re- temporary Obstetrics and Gynecology 11: garding Estrogen Replacement Therapy After 137-143, 1978. Treatment of Endometrial Cancer,” Oncology 25. Gambrell, R., “Prevention of Endometrial 6(7):23-26, 1992. Cancer with Progestins,” Maturitas 8: 15. Cust, M., Gangar, K., Hillard, T., et al., “A 159-168, 1986. Risk-Benefit Assessment of Estrogen Thera- 26. Gambrell, R., Massey, F., Castaneda, T., et py in Postmenopausal Women,” Drug Safety al., “Reduced Incidence of Endometrial Can- 5(5):345-358, 1990. cer Among Postmenopausal Women Treated 16. Deligdisch, L., and Holinka, C., “Endome- with Progestogens,” Journal of the American trial Carcinoma: Two Diseases?” Cancer Geriatrics Society 27(9):389-394, 1979. Detection and Prevention 10:237-246, 1987. 27. Gillet, G. Y., Andre, G., Faguer, B., et al., “In- 17. Eisenberg, J. M., “Cost Effectiveness of Pre- duction of Amenorrhea During Hormone Re- ventive Services for the Elderly,” unpublished placement Therapy: Optimal Micronized Pro- paper presented at AGS/AFAR Annual Meet- gesterone Dose. A Multicenter Study,” ing, Fifty Years of Caring for Older Adults, Maturitas 19(2):103-115, 1994. conference sponsored by the American Geri- 28. Grady, D., Gebretsadik, T., Kerlikowske, K., atrics Society, Washington, DC, Nov. 14-18, et al., “Hormone Replacement Therapy and 1992. Endometrial Cancer Risk: A Meta-Analysis,” 18. Elwood, J. M., and Boyes, D., “Clinical and Obstetrics & Gynecology 85(2): 1-10, 1995. Pathological Features and Survival of Endo- 29. Grady, D., Rubin, S. M., Petitti, D. B., et al., metrial Cancer Patients in Relation to Prior “Hormone Therapy to Prevent Disease and Use of Estrogens,” Gynecologic Oncology Prolong Life in Postmenopausal Women,” 10:173-187, 1980. Annals of Internal Medicine 117: 1016-1037, 19. Ernster, V.L., Bush, T. L., Huggins, G.R., et 1992. al., “Benefits and Risks of Menopausal Estro- 30. Gray, L. A., Sr., Christopherson, W. M., and gen and/or Progestin Hormone Use,” Preven- Hoover, R. N., “Estrogens and Endometrial tive Medicine 17(2):201-223, 1988. Cancer,” Obstetrics & Gynecology 49(4): 20. Ettinger, B., Golditch, I., and Friedman, G., 385-389, 1977. “Gynecologic Consequences of Long-Term 120 Cost Effectiveness of Screening for Osteoporosis

31. Gusberg, S., and Kaplan, A., “Precursors of 41. Hunt, K., Vessey, M., and McPherson, K., Corpus Cancer. IV. Adenomatous Hyperpla- “Long-Term Surveillance of Mortality and sia as Stage O Carcinoma of the Endome- Cancer Incidence in Women Receiving Hor- trium,” American Journal of Obstetrics and mone Replacement Therapy,” British Journal Gynecology 87:662-678, 1963. of Obstetrics and Gynecology 94:620-635, 32. Gusberg, S., and Kardon, P., “Proliferative 1987. Endometrial Response to Theca-Granulosa 42. Hunt, K., Vessey, M., and McPherson, K., et Cell Tumors,” American Journal of Obstet- al., “Mortality in a Cohort of Long-Term Us- rics and Gynecology 11 1:633-643, 1971. ers of Hormone Replacement Therapy: An 33. Hammond, C., Jelovsek, F., Lee, L., et al., Updated Analysis,” British Journal of Obstet- “Effects of Long-term Estrogen Replacement rics and Gynecology 97: 1080-1086, 1990. Therapy. I. Metabolic Effects,” American 43. Jelovsek, F.R., Hammond, C. B., Woodard, Journal of Obstetrics and Gynecology B.H., et al., “Risk of Exogenous Estrogen 133(5):525-536, 1979. Therapy and Endometrial Cancer,” American 34. Henderson, B., Casagrande, J., Pike, M., et Journal of Obstetrics and Gynecology 137; al., “The Epidemiology of Endometrial Can- 85-91, 1980. cer in Young Women,” British Journal of 44. Jick, H., Watkins, R. N., Hunter, J. T., et al., Cancer 47:749-756, 1983. “Replacement Estrogens and Endometrial 35. Henderson, B., Paganini-Hill, A., and Ross, Cancer,” New England Journal of Medicine R., “Decreased Mortality in Users of Estrogen 300(5):218-222, 1979. Replacement Therapy,” Archives of Internal 45. Jick, S. S., Walker, A. M., and Jick, H., “Estro- Medicine 151:75-78, 1991. gens, Progesterone, and Endometrial Can- 36. Hoogerland, D.L., Buchler, D. A., Crowley, cer,” Epidemiology 4(1):20-24, 1993. J.J., et al., “Estrogen Use—Risk of Endome- 46. Kelsey, J. L., LiVolsi, V. A., Holford, T. R., et trial Carcinoma,” Gynecologic Oncology al., “A Case-Control Study of Cancer of the 6(5):451-458, 1978. Endometrium,” American Journal of Epide- 37. Horowitz, R., and Feinstein, A., “Alternative miology 116(2):333-342, 1982. Analytic Methods for Case-Control Studies 47. Kennedy, D., Baum, C., and Forbes, M., of Estrogens and Endometrial Cancer,” “Noncontraceptive Estrogens and Progestins: New England Journal of Medicine 299: Use Patterns Over Time,” Obstetrics & Gy- 1089-1094, 1978. necology 65:441-446, 1985. 38. Horowitz, R., and Feinstein, A., “Estrogens 48. La Vecchia, C., Franceschi, S., Decarli, A., et and Endometrial Cancer: Responses to Argu- al., “Risk Factors for Endometrial Cancer at ments and Current Status of an Epidemiologic Different Ages, ’’Journal of the National Can- Controversy,” American Journal of Medicine cer Institute 73(3):667-671, 1984. 81:503-507, 1986. 49. Lafferty, F. W., and Helmuth, D. O., “Post- 39. Hulka, B., Fowler, W., Kaufman, D., et al., Menopausal Estrogen Replacement: The Pre- “Estrogen and Endometrial Cancer: Cases vention of Osteoporosis and Systemic Ef- and Two Control Groups from North Caroli- fects,” Maturitas 7:147-159, 1985. na,” American Journal of Obstetrics and Gy- 50. Lawrence, C., Tessaro, I., Durgerian, S., et necology 137:92-101, 1980a. al., “Advanced-Stage Endometrial Cancer: 40. Hulka, B. S., Kaufman, D. G., and Fowler, Contributions of Estrogen Use, Smoking, and W. C., Jr., et al., “Predominance of Early En- Other Risk Factors,” Gynecologic Oncology dometrial Cancers After Long-Term Estrogen 32:41-45, 1989. Use,” Journal of the American Medical 51. Levi, F. La Vecchia, C., Gulie, C., et al., "Oes- Association 244(21):2419-2422, 1980b. trogen Replacement Treatment and the Risk Appendix G Evidence on HRT and Endometrial Cancer 121

of Endometrial Cancer: An Assessment of the Hormone Replacement Therapy,” Gyneco- Role of Covariates,” European Journal of logic Oncology 55:126-132, 1994. Cancer 29: 1445-1449, 1993. 60. Mortel, R., “The Creasman Article Re- 52. Lucas, W. E., “Causal Relationships Between viewed,” Oncology 6(7):26 and 28, 1992. Endocrine-Metabolic Variables in Patients 61. Obrink, A., Bunne, G., Cohen, J., et al., “Es- with Endometrial Carcinoma,” Obstetrics trogen Regimen of Women with Endometrial and Gynecology Survey 29:507-528, 1974. Carcinoma,” Acta Obstetrician et Gynecolog- 53. Lunenfeld, B., Diedrich, K., and Creatsas, G., ical Scandinavia 60: 191-197, 1981. “Follow-Up and Hormone Replacement in 62. Paganini-Hill, A., Ross, R., and Henderson, Gynecological Cancer Patients,” European B., “Endometrial Cancer and Patterns of Use Journal of Obstetrics & Gynecology and Re- of Estrogen Replacement Therapy: A Cohort productive Biology 41:27-29, 1991. Study,” British Journal of Cancer 59: 54. Lyles, K. W., “Diagnosing and Managing 445-447, 1989. Common Metabolic Bone Disease: Paget’s 63. Paterson, M., Wade-Evans, T., Sturdee, D., et Disease of Bone, Vitamin D Deficiency, and al., “Endometrial Disease After Treatment Related Disorders,” unpublished paper pres- with Oestrogens and Progestogens in the Cli- ented at AGS/AFAR Annual Meeting, Fifty macteric,” British Medical Journal Years of Caring for Older Adults, conference 280:822-824, 1980. sponsored by the American Geriatrics Soci- 64. Persson, I., Adami, H., Bergkvist, L., et al., ety, Washington, DC, Nov. 14-18, 1992. “Risk of Endometrial Cancer After Treatment 55. Lynn, J., “Current Issues in Medical Decision with Oestrogens Alone or in Conjunction Making: Futility, Family, Feeding, and Fore- with Progestogens: Results of a Prospective going,” unpublished paper presented at AGS/ Study,” British Medical Journal 298: AFAR Annual Meeting, Fifty Years of Caring 147-151, 1989. for Older Adults, conference sponsored by the 65. Petitti, D., Perlman, J., and Sidney, S., “Non- American Geriatrics Society, Washington, contraceptive Estrogens and Mortality: Long- DC, Nov. 14-18, 1992. Term Follow-Up of Women in the Walnut 56. Mack, T., and Ross, R., “A Current Percep- Creek Study,” Obstetrics & Gynecology 70(3, tion of HRT Risks and Benefits,” Osteoporo- part 1):289-292, 1987. sis: Physiological Basis, Assessment, and 66. Pettersson, B., Adami, H., Persson, I., et al., Treatment, H.F. DeLuca and R. Mazess (eds.) “Climacteric Symptoms and Estrogen Re- (New York, NY: Elsevier Science Publishing placement Therapy in Women with Endome- Co., Inc., 1990). trial Carcinoma,” Acta Obstetrician et Gyneco- 57. Mack, T., Pike, M., and Henderson, B., “Es- logical Scandinavia 65:81-87, 1986. trogens and Endometrial Cancer in a Retire- 67. Reifenstein, E., “The Treatment of Advanced ment Community,” New England Journal of Endometrial Cancer with Hydroxyprogester- Medicine 294:1262-1267, 1976. one Caproate,” Gynecologic Oncology 2: 58. McDonald, T. W., Annegers, J. F., O’Fallen, 377-414, 1974. W. M., et al., “Exogenous Estrogens and En- 68. Robboy, S., and Bradley, R., “Changing dometrial Carcinoma: Case-Control and Inci- Trends and Prognostic Features in Endome- dence Study,” American Journal of Obstet- trial Cancer Associated with Exogenous Es- rics and Gynecology 127(6):572-580, 1977. trogen Therapy,” Obstetrics & Gynecology 59. McGonigle, K.F., Karlan, B. Y., Barbuto, 54(3):269-277, 1979. D. A., et al., “Development of Endometrial 69. Ron, E., Lunnfeld, B., Menczer, J., et al., Cancer in Women on Estrogen and Progestin “Cancer Incidence in a Cohort of Infertile 122 Cost Effectiveness of Screening for Osteoporosis

Women,” American Journal of Epidemiology Paoletti, and J.L. Ambrus (eds.) (Lancaster, 125:780-790, 1987. England: MTP Press, 1980). 70. Rubin, G., Peterson, H., Lee, N., et al., “Es- 79. Thorn, M., White, P., Williams, R., et al., trogen Replacement Therapy and the Risk of “Prevention and Treatment of Endometrial Endometrial Cancer: Remaining Controver- Disease in Climacteric Women Receiving Es- sies,” American Journal of Obstetrics and trogen Therapy,” Lancet 2:455-457, 1979. Gynecology 162:148-154, 1990. 80. Vakil, D., Morgan, R., Halliday, M., “Exoge- 71. Salmi, T., “The Endometrial Carcinoma Risk nous Estrogens and Development of Breast Factors, with Special Reference to the Use and Endometrial Cancer,” Cancer Detection of Oestrogens,” Acta Endocrinologica 233 and Prevention 6:415-424, 1983. (Suppl.):37-43,1980. 81. Voigt, L., Weiss, N., Chu, J., et al., “Progesto- 72. Shapiro, S., Kelly, J., Rosenberg, L., et al., gen Supplementation of Exogenous Oestro- “Risk of Localized and Widespread Endome- gens and Risk of Endometrial Cancer,” Lan- trial Cancer in Relation to Recent and Discon- cet 338:274-277, 1991. tinued Use of Conjugated Estrogens,” New 82. Weiss, N., “Noncontraceptive Estrogens and England Journal of Medicine 313:969-972, Abnormalities of Endometrial Proliferation,” 1985. Annals of Internal Medicine 88:410-412, 73. Smith, D., Prentice, R., Thompson, D., et al., 1978. “Association of Exogenous Estrogens and 83. Weiss, N., Szekely, D., English, D., et al., Endometrial Cancer,” New England Journal “Endometrial Cancer in Relation to Patterns of Medicine 293: 1164-1167, 1975. of Menopausal Estrogen Use,” Journal of the 74. Somers, S., Carcinoma of the Endometrium, American Medical Association 242(3): in the Uterus (Baltimore, MD: William and 261-264, 1979. Wilkins, 1973). 84. Whitcroft, S., and Stevenson, J., “Hormone 75. Spengler, R. F., Clarke, E. A., Woolever, C. A., Replacement Therapy: Risks and Benefits,” et al., “Exogenous Estrogens and Endometrial Clinical Endocrinology 36: 15-20, 1992. Cancer: A Case-Control Study and Assess- 85. Whitehead, M., “A Risk-Benefit Assessment ment of Potential Biases,” American Journal of Estrogen Therapy in Postmenopausal of Epidemiology 114(4):497-506, 1981. Women,” Drug Safety 5(5):354-358, 1990. 76. Stampfer, M. J., Colditz, G.A., Willett, W. C., 86. Whitehead, M., King, R., McQueen, J., et al., et al., “A Prospective Study of Exogenous “Endometrial Histology and Biochemistry in Hormones and Risk of Endometrial Cancer” Climacteric Women During Oestrogen and (abstract), American Journal of Epidemiolo- Oestrogen/Progestin Therapy, ’’Journal of the gy 124:520, 1986. Royal Society of Medicine 72:322-327, 1979. 77. Stavraky, K. M., Collins, J. A., Dormer, A., et 87. Whitehead, M., Townsend, P., Pryse-Davies, al., “A Comparison of Estrogen Use by J., et al., “Effects of Various Types and Dos- Women with Endometrial Cancer, Gyneco- ages of Progestogens on the Postmenopausal logic Disorders, and Other Illnesses,” Ameri- Endometrium,” Journal of Reproductive can Journal of Obstetrics and Gynecology Medicine 27:539-548, 1982. 141(5):547-555, 1981. 88. Wigle, D.T., Grace, M., and Smith, E. S., “Es- 78. Studd, J., Thorn, M., Paterson, M., et al., trogen Use and Cancer of the Uterine Corpus “The Prevention and Treatment of Endome- in Alberta,” Canadian Medical Association trial Pathology in Postmenopausal Women Journal 118(6): 1276-1278, 1978. Receiving Exogenous Estrogens,” The Meno- 89. Williams, D. B., Voigt, B.J., Fu, Y. S., et al., pause and Postmenopause, N. Pasetto, R. “Assessment of Less than Monthly Progestin Appendix G Evidence on HRT and Endometrial Cancer 123

Therapy in Postmenopausal Women Given sal Women Taking Conjugated Estrogens Estrogen Replacement,” Obstetrics & Gy- (Premarin) with Medroxyprogesterone Ace- necology 84(5):787-793, 1994. tate or Conjugated Estrogens Alone,” Ameri- 90. Williams, S. R., Frenchek, B., Speroff, T., et can Journal of Obstetrics and Gynecology al., “A Study of Combined Continuous Ethi- 170(5 pt 1):1213-1223, 1994. nyl Estradiol and Norethindrone Acetate for 92. Ziel, H., and Finkle, W., “Increased Risk of Postmenopausal Hormone Replacement,” Endometrial Carcinoma Among Users of American Journal of Obstetrics and Gynecol- Conjugated Estrogens,” New England Jour- ogy 162(2):438-446, 1990. nal of Medicine 293: 1167-1170, 1975. 91. Woodruff, J.D., and Pickar, J. H., “Incidence of Endometrial Hyperplasia in Postmenopau- 124 Cost Effectiveness of Screening for Osteoporosis Appendix G Evidence on HRT and Endometrial Cancer 125

G-

00 126 I Cost Effectiveness of Screening for Osteoporosis

v o

13333333333 Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of Author Description of cases and controls and controls to estrogen usea,b estrogen usea,b estrogen a,b estrogen use a,b Wigle (1 978) Cases were women aged 55 to 74 202 cases (47.2% Any use, 2,2 1-4 years 1.8 Current use. 2.7 ( years with histologically confirmed estrogen users), (p< 0.01) (p< o 05) p < 0,01) endometrial cancer who first 1,243 controls Estrogen users was > 5 years 5,2 Past use. 2.0 attended an Alberta, Canada cancer (26.3% estrogen defined as users of (p< 0,05) (p< 0,01) clinic during the period 1971 to users) hormonal 1973. Controls were women aged 55 replacement to 74 years who attended the cancer therapy or oral clinic for any primary cancer other contraceptives. than breast, cervix, uterus, ovary, or other female genital organs. Information on HRT use and risk factors was gathered by questionnaire. Jick (1979) Cases were women 50 to 64 years of 67 cases (89.6% Ever use: Duration of use: Dose: age who were members of Group estrogen users); 74 relative risk 11.2 (4,2 O-4 years: 3.0 0.3 mg CEE: 4.3 Health Cooperative of Puget Sound, controls (43.2% -21 .1) (0.5-14.9) (1 .2-15 .6) Seattle, Washington, who were estrogen users) 5-8 years: 36.0 0.625 mg CEE: 7.1 diagnosed with endometrial cancer (5.6-300.9) (2,8-1 7.6) from January 1972 to June 1977. 9-12 years 63.0 1.25 mg CEE: (8.4 Controls were members of the same (10,4-502.9) (2,0-36.5) age that were hospitalized for other 13 years. 21.0 conditions at the same age as cases. (4.6-107.9) Information was obtained from Relative risk telephone interviews and clinic estimates were records, calculated by Grady (1995) from published crude data. — Relationship of Relationship of Relationship of o Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of b a,b a,b a,b Author Description of cases and controls and controls to estrogen us& estrogen USe estrogen estrogen use Antunes Cases were all patients with 451 cases (20% Unadjusted relative None: 1.0 <1 mg: 3,5 (1 ,6-7,6) (1979) endometrial cancer admitted to six of estrogen users); risk <1 yr.: 2.2 (0,9-5,5) 1-2 mg: 7,1 (2,8-18) the 24 hospitals in the Greater 446 controls from 6.0 (3.7-9.7) 1-5 yrs.: 2.9 (1 ,3-6.7) >2 mg.: 3.7 (0,8-16) Baltimore area from 1973 to February other services; 442 compared with >5 yrs.: 15 (4,9-45) 1977. Cases were ascertained from gynecology hospital controls hospital tumor registries, admissions controls Unadjusted relative records, and pathology records. risk Controls were female patients who 2.1 (1.5-NA) were matched with cases for compared with hospital, race, age, and date of controls from admission, One set of controls were gynecology service taken from hospital services other Adjusted relative risk than gynecology, obstetrics, and psychiatry services. A second set of 5.5 (2.3-12.9) controls was taken from the compared with “ hospital controls gynecology service, Information was gathered through personal Adjusted relative risk interviews, medical records, and 2.4 (1 ,5-3.7) pathologenic specimens. compared with gynecology controls Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of Author Description of cases and controls and controls to estrogen use a,b estrogen use a,b estrogena,b estrogen use a,b Hulka (1980a) Cases were all women who had 256 cases (32.8% White women. <3.5 yrs.: <0,625 mg: 1.6 (NS) In comparison of received their initial therapy for estrogen users); 1.8 (0,9-2.5) 0.8 (0.4-1.7 compared with cases to endometrial cancer at North Carolina 224 gynecology compared with compared with gynecology community Memorial Hospital (NCMH) from controls (22.9% gynecologic gynecology controls; 2.3 (NS) controls, risk drops 1970 through 1976; patients with users); 321 controls; 1.4 controls; compared with to that of non-users carcinoma “in situ” were excluded. community controls (0.9-2.1) compared 0.7 (0.4-1 .3) community controls of estrogen after a Cases were 60 years old, on (27.1% users) with community compared with >0,625 mg: 1.8 (NS) 20-month average, at admission. Gynecology controls community controls compared with estrogen-free controls were women, average age Black women: >3.5 yrs.: gynecology interval; in 60 years, with intact uteri selected 0.7 (0.3-2.1) 4.1 (1 ,8-9.6) controls, 1.4 (NS) comparison to the from the pool of all gynecologic compared with compared with compared with gynecology control admissions and consultations on gynecologic gynecology community controls group, excess risk surgical or medical services of the controls; 1.5 controls; disappeared 28 NCMH from 1970 through 1976 (0.4-5.1) compared 3.6 (1 .9-6.8) months after matched for age, race, and year of with community compared with cessation of admission, and with intact uterus; controls community controls estrogen. excluded were women admitted to the gynecologic oncology service and women admitted primarily for curettage or endometrial biopsy. Community controls were a sample of women, average age 55 years, with intact uteri residing in a major referral area of NCMH, and matched for age and racial group. Sources of information included interviews and review of medical records. Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of a,b Author Description of cases and controls and controls to estrogen use a,b estrogen use a,b estrogen a,b estrogen use Hulka (1980b) Cases were women, mean age 61 256 cases (32.8% Duration of estrogen Estrogen Estrogen-free interval years, with endometrial cancer estrogen users), use strength= 6 me,, receiving their initial therapy at North 224 gynecology < 3,5 years: mg. Stage 1A: 2.5 Carolina Memorial Hospital (NCMH) controls (22.9% Stage 1A: 1.2 (NS) Stage 1A: 5.8 (NS) between 1970 and 1976. users), 321 Stage 1B: 0.9 (NS) (p < 0,05) Stages IB-IV: 1.3 Gynecologic controls were selected community controls Stage 11: 0.7 (NS) Stage IB-IV: 2.3 (NS) from patients admitted to the (27.1 % users) Stage III-IV: 0.6 (NS) (NS) Grade 1. 2.2 (NS) gynecology service and from Grade 1. 1.0 (NS) Grade 1: 4.0 Grades 2-3: 1.3 (NS) patients receiving gynecologic Grade 2: 0.7 (NS) (p< 0.05) invasion: consultations while inpatients on Grade 3: 0.6 (NS) Grades 2-3.2.5 (NS) endometrium. 2.1 surgical or medical services of the invasion: invasion: (NS) NCMH during 1970 through 1976. myometrium and endometrium: 5.2 myometrium: 2.0 (NS) Admissions to the gynecologic beyond. 0.5 (NS) (p< 0,05) oncology service, women admitted Estrogen-free interval myometrium and for dilation and curettage, and Duration of estrogen < 6 months: beyond: 2.1 (NS) women with a previous hysterectomy use Stage 1A. 8.8 (p were excluded. Controls were >3.5 yrs.: Estrogen strength > < 0.05) Stages matched for age, year of admission, Stage 1A: 7.6 0.625 mg: IB-IV: 2.1 (NS) and race with cases. Community (p< 0,05) Stage 1A: 8.5 Grade 1. 6.2 controls were from a sample of Stage IB: 1.6 (NS) (p< 0,05) (p < 0,05) women over 30 years old (mean age Stage Il: 3.3 Stages IB-IV: 1.5 Grades 2-3. 2.3 (NS) 56 years) residing in the major (p< 0,05) (NS) referral areas of NCMH, stratified by Stage III-IV 1.5 Grade 1. 5.4 age and within racial group. All had (NS) (p< 0,05) intact uteri. Information was gathered Grade 1 55 Grades 2-3. 2.0 (NS) from interviews and medical records. (p< 0,05) Grade 2: 19 (NS) Grade 3 2.9 (p< 0.05) — Appendix G Evidence on HRT and Endometrial Cancer 1131

I .

Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of Author Description of cases and controls and controls to estrogen use a,b estrogen use a,b estrogen a,b estrogen use a’b Salmi (1980) Cases were all patients with 318 cases (33% Matched pairs endometrial cancer diagnosed and hormone users); analysis treated in the Department of 282 matched Any use of hormones. Obstetrics and Gynecology at the controls; 585 total 0.6 (0.4-0.9) University Central Hospital of Turku, controls (43.6% Use of hormones for Finland, from 1970 to 1976. Controls users) gynecological were women between the ages of 35 conditions: 0.6 and 60 identified from Turku’s (0.4-0.9) continuing mass screening program for cervical and breast cancer. Use of estrogen: Women over 60 were identified from 0.4 (0.2-0.7) the National Population Registry. Estrogen use was There were 585 controls, 282 of defined as use of 6 which were matched for age, height, months or more. weight, and social class. Information Estradiol only or on HRT use was gathered by combined with interviews. androgen. 0.3 (0.2-0.7) Estriol only. 0.4 (0.1-1 .0) Conjugated estrogens: 5.0 (p < 0.05) Other estrogens: 0.6 (0.2-1.4) Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of a,b a,b a,b a,b Author Description of cases and controls and controls to estrogen USe estrogen use estrogen estrogen use Stavraky Cases were all new patients between 206 cases (58% unadjusted relative All durations: Gynecologic controls. Use <5 years (1981) 40 and 80 years of age with a estrogen users), risk 2.4 (1 ,6-3.7) unadjusted relative <0,625 mg 1.9 duration diagnosis of endometrial carcinoma 191 gynecologic compared with risk (0.9-3.6) Current users (use admitted to Victoria Hospital, controls (38% gynecologic 2.9 (1 .6-5.1) >0,625 mg 3.1 within past year). London, Ontario between September users); 199 controls adjusted relative risk (1 ,5-6.3) adjusted relative risk 1976 and October 1978 for nongynecologic unadjusted relative 1.3 (0,5-3.7) Nongynecologic 1.3 (0,6-3,1) preoperative radiation. Two controls controls (28% risk 4,3 (2.7-6,7) Risk by duration of controls. compared with for each patient were selected from users) compared with use among patients <0.625 mg 2.9 gynecologic the hospital’s daily patient register, nongynecologic and gynecologic (1 .4-57) controls; one control was a woman with a controls controls who >0.625 mg 6.4 7,2 (2.6-20.3) gynecologic disorder, matched for compared with unadjusted relative presented with (2.5-14,5) age within.5 years; another control bleeding: nongynecologic was a woman with a nongynecologic risk for controls postmenopausal unadjusted relative disease within the same age range; risk Past users (cessation hysterectomized women were not women only 2,3 (1 ,5-3,7) < 2 years: 2.3 of use included in control group. (0.5-7,9) > 1 year ago): compared with - . Information was gathered by 2-4 years: 1,1 adjusted relative risk questionnaire. gynecologic controls (0.5-2.5) 0.5 (0.2-1 .3) 5-9 years: 4.1 compared with unadjusted relative (1 .4-10 .5) gynecologic risk for 1 O+ years, 11,0 controls, postmenopausal (2.1-39.0) 1,0 ( 0,4-2.5) women only compared with 4.8 (2.9-7.7) adjusted relative risk. O-4 yrs.: 0.7 (O 2-2,5) nongynecologlc compared with controls nongynecologic 5-10+ years: 2.3 controls (1 .8-8,4) adjusted relative riskc Risk of endometrial 1.5 (0.9-2.7) cancer among compared with patients and two gynecologic control groups by controls duration of estrogen use 14,4 (5,0-41 .8) compared with nongynecologic controls Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of Author Description of cases and controls and controls to estrogen use a,b estrogen use a,b estrogen a,b estrogen use a,b adjusted relative riskc < 2 years, Use >= 5 years 4.8 (2.7-8.4) adjusted relative risk duration: compared with 0.7 (0.3-1.9) Current users. non-gynecologic compared with adjusted relative risk controls gynecologic 4,3 (1 .9-9,7) o adjusted relative riskc controls, compared with for 1.6 (0,6-4.3) gynecologic postmenopausal compared with controls, women only nongynecologic 11.3 (4.9-25.5) 1.5 (0.8-2.8) controls compared with compared with 2-4 years. nongynecologic gynecologic adjusted relative risk controls controls 1.0 (0.4-2.2) Past use. adjusted relative riskc compared with adjusted relative risk for gynecology 0.7 (0.2-2.7) postmenopausal controls; compared with women only 4.0 (1 .6-10 .1) gynecologic compared with controls, 4.2 (2,2-8,0) nongynecologic 2.3 (0.6-8.5) compared with controls compared with nongynecologic nongynecologic controls 5-9 years: adjusted relative risk controls Estrogen use was 1.7 (0.7-4 1) defined as use six compared with months or more. gynecologic controls; 5.3 (2.2-1 2.4) compared with nongynecologic controls 1 O+ years. adjusted relative risk 6.4 (2.1-19.3) compared with gynecologic controls, Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of Author Description of cases and controls and controls to estrogen use a,b estrogen use a,b estrogen a,b estrogen use a,b Kelsey (1 982) Cases were women ages 45-74 167 cases (47% Use >5 years: Use <1 yr.: odds years old who were admitted from estrogen users), odds ratio 1.6 ratio 1.1 (no 1977 to 1979 to seven Connecticut 903 controls (38% (1 .3-2,0) confidence hospitals with newly diagnosed estrogen users) intervals provided) endometrial cancer. Controls were 1-2.5 yrs.: odds ratio other women in the same age group 1,0 admitted to surgical services (except 2.6-5.0 yrs.: odds gynecology) of those hospitals at the ratio 2.9 same time as the cases. Information 5.1-7.5 yrs.: odds on HRT use was obtained by ratio 4.3 questionnaire. 7.6-10,0 yrs.: odds ratio 8.2 > 10 yrs,: odds ratio 2.7 (test for trend: p < 0,001) Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of Author Description of cases and controls and controls to estrogen us&b estrogen use a,b estrogen a,b estrogen use a,b La Vecchia Subjects were women admitted to 283 cases (25% Relative risk There was a (1984) university and general hospitals in estrogen users); 2.3 (1 .6-3.2) adjusted significant trend of the Greater Milan area between 1979 566 controls (17% for body mass increasing risk with and 1983. Cases were diagnosed users) index and age increasing duration with endometrial cancer within the of use (test for year prior to interview. Cases were trend: p = 0.001). between 33 and 74 years old Age <55 years: (median age 60); 30 cases were <2 years use: 1,8 below 50 years of age. Controls were (0.9-3.6) women less than 75 years admitted >2 years: 5.1 for acute conditions unrelated to risk (1 .5-17.1) factors for endometrial cancer. (test for trend: Women with gynecological, p = 0.002) hormonal, or neoplastic diseases or Age 55-64 years: who had undergone hysterectomy <2 yrs,: 1.5 (0.8-2-6) were excluded from controls. >2 yrs,: 1,8 (0,7-4.5) Information was gathered by (test for trend: personal interview. p = 0.12) Age >= 65 years: <2 yrs.: 1,6 (0.7-3.5) >2 yrs.: 1,4 (0.4-5.4) (test for trend: p = 0.29) Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of oases endometrial cancer to duration of to dose of to recency of a,b a,b a,b a,b Author Description of cases and controls and controls to estrogen USe estrogen USe estrogen estrogen use Shapiro (1985) Cases were women with endometrial 425 cases (31 % Relative risk <1 year: 0.9 (0,4-1 ,8) < 1 year since last carcinoma admitted to hospitals in estrogen users); 3.5 (2,6-4.7) 1-4 years: 2.9 use. Boston, MA; Philadelphia, PA; 792 controls adjusted for age, (1 .8-4.7) < 1 yr. duration. — Baltimore, MD; Tucson, AZ; New body-mass index, 5-9 years: 5.6 1-4 yrs.: 2.1 (0.9-4.7) York, NY; Kansas City, MO; San and geographic (3,4-9.3) 5-9 yrs.: 6.3 (3.0-13) Francisco, CA; and London, Ontario; area. > 10 years: 10 >10 yrs.: 12 (5,9-24) ages 50-69 years, with no history of (5.9-18) Estrogen use was 1-4 years since last other cancers. Controls were other defined as use of use: female patients on medical, surgical, conjugated < 1 yr. duration: 0,6 and orthopedic wards, ages 50-69 estrogen, (0.2-2,0) years, with no history of cancer, beginning at least 1-4 years: 3.1 admitted for conditions judged not to two years prior to (1 .3-7,4) be related to estrogen use. Patients the date of 5-9 years: 5.2 were interviewed between interview. (2.1-13) September 1976 and December > 10 years: 12 1982. (4.8-32) 5-9 years since last use: < 1 yr. duration: 1,0 (0.3-3.5) 1-4 years: 4.0 (1 .4-12) 5-9 years: 6.3 (2,0-20) > 10 years, 3,7 (0.8-18) Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of o a,b a,b a,b a,b Author Description of oases and controls and controls to estrogen use estrogen USe estrogen estrogen use > 10 years since last use: < 1 yr. duration: 1,2 (0.4-3.6) 1-4 years: 3.5 (1 ,4-8.3) 5-9 years: 4.1 (1 .1-15) > 10 years: —

Buring (1986) Cases were white women, aged 188 cases (39% Ever use: 2.4 < 1 yr.: 1,4 (no 0.3 mg, 0.625 mg: < 1 yr.: 2,4 (no 40-80 years, who were admitted to estrogen users); (1 ,7-3,6) confidence interval 2.7 (1 ,6-4.9) confidence interval the Boston Hospital for Women’s 428 controIs (17%. current use. 2.8 provided) 1.25 mg, 2.5 mg: provided) Parkway Division with first diagnosis estrogen users) (1 ,8-4.2) I-4 yrs.: 2.0 3.8 (2.2-6.6) 1 + yrs.: 4.6 of endometrial cancer made (current use defined 5-9 yrs.: 6.4 1-2 yrs.: 4.2 between January 1970 and June as use within the 10+ yrs.: 7.6 3-4 yrs.: 5.9 1975. Controls consisted of all white year before index 5+ yrs.: 4.5 women, aged 40-80 years, admitted admission) An excess risk of to the same hospital during the same endometrial cancer period for nonmalignant conditions was noted to requiring surgery. Information was continue among gathered from hospital and clinic estrogen users who records, had discontinued 5 or more years ago, although there were small numbers of former users......

Weiss (1979) Cases were all female residents of 322 cases (81 % ever Age-adjusted relative Age-adjusted relative Time since last use: King County, Washington, aged 50 to users); risk: risk. >8 years: 3.0 74 years with newly diagnosed 289 controls (34%. 1-2 years 1.2 <0.5 mg per day: (0.9-10.6) endometrial cancer between January ever users) (0.4-3.7) 2.5 (1 .1-5,3) 3-7 years: 3.8 1975 and April 1976. Cases were 3-4 years: 5.4 (1 .5-9.5) 0.6-1.2 mg per day: identified from the Cancer (2.5-1 1,5) 1-2 years: 5.3 8.8 (5.0-12.7) Surveillance System, a 5-7 years: 4.7 (2.6-10.8) population-based tumor registry (2.6-8.4) > 1.25 mg per day: current use: 8.7 serving western Washington. 8-10 years. 11,7 7.6 (5.0-1 1.6) (6.4-1 1.8) Controls were white women aged 51 (6.2-21 .8) to 74 years from King County 11-14 years. 24.2 Identified from household surveys. (1 1,8-49,4) Information on HRT use and risk 15-19 years 102 factors was gathered through (5.3-20,0) interviews, > 20 years, 83 (2.8-24.5)

II Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of Author Description of cases and controls and controls to estrogen usea,b estrogen use a,b estrogen a,b estrogen use a,b Obrink (1981) Swedish study comparing use of 622 cases (19.27. 6-36 months: estrogens among 622 cases of estrogen users); 7.5% cases, 8.O% endometrial cancer treated at 1,866 controls controls (NS) Radiumhemmet (Stockholm) 37-72 months: between 1974 and 1977 with 1 0.3% cases, 2.2% estrogen use of the average female controls (p < 0.001) population, represented by a randomly selected sample of 1,866 More than 6 years of treatment was age-matched controls. Progestin treatment was rare among cases and uncommon. controls.

Spengler Cases were newly diagnosed with 88 cases (45% Odds ratio 2.9 1-6 months: 1.4 Conjugated equine (1981) endometrial cancer between April 1, estrogen users), (1 .7-5,1) (0.5-4.4) estrogens: 1977 and December 31, 1977, and 177 controls (22% Odds ratio matched 7-24 months: 2.6 <1 mg: 2.0 (0,9-4.6) were residents of metropolitan estrogen users) 3.2 (p= 0.0001) (1 .0-6,5) >1 mg: 4.0 (1 .9-8.4) Toronto between 40 and 74 years of 25-60 months: 2.2 total: 3.0 (1 .7-5.3) Relative risk age. Cases were identified from the (0.7-6.5) records of the pathology (adjusted for age, >60 months: 8,6 obesity, age at departments of 21 Toronto hospitals. (3.2-23.0) menopause, Two age-matched controls were nulliparity, and selected from the same educational level) 3.7 neighborhood and type of dwelling (1 .8-7.6) as their respective case. Neighborhood controls were Estrogen use was obtained by door-to-door canvassing defined as use 1 or which started at the fourth dwelling more months during to the right of the case’s residence or after menopause. and proceeded sequentially around the block or through the apartment building. No control had history of hysterectomy or cancer. Information was gathered by questionnaire and by review of hospital and clinic records. 144 I Cost Effectiveness of Screening for Osteoporosis Relationship of Relationship of endometrial cancer Number of cases endometrial cancer to recency of Author Description of cases and controls and controls to estrogen use a,b estrogen use a,b Lawrence Cases were women ages 40 to 69 84 cases (27% < 1 year: The risk of advanced No significant No significant (1989) years from hospitals in upstate New estrogen users); odds ratio 0.84 (no endometrial cancer association was association was York who had been diagnosed as 168 controls (24% confidence interval) increased found between dose found with recency having advanced-stage (stages 2-4) estrogen users) 1-5 years: significantly (p < and risk of interval and risk of endometrial cancer in 1979-1981. odds ratio 1.47 0.05) with duration of endometrial cancer. endometrial cancer. Controls were selected from the files use of estrogen pills. of licensed drivers maintained by the > 5 years: No significant New York State Department of Motor odds ratio 2.21 association was Vehicles. Two controls were selected found for any other for each case, matched by county of variables or for residence and age. Information on interaction between HRT use was gathered through longer estrogen use structured interviews. and dosage greater than 0.625 mg, continuous mode of administration, or recency interval (the time interval from the last use of estrogen to diagnosis). Despite a statistically significant correlation between duration of estrogen use and advanced-stage endometrial cancer, estrogen use actually contributed little to the risk of advanced-stage disease. Odds ratio=1.01 (1.00-1.03). Relationship of Relationship of Relationship of 0 Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of a,b a,b a b a,b Author Description of cases and controls and controls to estrogen use estrogen USe estrogen ’ estrogen use Rubin (1990) Results from Cancer and Steroid 196 cases (24% 1.9 (1 .3-2.8) ever <2 yrs.: 1,3 (0.7-2,4) <0.625 mg per day, Time since last use < Hormone (CASH) Study, a estrogen users); user vs. never user 2-5 years. 2.1 1.2 (0,5-2.7) 2 years multicenter study conducted in 8 986 controls (14’%. Estrogen use was >6 years, 3.5 > 1.25 mg per day, all use. 1.9 (1 .2-3.2); areas of the United States (Atlanta, estrogen users) defined as 3 months 3.8 (1 .7-8.5) duration <2 yrs.. 1.4 GA; Detroit, Ml; San Francisco, CA; consecutive use of (07-3.0); SeattIe, WA; Connecticut, Iowa, New estrogen duration >= 2 yrs., 2.4 Mexico, and Utah). Cases were replacement therapy. (1 ,3-4.4) postmenopausal women 40 to 54 Time since last use years of age who resided in one of 2-5 years: the eight areas and who had an all use. 1.5 (0.8-3.1), endometrial cancer diagnosed duration <2 yrs.: 1.1 between December 1, 1980 and (o 4-3.3); December 31, 1982. Controls were duration >= 2 yrs.: 2.0 women with an intact uterus, (0.8-4.9) matched for age and geographic area to cases. Information on HRT Time since last use >= use was obtained through interview. 6 years. all use. 2.7 (1.1 -6.4); duration <2 yrs.: 1.4 (0.4-5.2); duration >= 2 yrs., 5.4 Relationship of Relationship of Relationship of Relationship of endometrial cancer endometrial cancer endometrial cancer Number of cases endometrial cancer to duration of to dose of to recency of Author Description of cases and controls and controls to estrogen usea,b estrogen use a,b estrogen a,b estrogen use a,b Voight (1 991 ) Cases were all women diagnosed 158 cases (38% HRT Estrogen alone: Estrogen only use 23 with endometrial cancer between users); O.R. 3.1 (1 ,6-5.8) years: Jan. 1, 1985 and Dec. 31, 1987 who 182 controls (27% Estrogen plus 5.7 (2.5-1 2.8) were residents of King County, HRT users) progesterone: Estrogen use >3 Washington, and who were 40 to 64 O.R. 1.3 (0.6-2.8) years plus any use years of age at diagnosis ; cancer Progestin use <10 of progestin: 1.6 cases were identified through the (0.6-3.9) cancer surveillance system at the days per month plus Fred Hutchinson Cancer Research estrogen: O.R. 2,0 Estrogen 23 years plus progestin <10 Center. Controls were recruited by (0.7-5.3) random telephone digit dialing; Progestin use >= 10 days per me,: 2.4 controls were nonhysterectomized days per month plus (0.6-9.3) women who were residents of King estrogen: O.R. 0.9 Estrogen 23 years County. Information on HRT use was (0.3-2.4) plus progestin >= 10 gathered through interviews. days per month: 1.1 (0.4-3.6) a 95 percent confidence intervals are shown in parentheses b Relationship is relative risk, unless stated otherwise. KEY: NS = not statistically significant, SOURCE: Office of Technology Assessment, 1995. Relationship of endometrial o Relationship of endometrial cancer to cancer to duration, recency, a,b a,b Author Description of cohorts Size of cohort use and dose of HRT and latency of HRT use

Gambrell (1979) Participants were postmenopausal 8,170 patient-years (81% HRT Endometrial cancer incidence (cases per outpatients at Obstetrics and Gynecology use); 1,000 pt.-years). Clinic, Wilford Hall USAF Medical Center, 14 endometrial cancer cases Estrogen alone: 6.8/1 ,000 Texas. Duration of estrogen therapy PERT, 0.5/1 ,000 (p <0.01 compared to ranged from 2.5 to 12 years. Recruitment estrogen) between 1976-1977 was prospective, progestin alone: 0/1 ,000 1975 recruitment was retrospective. no therapy: 2/1 ,000 (NS compared to estrogen)

Bush (1983) Participants were white women, aged 40 2,270 white women (593 users, Endometrial cancer deaths. to 69 years at baseline, and followed for 1,677 nonusers) Nonusers: 1 death from an unspecified an average of 5.5 years. All women in the genitourinary cancer. cohort were participants in the Lipid Users: 1 death from uterine cancer, Research Clinics Program Follow-up Study, conducted in 10 North American clinics, between 1972 and 1976. All subjects were examined at initiation, and were followed with clinic visits and review of death certificates. Information on descendants was gathered from medical records and family members. . ..._ . .

Relationship of endometrial Relationship of endometrial cancer to cancer to duration, recency, Author Description of cohorts Size of cohort use and dose of HRTa,b and Iatency of HRT usea,b Lafferty (1985) Cohort members were postmenopausal 61 estrogen-treated women, 63 One case of endometrial cancer occurred women 45 to 60 years old followed at a untreated controls in untreated controls, and two in single private practice in Cleveland, OH. estrogen-treated women. No endometrial All treated patients received conjugated cancer deaths occurred in untreated equine estrogen 0.6 mg daily for three out controls and two deaths in of four weeks. Study was carried out estrogen-treated women. The difference in between 1966 and 1981, and patients rates of endometrial cancer deaths were were followed for an average of 8.6 years. not statistically significant, but the Patients were followed with physical population was very small. exams twice annually.

Gambrell (1986) Participants were post-menopausal 2,905 postmenopausal women NO use: 245.5 endometrial cancer cases women seen at Wilford Hall USAF Medical with 27,243 patient-years of per 100,000 patient-years. Center (Texas) using various hormone observation between 1975 and Unopposed estrogen: 390.6 per 100,000 regimens were compared to untreated 1983 (31 endometrial cancer (NS vs. no use) women. Three years of retrospective data cases). Estrogen and progesterone: were gathered for 1972-74 from medical and pharmacy records and tumor registry. 49.0 per 100,000 (p <0,0001 vs. Women were recruited between 1975 and unopposed estrogen users) (p =< 0.005 1979, and followed until 1983. Information vs. no use) on HRT use and risk factors gathered at Estrogen vaginal cream: 73.6 per 100,000 clinic visits. users (p <0,005 vs. unopposed estrogen users), None of the differences between the other groups were statistically significant,

I Relationship of endometrial cancer to Relationship of endometrial cancer to duration, recency, and Author Description of cohorts Size of cohort use and dose of HRT a,b latency of HRT usea’b

Stampfer (1986) Subjects were members of the nationwide 96,356 women in cohort with Current use of postmenopausal HRT: 4.4 Current use and duration of Nurses Health Study cohort. Cohort intact uterus who were free of (2,2-7.1); Among past HRT users, there use >5 years: 6.9 (3.6-13.2) members were registered nurses ages 30 cancer at baseline (no was an increased risk with increasing Current use and duration of to 55 years old in 1976. Subjects of this information on number of duration. use <1 year. 3.5 (1 .2-10 .8) study were cohort members who were postmenopausal), among free of cancer and had intact uteri. postmenopausal women in Information on HRT use and risk factors cohort, there were 70 cases of was gathered by questionnaire every two endometrial cancer in 114,896 years, and deaths were identified through years) state vital statistics records. There was 114,896 person-years of follow-up among postmenopausal women in cohort.

Petitti (1987) Subjects were participants in the Walnut 3,437 never users of estrogen, Endometrial cancer deaths. Creek (California) Contraceptive Drug 2,656 ever users of estrogen nonusers: 1 Study. Subjects were 18 to 54 at study users. 5 initiation, and were recruited between RR endometrial cancer death December 1968 and 1972. All subjects 2.6 (0.4-1 5.5) received a complete history and physical at study entry. Through 1977, reformation was gathered from clinic visits and mailed questionnaires. From 1978 to 1983, information was gathered from the California Death Index and death certificates. Oral contraceptive users were excluded from this analysis. . .

II Relationship of endometrial cancer to Relationship of endometrial cancer to duration, recency, and Author Description of cohorts Size of cohort use and dose of HRTa,b latency of HRT usea,b Pagnini-Hill Subjects were non-hysterectomized 5,160 non-hysterectomized Risk ratio for endometrial cancer in users Recency (years since (1989) women, aged 44-100 years (73 median) women is 10 (p < 0,0001) compared with cessation of estrogen), at baseline from the Leisure World nonusers, No effect of dose on risk was O-1 years. 25 (9,2-69) (California) Retirement Community, found. The relationship between HRT use 2-7 years: 12 (no confidence Subjects were recruited from June 1981 to and incidence of endometrial cancer is interval) January 1987. Of estrogen users, 99% reported in Henderson (1991) 8-14 years: 8.1 (non had used unopposed estrogen. Average (Henderson, 1991), confidence interval) duration of follow-up was 4.6 years. 15+ years: 5,8 (2.0-1 7) Information was gathered by periodic Duration of estrogen use,* questionnaires. < 2 years: 5.2 (no confidence interval) 3-7 years: 7.0 (no confidence interval) 8-14 years: 4 (no confidence interval) 15+ years: 20 (7,2-54)

*Paper also has table showing interaction of duration and years since cessation of therapy Relationship of endometrial cancer to Relationship of endometrial cancer to duration, recency, and Author Description of cohorts Size of cohort use and dose of HRPa,b latency of HRT usea,b Henderson (1991) Participants were residents of a Southern 8,881 postmenopausal women Relative risk endometrial cancer death: California retirement community (Leisure 3.0 (no confidence interval provided) in World), were almost entirely white, ever users vs. never users of estrogen. moderately affluent, and well educated. The relationship between use of HRT and Subjects were recruited between June endometrial cancer incidence in this 1981 and January 1987. The resident’s cohort is described in Pagnini-Hill (1989) median age at study initiation in 1981 was (Pagnini-Hill, 1989), 73 years. Information was gathered by periodic questionnaires and review of local county death certificates. Virtually all HRT users took unopposed estrogen. Reported here are the results of 7.5 years a 95 percent confidence intervals are shown in parentheses b Relationship is relative risk, unless stated otherwise. KEY: NS = not statistically significant; O.E. ratio = observed to expected ratio; PERT = estrogen/progestin combination therapy SOURCE: Office of Technology Assessment, 1995 Relationship of endometrial Relationship of endometrial cancer to cancer to duration, recency, Author Description of cohorts Size of cohort use and dose of HRPa,b and latency of HRT usea,b Hammond (1979) Participants were diagnosed between 301 “hypoestrogenic” patients O.E. ratio 9.3 (4.7-16.7) in white women 1940 and 1969 with diseases related to who received ERT; receiving estrogen: 1.1 (0.3-3.9) in white estrogen deficiency and followed for at 309 hypoestrogenic patients women not receiving estrogen least 5 years by the Duke University never receiving estrogen, All patients who developed Obstetrics and Gynecology Service 14 patients developed adenocarcinoma of the endometrium (Durham, NC). Expected rates of endometrial cancer during estrogen therapy had received this endometrial cancer were obtained from compound for at least five years. the Third National Center Survey for the Atlanta (Southeastern United States) area; 95.5% Of estrogen users received conjugated estrogens. Data on ERT use was obtained from hospital and clinic records. Vakil (1983) Study examined the incidence of 1,483 postmenopausal women Relative risk of endometrial cancer in ever endometrial cancer in a cohort of women, users 32-62 years of age, receiving estrogen 1.3 (no confidence interval provided treatment for menopausal symptoms among the patients of 20 gynecologists in the metropolitan Toronto area. Incidence rates in the cohort were compared to two control groups: the age-specific endometrial cancer incidence rates of the female populations of Ontario and of Saskatchewan. Estrogen therapy was begun between 1960 and 1970, and subjects were followed for up to 17 years Relationship of endometrial Relationship of endometrial cancer to cancer to duration, recency, a,b Author Description of cohorts Size of cohort use and dose of HRFa,b and latency of HRT USe Hunt (1987) This is the same cohort as described in 4,544 British women receiving O.E. ratio of endometrial cancer is 2.84 Latency (time since first use): Hunt (1990) (Hunt, 1990). Cohort HRT (43% PERT users) (1 ,46-4.96) for current users of at least 1 O-4 years. O.E. ratio 2.11 members were British women receiving year duration compared with expected (0.57-5.39) hormone replacement therapy recruited at incidence. No deaths from endometrial 5-9 years. 3.03 (1.1 1-6.60) 21 menopause clinics. Subjects were cancer occurred in the cohort. The 10+ years, 5,71 (0.64-20.63) recruited prospectively between 1978 and relationship of HRT use to endometrial There was evidence of a 1982, and retrospectively before 1978; cancer death are reported in Hunt (1990), rising trend in O.E. ratio with nearly equal proportions were recruited below. interval since first use, retrospectively and prospectively. Most although the trend does not cohort members were aged 45-54 years reach statistical significance. at recruitment. Thirty-six percent of cohort had undergone hysterectomy, 2-2.5 times the proportion in the British population. Mean duration of follow-up was 67 months. Cancer registry rates for England and Wales were used for determining expected incidence.

II Relationship of endometrial Relationship of endometrial cancer to cancer to duration, recency, Author Description of cohorts Size of cohort use and dose of HRTa,b and latency of HRT usea,b Ettinger (1 988) Subjects were female members of Kaiser 181 estrogen users, 220 Risk ratio for endometrial cancer is 7.7 Foundation Health Plan, San Francisco, nonusers controls (2.4-24.5) for users compared with CA, all who had filled at least 2 nonusers. prescriptions for an oral estrogen Endocarconima developed in 9.9% of preparation and were aged at least 53 users compared with 1.4% of nonusers. years in 1986. Estrogen users were menopausal women whose estrogen therapy was begun within three years of menopause and was used regularly for at least 5 years. Nonuser controls were women who had undergone spontaneous (nonsurgical) menopause, were identified from pharmacy records of health plan and were matched for age and length of membership in health plan. Mean age for estrogen users was 67 years, mean age of nonusers was 68.8 years. Clinical material was obtained from 1965 to 1980. Relationship of endometrial Relationship of endometrial cancer to cancer to duration, recency, o Author Description of cohorts Size of cohort use and dose of HRT a,b and latency of HRT use a,b Persson (1989) Cohort members were women age 35 23,233 women on estrogens All HRT users: Duration of estrogen use: years or older who had been prescribed (133,373 person-years); 74 1.4 (0.4-2.1) estrogen alone. estrogens for the treatment of menopausal cases of endometrial cancer estrogen alone. 1.4 (1.1 -1 .9) <6 mos.: 1.1 (0.5-2,5) problems in the Uppsala health care and 33 pre-malignant lesions estrogen and progestin: 0.9 (0.4-2.0) 7-36 mos.: 1.4 (0,8-2.4) region of Sweden during April 1977 to 37-72 mos.: 1.2 (0.6-2.2) March 1980, identified through >73 mos., 1,8 (1,1 -3.2) prescription records. Compliance, estrogen and progestin: sociodemographic data, and lifetime <6 mos.: O (0,0-1 2.7) exposures to estrogen and progesterone 7-36 mos.: 1.4 (0.5-3.6) were assessed by a mailed questionnaire 37-72 mos.: 1.2 (0.3-5.5) to 735 randomly selected members of the >73 mos.: O (0,0-456.1) cohort. In addition, characteristics of all women with endometrial cancer were Endometrial cancer and pre-malignant lesions assessed by questionnaires. Cases of estrogen alone. endometrial cancer were identified from a <6 mos.: 0,9 (0,4-2,1) cancer registry and medical records. 7-36 mos.: 1.6 (1 .0-2.5) Expected outcome in the cohort was 37-72 mos.: 1.6 (1 .0-2.6) determined from age-specific incidence >73 mos.: 2,7 (1 ,8-4,2) rates of endometrial cancer in the region in the same years. Pathologic specimens estrogen and progestin: from all endometrial cancers and <6 mos.: 0,9 (0.2-4.3) pre-malignant lesions in the cohort and 7-36 mos.: 1.6 (0.7-3.5) the background population were 37-72 mos.: 0.9 (0.2-4.1) reviewed. Average observation period >73 mos.: O (0.0-211 8) Relationship of endometrial Relationship of endometrial cancer to cancer to duration, recency, Author Description of cohorts Size of cohort use and dose of HRTa,b and Iatency of HRT usea,b Hunt (1990) This is the same cohort as described in 4,544 long term users of HRT Observed endometrial cancer deaths. O Hunt (1987) (Hunt, 1987). Subjects were (43% PERT users) expected endometrial cancer deaths: women recruited from 21 menopause 2,70 (taking into account uterine status) clinics around Britain; all had received at least 1 year continuous treatment with O/E ratio: 0.00 (0.00-0.97) hormonal replacement therapy before The previous report, Hunt (1987), recruitment, All subjects were interviewed however, noted an elevated risk of at recruitment. Most subjects were age 45 incident endometrial cancer (see above) to 54 at first use of HRT. Mean duration of HRT use was 66.9 months; 59& were current estrogen users. The observed mortality was compared to the expected rates in the female population of England and Wales. a 95 percent confidence intervals are shown in parentheses. b Relationship is relative risk, unless stated otherwise KEY: NS = not statistically significant; O.E. ratio = observed to expected ratio; PERT = estrogen/progestin combination therapy, SOURCE” Office of Technology Assessment, 1995 Appendix H: Evidence on HRT and Gallbladder Disease H

here are several theoretical reasons to ex- der disease and HRT. A small number of studies pect a causal link between estrogens and have shown that the incidence of symptomatic gallbladder disease. In particular, estro- gallbladder disease increases approximately two- gens increase the risk of gallstone forma- fold in current users of oral estrogen replacement Ttion. Gallstones form in the gallbladder, a therapy. The first report of an association came muscular sac in the abdomen that stores and re- from a case-control study conducted in the leases bile, a substance that aids in fat digestion. mid- 1970s by the Boston Collaborative Drug Sur- Gallstones are solidified bile. Bile is highly satu- veillance Program, which showed an increased in- rated with cholesterol, and it is thought that estro- cidence of surgically confirmed gallbladder gen raises the concentration of cholesterol in bile, disease in current users of either oral contracep- increasing the risk of stone formation (2). Thus, tive or oral estrogen replacement therapy (3).2 one would expect an increased prevalence of gall- One of the two prospective cohort studies of es- stones and symptomatic gallbladder disease in estrogen replacement therapy and gallbladder dis- trogen users (1 1).l ease found that women who were current or past Tables H-1 to H-4 summarize the clinical stud- users of noncontraceptive estrogen had an age-ad- ies evaluating the relationship between gallblad- justed relative risk of symptomatic gallstone dis-

] It is not known whether estrogens not taken by mouth would also increase the risk of gallstones. Some argue that estrogen taken by skin patch or injection would not increase the risk of gallstone formation because estrogen taken by nonoral routes does not pass directly from the intestine to the liver. By avoiding the first-pass effect on liver metabolism, nonoral routes of estrogen administration may reduce this increased risk. A study by D’Amato and colleagues compared the effect of 17-beta estradiol given by skin patch and estradiol valerate given by mouth on bile lipid levels in a postmenopausal woman (4). While both therapies increased the cholesterol level in the bile, only oral estrogen induced the formation of cholesterol crystals. 2 The Boston Collaborative Drug Surveillance Program study was criticized for using hospitalized controls, one half of whom were being treated for fracture or some other orthopedic problem. One commentator has argued that, since women with osteoporosis are less likely to be taking estrogen replacement therapy, this design could have led to a spuriously low rate of estrogen use in the comparison group, compared with usual use in the cases, and thereby a falsely elevated relative risk (1). 159 Number of cases Author Description of cases and controls and controls Resultsa — Boston Collaborative Drug Cases were postmenopausal women 45 to 69 years old 152 cases, Relative risk 2,5 (1 .5-4,2), there was “no Surveillance Program with a diagnosis of “cholelithiasis” or “cholecystitis” 774 controls evidence of a relation with duration of use in (1974) and subsequent cholecystectomy, who were postmenopausal estrogen users. ” admitted to general medical and surgical wards of 24 o hospitals in the Greater Boston area between January and November 1972. Patients with diseases that might either contraindicate estrogen therapy or be related to their use were excluded. Controls were hospital patients without a diagnosis of gallbladder disease, venous thromboembolism, or breast tumors,

Honore (1980) Cases were 262 perimenopausal women (ages 41 to 60 262 cases; Relative risk 3.72 (p < 0.005). There was a years) with symptomatic gallbladder disease treated 290 controls significantly greater incidence of gallbladder by cholecystectomy from 1975 to 1978 at a hospital disease in obese HRT users than in nonobese in Newfoundland, Canada, and diagnosed HRT users (p < 0.05), pathologically as having cholesterol gallstones. A control group, matched for age, consisted of women treated surgically for diseases that have no known association with estrogen replacement therapy. Information on HRT use was obtained from a review of medical records.

Scragg (1984) Cases were patients in 2 public hospitals in Adelaide, 200 cases, Mean duration of estrogen use was not Australia with gallstone disease diagnosed by 234 hospital controls, substantially different between cases and both ultrasound or cholecystectomy between December 82 community control groups 1978 and September 1980, Two control groups were controls used for comparison. Hospital controls were women who were hospitalized and had negative cholecystograms. Community controls were women from the community, matched to cases for age and area of residence.

Kakar (1988) Subjects were women ages 41 to 74 enrolled in a 102 cases, Relative risk 1 18 (0,65-2,13) for users of 1 year prepaid health plan in western Washington’s Group 98 controls or more vs nonusers, Standardization for the Health Cooperative of Puget Sound Cases were effects of age, race, obesity, parity, thiazide women who underwent gallstone surgery between use, and diagnosis of high blood pressure did January 1979 and September 1988. Controls were not alter appreciably the estimate of relative matched for sex, age, and residence with cases. risk.

II Number of cases Author Description of cases and controls and controls Resultsa LaVecchia(1992) Subjects were women admitted to one of four hospitals 235 cases; Users of any duration: in Milan, Italy between 1987 and 1990. Cases were 583 controls unadjusted relative risk: 1.7 (0.9-3,1) women, ages 23 to 74 (median age 54), who adjusted relative riskb: 1.9 (1 ,0-3.6) underwent cholecystectomy, and were discharged Use less than 2 years: with the diagnosis of cholelithiasis or cholecystitis. unadjusted relative risk: 1.7 (0.8-3.6) Controls were women, ages 21 to 74 (median age adjusted relative riskb: 1.8 (0.9-4,2) 54), admitted for acute diseases other than digestive Use 2 or more years: or hormonal diseases or those potentially influencing unadjusted relative risk: 1.3 (0.5-3.8) the use of female hormone preparations. adjusted relative riskb: 1.5 (0,5-4.5) x Less than 10 years since last use: unadjusted relative risk: 1.1 (0.5-2.7) adjusted relative riskb: 1.3 (0.5-3,3) Last use 10 or more years ago: unadjusted relative risk: 2.3 (1 ,0-5.3) adjusted relative riskb: 2.4 (1 ,0-5.1) a The results are followed by 95% confidence intervals in parenthesis. b Relative risk was adjusted for age, education, area of residence, body mass index, parity, and age atmenopause. SOURCE: Office of Technology Assessment, 1995. Number of Author Description of study participants participants Resultsa Diehl (1980) Subjects were obtained from a review of a sample of 1,018 records “No trends in the prevalence of gallbladder o medical records from patients enrolled in the Family disease were seen in relationship to use of Health Center of the University of Texas Health conjugated estrogens. Our failure to Sciences Center at San Antonio. Gallbladder disease find associations with estrogen-containing was defined as history of cholecystectomy, drugs may be related to our inability to gallbladder surgery, or abnormal cholecystogram. quantitate their cumulative use in our study population. ”

Petitti (1981) Subjects were adult female twins who volunteered to 868 female twins Relative risk 2.0 (1 .1-3.6) for history of undergo a health examination for a study. Subjects physician-diagnosed gallbladder disease in were considered to have gallbladder disease if they estrogen users versus nonusers. answered “yes” to the question, “Has a doctor ever told you that you have had gallstones or gallbladder trouble?”

Pixley (1 985) Subjects were women aged 40 to 69 registered at two 632 women recruited from Study concludes “no association [of gallstones] Oxford, England general practices. All subjects were general practice registers was found with parity or use of exogenous screened with ultrasound for gallstones. Gallbladder and 130 vegetarians. estrogens. ” No further information or statistical disease was defined as cholelithiasis on ultrasound analyses was provided on this issue. cholecystectomy.

Jorgensen (1988) Subjects were a random sample of women from 2,301 women Odds ratio 1.02 (0.25-4.26) for current or past Copenhagen county, Denmark, ages 30, 40, 50, and gallbladder disease in estrogen users versus 60 years, drawn in 1982 from the National Person nonusers. Odds ratio 1.86 (0.89-3.86) for Register. Subjects were examined and/or interviewed estrogen users of 8 or fewer years versus by telephone or mailed questionnaire. Examined users of more than 8 years. patients received ultrasonography to identify current gallstone disease. a The results are followed by 95% confidence Intervals in Parentheses unless otherwise specified SOURCE: Office of Technology Assessment, 1995

II Author Description of study participants Size of cohort Resultsa

Petitti (1 988) Subjects were women 18 to 54 years old at time of entry into the 16,638 women All ever users: Walnut Creek (California) Contraceptive Drug Study cohort between unadjusted relative risk 2.4 (1 .7-3.2) age-adjusted December 1968 and February 1972. Women who ever used oral relative risk 2.1 (1 .5-3.0) contraceptives were excluded from this analysis. Women with a past usersb: previous cholecystectomy were also excluded from this analysis. unadjusted relative risk 1.8 (1.1 -2.9) Results of 10 to 13 year followup are presented. Patients were age-adjusted relative risk 1.6 (1.1 -2.5) examined at initiation of study and followed by examination, current users: questionnaire and/or reexamination. Cases were women who unadjusted relative risk 3.1 (2.2-4.9) underwent cholecystectomy for cholelithiasis or cholecystitis. age-adjusted relative risk 2.7 (1 .8-4.0) There was no evidence of a relationship of incidence of cholecystectomy with duration of estrogen use.

Grodstein (1 993) Subjects were postmenopausal U.S. registered nurses who were 54,845 postmenopausal Current users: enrolled in the Nurses Health Study. Information on postmenopausal women risk-factor adjusted relative risk 2.1 (1.9-2.4) estrogen use and cholecystectomy was gathered by mailed Current users of 10 years or more. questionnaires every two years. Duration of follow-up was 8 years. risk-factor adjusted relative risk 2.6 (2.2-3.1 ) Current users of 1.25 mg CEE per day or more: risk-factor adjusted relative risk 2.4 (2.0-2.9) Past users of less than 2 years duration. relative risk 1.4 adjusted for recency of use Past users of 10 or more years: relative risk 1,7 adjusted for recency of use Most recent use 1 to 2.9 years ago. risk-factor adjusted relative risk 1.6 (1.2 to 2.0) Most recent use 5 or more years ago. risk-factor adjusted relative risk 1.3 (1.1 to 1.6)

a The results are followed by 950/o confidence Intervals in parenthesis b Author notes that "We reviewed the medical records at the 39 past estrogen users who had cholecystectomies after 1977 and discovered that 12 Of them had reinitiated estrogen use after 1977 and before their hospitalization for gallbladder disease When those 12 women were removed from the cases that had been considered past users, the age-adjusted relative risk of gallbladder disease in past users decreased to 1.1 (95% confidence interval 0.7-1.8) When these 12 women were added to the current users, this relative risk estimate for current use increased to3 9 (95% confidence interval 2,6-5 .9. As study subjects who never experienced cholecystectomy and who initiated estrogen use after followup could not be relocated, these risk estimates where biased downward for past estrogen use and upward for current estrogen use “ KEY: CEE = conjugated equine estrogen SOURCE Office of Technology Assessment, 1995 164 Cost Effectiveness of Screening for Osteoporosis

Duration of Results Author Description of study participants study — Nachtigall Subjects were 84 pairs of postmenopausal 10 years Incidence of cholelithiasisa in treatment (1979) inpatients at Goldwater Memorial Hospital in group: 4/84 (0.48%) New York City, a hospital for chronic Incidence of cholelithiasis control group diseases, matched for age and diagnosis. 2/84 (0.24%) Treatment group received 2.5 mg CEE daily P>=0.05 (nonsignificant)b with 10 mg medroxyprogesterone acetate for 7 days each month. Control group received placebos.

a Author did not define “cholelithiasis. ” b Author notes "These findings, however, should be interpreted with Caution Since the power of the test differences was generally low due tO the small sample size. ” SOURCE: Off Ice of Technology Assessment, 1995 ease of approximately 2.1, while current users had Although empirical studies have found an a relative risk of 2.7 (16) (table H-3). The other co- increase in symptomatic episodes, hospitaliza- hort study found, after adjusting for confounding tion, and gallbladder removal (cholecystectomy) factors, a relative risk of cholecystectomy of 2.6 among current estrogen users, they have failed to in long-term current estrogen users. The only con- detect an increased prevalence of gallstones trolled clinical trial of estrogen use and gallblad- among estrogen users using imaging techniques der disease found doubled the incidence of capable of detecting silent gallstones (6,11,17). gallbladder disease in HRT users (12) (table H-4). The failure to find increased incidence of asymp- This difference did not reach statistical signifi- tomatic gallstones raises the possibility that the cance, which may be due to the small number of studies examining symptomatic disease may be women who participated in this study. subject to surveillance biases (i.e., estrogen- Some studies of symptomatic gallstone dis- treated women are seen more frequently by their ease, however, have found no effect of estrogen doctors and are therefore more likely to be diag- use on gallbladder disease (9). The differences nosed and undergo surgery) (l). among studies may be due to differences in the Studies have not been able to consistently dem- doses of estrogens used by participants, the aver- onstrate an increased risk of gallbladder disease age duration of use of estrogen, or the small num- with increased duration of use of estrogen replace- bers of persons involved in these studies (14). ment therapy (3, 16, 16a). Results from the Nurses These studies also were either case-control studies Health Study cohort demonstrated an increased or cross-sectional studies, which may have biased risk with duration of use in current users, but little their outcomes. or no effect of duration in past users (16a) (table Although the strongest evidence, including the H-3). prospective cohort studies of the issue, points to The results from the Nurses Health Study also an elevated risk of symptomatic gallstone disease showed an increased risk with larger doses of es- among current users, it is less certain whether the trogen (16a). This result is consistent with an ear- risk of symptomatic gallstone disease remains ele- lier cohort study of oral contraceptive users that vated in those who have ceased estrogen therapy. found an increase in risk with increasing dose The studies to date have not found a statistically (19). significant relationship between past use and gall- The addition of progestins is unlikely to miti- stone disease. gate estrogen-induced increases in gallbladder Appendix H Evidence on HRT and Gallbladder Disease 165

disease, since progestins also promote gallstone 4. D’Amato, G., Cavallini, A., Mesa, C., et al., formation (11). “Serum and Bile Lipid Levels in a Postmeno- On the basis of the studies outlined in tables pausal Woman After Percutaneous and Oral H-l to H-4, including the cohort studies of this is- Natural Estrogens,” American Journal of Ob- sue(16), 0TA adopted a base case assumption that stetrics and Gynecology 160:600-601, 1989. the risk of symptomatic gallbladder disease would 5. Diehl, A.K., Stern, M. P., Ostrower, V. S., et be elevated by a factor of 2.5 while a woman is on al., “Prevalence of Clinical Gallbladder Dis- HRT. The risk would subside to that of the general ease in Mexican-American, Anglo, and Black population of women at the time that HRT ceases. Women,” Southern Medical Journal We believe that the possible values of the relative 73(4):438-443, 1980. risk of symptomatic gallbladder disease due to 6. Everson, R., Byar, D., and Bischoff, A., “Es- current HRT range from 1.0 (best case) to 3.0 trogen Predisposes to Cholecystectomy but (worst case). not to Stones,” Gastroenterology 82(1):4-8, The definitive treatment for gallbladder disease 1982. is cholecystectomy, a standard surgical procedure 6a. Grodstein, F., Colditz, G. A., Stampfer, M. J., that is rarely fatal (13). For this analysis, we have “Prospective Hormone Use and Cholecystec- assumed that gallbladder disease results in health tomy in a Large Prospective Study,” Obstet- care costs for surgical removal of the gallbladder rics and Gynecology 83(1):5-11, 1993. and hospitalization. We have assumed, however, 7. Honore, L. H., ’’Increased Incidence of Symp- that gallbladder disease does not affect the years tomatic Cholesterol Cholelithiasis in Peri- of life lived. menopausal Women Receiving Estrogen OTA’s sensitivity analysis shows that our as- Replacement Therapy: A Retrospective sumptions about the risk of gallbladder disease in Study,” Journal of Reproductive Medicine HRT users does not affect the outcome of the anal- 25(4):187-190, 1980. ysis greatly, since gallbladder disease affects 8. Jorgenson, T., “Gall Stones in a Danish Popu- health care costs but not years of life lived. lation: Fertility Period, Pregnancies, and Ex- ogenous Female Sex Hormones,” Gut REFERENCES 29:433-439, 1988. 1. Barrett-Connor, E., “Putative Complications 9. Kakar, F., Weiss, N., and Strite, S., “Non- of Estrogen Replacement Therapy: Hyperten- Contraceptive Estrogen Use and the Risk of sion, Diabetes, Thrombophlebitis, and Gall- Gallstone Disease in Women,” American stones,” The Menopause. Biological and Journal of Public Health 78(5):564-566, Clinical Consequences of Ovarian Failure: 1988. Evolution and Management, S. Korenman 10. La Vecchia, C., Negri, E., D’Avanzo, B., et (cd.) (Norwell, MA: Serono Symposia, USA, al., “Oral Contraceptives and Non-Contra- 1990). ceptive Oestrogens in the Risk of Gallstone 2. Bennion, L., Ginsberg, R., Garnick, M., et al., Disease Requiring Surgery,” Journal of Epi- “Effects of Oral Contraceptives on the Gall- demiology and Community Health bladder of Normal Women,” New England 46(3):234-236, 1992. Journal of Medicine 294(4): 189-192, 1976. 11. Mack, T., and Ross, R., “A Current Percep- 3. Boston Collaborative Drug Surveillance Pro- tion of HRT Risks and Benefits,” Osteoporo- gram, “Surgically Confirmed Gallbladder sis: Physiological Basis, Assessment, and Disease, Venous Thromboembolism, and Treatment, H.F. DeLuca and R. Mazess (eds.) Breast Tumors in Relation to Postmenopausal (New York, NY: Elsevier Science Publishing Estrogen Therapy,” New England Journal of Co., Inc., 1990). Medicine 290(1):15-19, 1974. 166 Cost Effectiveness of Screening for Osteoporosis

12. Nachtigall, L.E., Nachtigall, R. H., Nachti- 16. Petitti, D., Sidney, S., and Perlman, J., “In- gall, R. D., et al., “Estrogen Replacement creased Risk of Cholecystectomy in Users of Therapy II: A Prospective Study in the Rela- Supplemental Estrogen,” Gastroenterology tionship to Carcinoma and Cardiovascular 94:91-95, 1988. and Metabolic Problems,” Obstetrics & Gy- 17. Pixley, F., Wilson, D., McPherson, K., et al., necology 54(1):74-79, 1979. “Effect of Vegetarianism on Development of 13. Nahrwold, D.L., “Acute Cholecystitis,” Text- Gall Stones in Women,” British Medical book of Surgery: The Biological Basis of Journal 291:11-12, 1985. Modern Surgical Practice, 14th Ed., D.C. Sa- 18. Scragg, R. K., McMichael, A.J., and Sea- biston, Jr., (cd.) (Philadelphia, PA: W.B. mark, R. F., “Oral Contraceptives, Pregnancy, Saunders Co., 1991). and Endogenous Oestrogen in Gall Stone Dis- 14. Petitti, D., “Gallbladder Disease and Use of ease - A Case-Control Study,” British Medi- Noncontraceptive Estrogens,” Geriatric cal Journal 288: 1795-1799, 1984. Medicine Today 9(3):56-63, 1990. 19. Strom, B., Tamragouri, R., Morse, M., et al., 15. Petitti, D. B., Friedman, G.D., and Klatsky, “Oral Contraceptives and Other Risk Factors A. L., “Association of a History of Gallblad- for Gallbladder Disease,” Clinical Phar- der Disease with a Reduced Concentration of macology and Therapy 39:335-341, 1986. High-Density-Lipoprotein Cholesterol,” New England Journal of Medicine 304(23): 1396-1398, 1981. Appendix I: Evidence on HRT and Coronary Heart Disease I

oronary heart disease (CHD) is the leading development of CHD (57,75). Estrogens also cause of death among U.S. women, sur- lower the serum concentration of low density lipo- passing the rates from cancer and other protein cholesterol (LDL), and LDL levels are di- diseases (13). Any change in the risk of rectly related to the development of coronary heart CHDc due to hormone replacement therapy (HRT) disease (8,15,16,23,46,53,57,86,). would profoundly alter the risk-benefit equation The Lipid Research Clinics Follow-up study of HRT. showed that women using conjugated equine es- Prior to menopause, women have a lower inci- trogens at the usual doses indicated for postmeno- dence of CHD than men. The Framingham study pausal women had HDL levels 16.8 percent showed that men had three times the incidence of higher than women not taking estrogens (15). Es- heart disease of age-matched premenopausal trogen users also had LDL levels approximately 7 women (50). Women within the first few years af- percent lower than those of nonusers. Coronary ter natural menopause have no substantial in- heart disease deaths were reduced by 65 percent in creased risk of heart disease over premenopausal estrogen users compared with nonusers. The in- women (21). However, by age 70, the incidence vestigators concluded that this benefit was sub- of CHD is approximately the same in women and stantially mediated by the increase in HDL levels. men. Moreover, after surgical or premature meno- Recent research has demonstrated that eleva- pause women develop a substantially increased tions of HDL and decreases in LDL may also oc- risk of CHD at an earlier age than women who un- cur with percutaneous and subdermal estrogen dergo natural menopause at a later age (89). administration (46,105). However, there is evidence that transdermal estrogens do not produce HOW ESTROGEN MAY AFFECT CHD the same degree of favorable alterations of lipo- One mechanism for a possible beneficial effect of protein cholesterol levels as oral estrogens (10, estrogen against CHD is the ability of estrogen to 20,1 15). Oral estrogen has a much greater lipid ef- favorably alter lipoprotein levels. Estrogen use fect (increasing HDL, decreasing LDL) than a has been shown to increase the level of high densi- comparable transdermal dosage, perhaps because ty lipoprotein cholesterol (HDL) (8,16,23,53,57, of higher concentrations of estrogen in the portal 75,101). Studies have demonstrated that serum circulation of the liver with oral therapy. concentrations of HDL are inversely related to the 1167 168 Cost Effectiveness of Screening for Osteoporosis

The cardioprotective effect of hormone re- Rosano demonstrated in a clinical trial the im- placement therapy may also be mediated through mediate effect that estrogens have on heart disease lowering lipoprotein(a), an independent risk fac- (87). The investigators studied the acute effect of tor for heart disease in postmenopausal women sublingual estradiol on exercise tolerance and (69,95,100). angina in 11 women with coronary artery disease. There is evidence that estrogen protects the The women did two exercise treadmill tests heart by reversing other changes in metabolism (EKG) on two separate days. Forty minutes before that occur at menopause ( 106). Estrogen has been each test, they took sublingual estradiol or pla- found to reverse the unfavorable effects of meno- cebo, in random order. Six patients developed ex- pause on glucose and insulin metabolism (69,74). ertional angina and EKG changes after admin- Central obesity is linked to heart disease risk, and istration of sublingual estradiol, whereas all 11 estrogen may reverse the changes in body fat dis- developed angina and EKG changes on placebo. tribution that results from loss of estrogen produc- The authors posited that this immediate beneficial tion at menopause (106). effect of estrogens maybe due to a direct coronary Estrogen may also exert its heart protective ef- artery relaxant effect of estrogen, dilation of pe- fects by favorably altering the balance between ripheral arteries and arterioles, or to a combination coagulation and fibrinolysis (18,82,97), by inhib- of these mechanisms. iting platelet function (5), or by relaxing arterial walls (58,124).1 Estrogen increases production of EVIDENCE ON ERT AND CHD prostacyclin, a prostoglandin in the arterial wall All but four of the more than 30 studies that have (82) that reduces platelet aggregation (70) and evaluated the effect of estrogen replacement thera- causes dilatation of the blood vessels (124). In py (ERT) on coronary heart disease (CHD) have coronary artery occlusion, the release of throm- shown a reduced risk in estrogen users. The fol- boxane may induce the aggregation of platelets lowing is a discussion of the evidence on the rela- and reduce blood flow. Prostacyclin counteracts tionship between ERT and cardiovascular disease the effect of thromboxane by reducing platelet ag- risk. Coronary evidence falls into five categories gregation and increasing blood flow, and in this based on methods and data sources: way may reduce the risk of coronary artery occlu- ■ hospital-based case-control studies, sion. ■ population-based case-control studies, Estrogen may also protect the heart by favor- ■ prospective cohort studies, ably altering cardiovascular hemodynamics. Re- ■ cross-sectional studies, and ceptors for estrogen have been found on arterial ■ randomized clinical trials. walls (52,68), and estrogen may directly relax the arteries throughout the body (22,25,58). By re- Each is discussed in turn. ducing the resistance to blood flow through the arteries, the work load on the heart is reduced (22). ❚ Hospital-Based Case-Control Studies By reducing the workload of the heart, its oxygen The earliest studies examining the risk of coro- needs are reduced. Thus, there is less likelihood nary heart disease in noncontraceptive estrogen that the oxygen requirements of the heart will ex- users used as “cases” individuals hospitalized for ceed the oxygen that is available from blood flow- myocardial infarction (heart attack) over a speci- ing through partially occluded coronary arteries fied time period. “Controls” were a comparison (87). group of patients with other diagnoses from the

1 For recent reviews of the potentially important nonlipoprotein-mediated mechanisms of reduction in coronary heart disease risk, see K.F. Ganger, B.A. Reid, D. Crook, et al., 1993; M. Riedel, W. Raffenbeul, and P. Lichtlen, 1993; J.C. Stevenson, D. Crook, I.F. Godsland, et al., 1994; M.J. Tikkanen, 1993. Appendix I Evidence on HRT and Coronary Heart Disease 169

same hospitals as the cases. The researchers then The first case-control study that did not detect a determined which women in each group had or lower risk of CHD in estrogen users was of hos- had not had ERT in the past through interviews pitalized patients aged 40 to 75 enrolled in the with the women, medical records, and other Boston Collaborative Drug Surveillance Pro- sources. gram. The relative risk of nonfatal myocardial in- Five hospital-based case-control studies have farction (MI) in estrogen users was 0.47, but the examined ERT among patients hospitalized for relative risk was not significantly different from myocardial infarction. (See table I-1.) Of these, one after statistical adjustment for differences in one showed an increased risk of coronary heart heart disease risk factors between the two groups disease among estrogen users (48), two showed (88). Also, heart disease risk is thought to be more virtually no change in risk (88,91), and two markedly reduced among those who are currently showed a decreased risk of CHD that was not sta- using estrogen (“current users”) compared with tistically significant.2 those that have used estrogens in the past (14), and A well recognized problem with case-control only eight of the 336 cases in the study (2 percent) studies is that the ascertainment of exposure to the were “current users” of estrogens. agent in question (e.g., estrogen) often depends on The second study finding no decreased risk was the recall of the study participants. Because cases of women aged 30 to 49 years old admitted to hos- may differ from controls in the accuracy of recall pital coronary care units (91). The adjusted rela- of exposure, a biased estimate of risk can occur. tive risk was near one both in patients who had A second problem particular to hospital-based ever used estrogen (“ever users”) and in “current case-control studies is that a control group com- users.” The results of this study may not be gen- posed of hospitalized patients is not likely to be eralizable to all postmenopausal women because representative of the general population from it was conducted among women under 50 years of which the cases were drawn with regard to expo- age. Because of their young age, these women had sure to estrogen. In the context of ERT, the results infrequent use of ERT and were at minimal risk of of hospital-based case-control studies are difficult coronary heart disease. Moreover, a substantial to interpret because many diseases are related in proportion of controls in this study were fracture some way to estrogen use. For example, some patients (13,103). members of the control group may have been hos- Jick and colleagues reported the highest rela- pitalized because of fracture, and women with tive risk of coronary heart disease in estrogen us- fracture are less likely to have used estrogen. ers among all studies (48). They reported a Even selecting controls from patients with dis- relative risk of first nonfatal MI of 7.5 (95 percent eases unrelated to estrogen use is problematical, confidence intervals 2.4 to 24) among estrogen us- because some physicians may be less willing to ers under 46 years of age and a relative risk of 4.2 prescribe ERT to patients who are already bur- (95 percent confidence intervals 1.0-18.8) among dened with other medications (103). The net effect postmenopausal estrogen users. This study had a of this last bias would be to underestimate the im- small number of cases and a large loss of study pact of estrogen on heart disease risks. participants over time. Sixteen of the 17 cases (94

2 The change in risk of disease in these studies is expressed either as a relative risk or as an odds ratio (42,94). The odds ratio is obtained from the exposure ratio in the cases divided by the exposure ratio in the controls. To determine the odds ratio in a hospital-based case-control study of myocardial infarction and estrogen use, one would calculate the ratio of estrogen users to nonusers among myocardial infarction patients (cases) and divide that by the ratio of estrogen users to nonusers in the comparison patients hospitalized with other diagnoses (controls). Results of a case-control study can also be expressed as a relative risk, which is the rate with which the disease occurs in exposed people divided by the rate of the disease’s occurrence in unexposed people. If these two rates of occurrence are very small and if no distortions have occurred in the four groups that make up the case-control study, the odds ratio will be approximately equal to the risk ratio. Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpointsa,b Rosenberg (1976) Cases were two sets of patients hospitalized Cases (set 1: 163; set 2: 173) Nonfatal Ml Current users: for myocardial infarction in the Boston (2.4% conjugated estrogen age-adjusted relative risk Collaborative Drug Study. First set of cases users); controls (set 1: 2,536; 0.71 (0.34-1 .46) o was from 21 hospitals in the United States, set 2: 4,194) risk-factor adjusted* relative risk Great Britain, Canada, Germany, New 0.97 (0.49-1 .95) Zealand, Italy, and Israel, admitted since 1969, Second set was from general medical *adjusted for age, history of Ml, angina, diabetes, and surgical wards of 24 Boston hospitals in hypertension, and smoking 1972. Study subjects were ages 40 to 75 years; average age 54 years. Controls were patients from same hospital admitted for neoplasm, gallbladder disease, and breast or reproductive organ disease. Data were obtained from interviews and hospital records. Current use was defined as use during the month prior to hospitalization. Jick (1 978a) Cases were women ages 39 to 45 years of 17 cases (53% estrogen users); First nonfatal Ml Current estrogen use: age discharged within the first 6 months of 34 controls (12% estrogen 7.5 (90% confidence interval 2.4 1975 with a diagnosis of AMI. Cases were users) to 24) identified from a national hospital discharge adjusted for type of sterilization database. Controls were drawn from the Ninety-four percent of cases, but only national hospital discharge database, were 47% of the controls, were cigarette about the same age as cases, were smokers. hospitalized for acute illnesses (other than Ml) or elective surgery, and were discharged about the same time as cases. Both cases and controls had no other illnesses that predisposed to Ml or contraindications to estrogen use. Cases and controls had a natural menopause, hysterectomies, or tubal ligation, or their husband had a vasectomy. Current estrogen use was defined as use of noncontraceptive estrogens within 3 months of admission.

I Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpointsa,b Jick (1 978) Cases were women ages 35 to 45 discharged 19 cases (53% estrogen users), First nonfatal Ml Current estrogen use during the first 6 months of 1975 with a 39 controls (10% estrogen 9.3 (lower 95% confidence interval 3.1) diagnosis of AMI. Cases and controls were users) adjusted for menopausal status identified from a national hospital discharge database. Controls were women about the same age as cases, who were hospitalized for acute illnesses (other than Ml) or surgery, and discharged about the same time as cases. Results are reported for cases and controls who had no serious chronic illnesses (other than Ml in cases) or contraindications to estrogen use. Current use was defined as use of noncontraceptive estrogens within three months of admission. Rosenberg (1980) Cases were women ages 30 to 49 years of 99 cases post menopausal First Ml Current users: age selected from interviews between July (18% current users) (24% age-adjusted relative risk 1976 and April 1979 with a discharge past users); 463 controls 1.39 (0.71 -2,74) diagnosis of first Ml. Hospitals were located in Greater Boston, Long Island, New York, and the coastal area of northern New York City and the Delaware Valley. Controls were selected from the same hospitals as cases but did not have a discharge diagnosis of Ml. Information was gathered by nurse interviewers. Current use was defined as use within the month preceding admission, Szklo (1 984) Cases were white female patients 35 to 64 39 cases (28% ever users), 81 First Ml Ever users years of age admitted to 5 general hospitals controls age-adjusted OR 0,8 (NS) in Maryland with a first Ml between risk-factor adjusted* OR 0,61 1971-1972. Two controls were matched to (0.20-1 .88) each case. Controls were females from the risk-factor adjusted* OR for surgical same hospitals as cases, with no history of Ml menopause only 0.37 (0.04-3,23) or abnormal Q waves on EKG, and matched by age and date of admission. Data were *adjusted for history of cardiovascular disease, obtained from interviews and review of smoking, educational level, and type of medical records, menopause Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpointsab

La Vecchia (1987) Cases were women less than 55 years of age 168 cases, 100 First Ml Current users: admitted between January 1983 and pre-menopausal (5% current age-adjusted relative risk December 1984 to the coronary care units of users) (3% past users); 251 1.85 (0.68-5.01) 30 hospitals in Northern Italy. Controls were controls risk-factor adjusted* relative risk matched to cases for index hospital and 2.95 (0.80-10.80) 5-year age group. Controls were admitted for Past users: acute conditions except cardiovascular, age-adjusted relative risk cancer, endocrine, gynecological, or primary 1,01 (0.31-3.27) diagnosis potentially related to cigarette risk-factor adjusted* relative risk smoking or hormone use. Data were gathered 0.77 (0.1 6-3.60) by trained interviewers. ‘adjusted for multiple heart disease risk factors “No relation was evident with duration of use. ” a Numbers in parentheses are 95 percent confidence intervals. b Risk estimates are in terms of relative risk, unless otherwise specified, KEY: AM I = acute myocardial infarction; EKG = electrocardiogram; ERT = estrogen replacement therapy; Ml = myocardial infarction;NS = not statistically significant; OR = odds ratio. SOURCE: Office of Technology Assessment, 1995.

II Appendix I Evidence on HRT and Coronary Heart Disease 173

percent) were smokers, which confounds inter- physicians’ practices were matched with two con- pretation of results. Also, the subjects were under trols from the same physician’s practice and of the 50 years of age, so the findings may not be gen- same age. Estrogen use was ascertained from eralizable to the overall population of postmeno- medical records and patient interviews. Thomp- pausal women (14). son showed a “weak” association between estrogen use and stroke and myocardial infarction, with ❚ Population-Based Case-Control Studies a relative risk of 1.36 in estrogen users (95 percent Among the seven population-based case-con- confidence intervals 1.01 to 1.81). An association trol studies of myocardial infarction and ERT, all between estrogen use and decreased risk of coro- but one demonstrated a trend toward decreased nary heart disease may have been obscured by relative risk of myocardial infarction in estrogen combining the myocardial infarction endpoint users, although the results were statistically sig- with the endpoint of stroke. nificant in only one of the studies that showed a trend toward decreased relative risk (92). (See ❚ Cohort Studies table I-2.) The published results of 15 cohort studies all Population-based case-control studies differ showed a reduced risk of coronary heart disease in from hospital-based case-control studies in that estrogen users, although the results of one cohort the cases and controls come from the community study, the Framingham study, are equivocal. or a sample of the general population. Controls se- Most cohort studies followed women with and lected from the community rather than a hospital without estrogen exposure, and thus had a control are likely to be more representative of the general group internal to the study. In three studies, how- population from which the cases were drawn than ever, mortality in a cohort of estrogen users was hospital-based case-control studies. compared with national mortality rates. These co- In one of the largest population-based case- hort studies without internal controls showed the control studies of myocardial infarction and estro- lowest apparent relative risk of cardiovascular dis- gen use, Pfeffer and colleagues found among ease with estrogen use. (See table I-3.) But women current users of estrogens an adjusted relative risk who take estrogens are on average of higher socio- of 0.7 (0.3-1.4) for fatal and nonfatalMI(81 ). Es- economic status and more educated and therefore trogen use in this study was ascertained by review are probably healthier than the general population of pharmacy records. In a reanalysis of Pfeffer’s (19,103). Consequently, cohort studies without data, Ross found that estrogen use among cases internal controls may overestimate the effect of was underestimated, because one-third of the estrogen exposure on cardiovascular disease. women who had estrogen usage noted on their The findings of cohort studies with internal medical records did not have records of estrogen controls, including the Framingham study, are prescriptions in the pharmacy records (92). The summarized in table I-4. One of the largest cohort mean duration of use was less than three months, studies of cardiovascular disease risk among post- which would also bias the findings toward an un- menopausal estrogen users is the Lipid Research derestimate because such a short duration is un- Clinics Follow-up study, initiated by the National likely to be sufficient for a plausible biological Heart, Lung, and Blood Institute in 1971 (12,15). effect (103). Almost 2,300 women have been followed in this Unlike the other case control studies in this study. A 1987 report noted a statistically signifi- group which used myocardial infarction as an end- cant reduction in incidence of CHD or stroke point, Thompson and colleagues used a combined death among current estrogen users (average endpoint of stroke and myocardial infarction length of use 8.5 years) compared with nonusers. (111). In that study, each of 603 women with The relative risk of cardiovascular death in estro- stroke or myocardial infarction identified in 83 gen users was 0.34. Adjustment for other potential Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpoints Talbott (1 977) Cases were white female residents of 64 cases (unknown number Sudden death Current users:b Allegheny County, Pennsylvania who had no postmenopausal) age-adjusted relative risk prior recorded history of heart disease and (5% current users); 0.34 (0.09-1 .30) who were ages 25 to 64 years old (mean age 64 controls 55.6) when they died suddenly of atherosclerotic heart disease outside of the hospital between September 1973 and April 1975. Cases were identified from county coroner records and death certificates. Cases were matched to controls who were females living on the same block and who were within 10 years of patient’s age. Information about cases was gathered from interviews of subjects’ family and physicians. Information about controls was gathered from interviews of subjects. Pfeffer (1 978) Cases and controls were women ages 50 to 171 cases First Ml Ever users: 98 years who were residents of a Southern (30% ever users) risk-factor adjusted* relative risk California retirement community between (8.7% current users); 171 0.86 (0.54-1 .37) 1964 and 1974. Cases had their first Ml while controls current users: in residence. Controls were drawn from a file risk-factor adjusted* relative risk containing all women in residence during the 0.68 (0.32-1 .42) study interval. There were no black members of the population. Data was obtained from *adjusted for age, hypertension, and diabetes review of medical clinic and pharmacy records,

ROSS (1981) Cases were women less than 80 years old 133 cases (percent ever users Fatal coronary Ever users. living in a retirement community near Los not provided), 133 living heart age-adjusted relative risk Angeles who died of coronary heart disease controls; 133 deceased disease living controls: 0.43 (0.24-0.75) between 1971 and 1975 inclusive. For each controls dead controls 0.57 (0.33-0.99) case a living and deceased female control risk- factor adjusted* relative risk were selected, matched for race, age, date of living controls. unchanged entry into the community, and, for deceased dead controls. unchanged control, date of death. The deceased control was used to remove bias for extra medical *adjusted for multiple heart disease risk factors attention the cases may have had toward the end of their Iives Data was gathered from medical clinic records

II Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpoints Bain (1981) Cases were postmenopausal female nurses 120 cases (53% ever users) First Ml Ever users. ages 30 to 55 in 1976 who reported (27% current users), age-adjusted relative risk 0.9 (0.6-1.2) hospitalization for Ml. Twenty female nurses 2,400 controls risk-factor adjusted* relative risk 0.8 hospitalized in the same year with no history (0.6-1 .3) of Ml were matched as controls to each case current users. on the basis of year of birth and menopausal age-adjusted relative risk 0.7 (0,5- 1,1 ) status at hospitalization. Information was risk-factor adjusted* relative risk 0.7 gathered by questionnaire. (0.4-1.1) age-adjusted relative risk in women with natural menopause 1,3 (0.5-3.4) age-adjusted relative risk in women with hysterectomy 1.0 (0,5-2,2) age-adjusted relative risk in women with hysterectomy and bilateral oophorectomy 0.4 (0.2-0.8)

*adjusted for multiple heart disease risk factors Adam (1981 ) Cases were women ages 50 to 59 who died 76 cases Fatal Ml Ever users: of Ml in England and Wales during November (12% ever users) unadjusted relative risk b 1978 identified from death certificates. Two (3% current users); 0.65 (0.29-1 .45) controls matched by age to cases were 152 controls current users: randomly selected from the practice list of the unadjusted relative risk b general practitioner responsible for the care 0.97 (0.41 -2.28) of the patient during life. Information was gathered from hospital records, postmortem reports, and questionnaires completed by the subject’s general practitioner.

Croft (1989) A nested case-control was carried out on 158 cases (9 estrogen users), First Ml Ever users. cohort data collected during the Royal 474 controls (32 estrogen unadjusted relative risk 0.8 (no c.i. College of General Practitioners’ oral users) provided) contraceptive study. Subjects were recruited adjusted relative risk* 0.8 (0.3 to 1.8) by U.K. general practitioners, and were followed between May 1968 and July 1969. *adjusted for social class, smoking, use of oral The cases were all women who had had their contraceptives, history of pre-eclampsia, first AMI while under observation in the study. hypertension, and hysterectomy Controls were chosen from randomly selected general practice registers, matched for age to cases. Medical records were examined. Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpointsa Beard (1989) Cases were female residents of Rochester, 86 cases Ml or sudden Ever users: Minnesota between 1960 and 1982 whose (27% ever users); 150 controls death risk-factor adjusted* odds ratio 0 first manifestations of heart disease were 0.55 (0.24-1 .30) sudden death or Ml occurring between the ages of 40 and 59. Two controls matched by *adjusted for age, year, menopausal status, age to each case were selected from women smoking, hypertension, and diabetes seen at the Mayo Clinic. Information was obtained from review of medical records.

Thompson (1989) Cases were white women ages 45 to 69 who 603 cases Ml and stroke Ever users of estrogen alone, developed Ml or stroke between 1981 to (94% past users); age-adjusted relative risk 1986 and whose general practitioners 1,206 controls 1.12 (0.79-1 ,57) reported to Northwick Park Hospital, England. risk-factor adjusted* relative risk Controls were white female clinic patients 1,09 (0,65-1 ,82) matched for age and general practitioner. past users of estrogen alone: Information gathered from review of medical age-adjusted relative risk records and interviews. 0,86 (0.43-1 ,74) risk-factor adjusted* relative risk 1,16 (0.43-3.12) ever users of progestin alone. age-adjusted relative risk 1.90 (1,1 1 -3,25) risk-factor adjusted* relative risk 1.02 (0.45-2.32) ever users of combined estrogen -progestin: age-adjusted relative risk 0,86 (0,43-1 .74) risk-factor adjusted* relative risk 1.16 (0.43-3,12)

*adjusted for marital status, smoking, history of hypertension venous thrombosis, stroke, Ml, diabetes, and family history of Ml Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpoints Rosenberg (1993) Cases were women, ages 45 to 69 (median 858 cases First Ml Ever users. age 60 years), who were residents of (21% used unopposed risk-factor adjusted* relative risk 0.9 Massachusetts from 1986 until 1990. Controls estrogens, 3% used (0.7-1 .2) were women, matched by metropolitan estrogen and progestins); recent users: precinct and 5-year age group, with no prior 858 controls (21 % used risk-factor adjusted* relative risk history of Ml, Ninety-eight percent of cases unopposed estrogens, 3.5% 0.8 (0.4-1 .3) and 97 percent of controls were white. Data used estrogens and past users: were gathered from interviews of physicians progestins) risk-factor adjusted* relative risk and patients, 0.9 (0.7-1 .3) unopposed estrogen users: risk-factor adjusted relative risk 1.3 (0.4-4.9) estrogen and progestin users: risk-factor adjusted relative risk 1.2 (0.6-2.4) progestin only users: risk-factor adjusted relative risk 1.3 (0.4-4.9)

*adjusted for multiple heart disease risk factors The estimated relative risk decreased with increased duration of unopposed estrogen use to 06 (O 4-1 1) (test for trend p= 0.08), The association of decreased risk with duration of use was stronger with recent use (test for trend p< O 05) than for past use (test for trend P=O. 86) o

Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpoints Mann (1994) Cases and controls were women ages 45 to 1,521 cases Ml Ever users. 64 years who were included in the generaI (7.7% ever users); age-adjusted odds ratio practice files of the VAMP database of the 6,084 controls (9.2% ever 0.82 (0.67-1 .01 ) British National Health Service beginning in users) risk-factor adjusted* odds ratio June 1987 to May 1993. Cases comprised all 0.83 (0.66-1 .03) incidents of both fatal and non-fatal cases of age-adjusted odds ratio for Ml where there were records of HRT estrogen-progestin 0.68 (0.47-0.97) prescriptions within 6 months of the date of age-adjusted odds ratio for unopposed the Ml. Controls were females in the same estrogen 0.93 (0.47-1 .86) age group with no prior history of Ml. Four controls were matched to each case. Data *adjusted for history of smoking, diabetes, was gathered from computerized medical hypertension, hysterectomy, and hyperlipidemia records. a 95 percent confidence intervals are reported after risk estimates. b Figure obtained from reanalysis of data inoriginal paper, included in meta-analysis by M.J. Stampfer, and G. A. Colditz,“Estrogen Replacement Therapy and Coronary Heart Disease A Quantitative Assessment of the Epidemiological Evidence,” Preventive Medicine 20:47-63, 1991 KEY’ AM I = acute myocardial infarction; c.i. = confidence interval; HRT = hormonal replacement therapy; Ml = myocardial Infarction. SOURCE: Office of Technology Assessment, 1995 ......

Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpointsa,b Byrd (1977) Cohort included women mean age 44 years, 1,016 women (all estrogen Fatal CHD Ever users: who received hysterectomies from one users); 13 cases of fatal unadjusted relative riskc 0,37 Nashville, TN, gynecologist from the late CHD (p

Relationship of hormonal Measured replacement therapy to heart 0 Author Description of study Number of study subjects endpoint disease endpointsa’b Falkeborn (1992) Cohort comprised 23,174 women aged 35 23,174 women (all HRT users, First Ml Ever users. and older (median age 54 at study entry) 21% current users), 227 age-adjusted relative risk estradiol/ from the Uppsala Health Care Region of cases of first Ml conjugated estrogens 0.74 Sweden, who had been treated with (0.61 -0.88) estrogen/progestin. Subjects were identified other estrogens only 0.90 (0.74-1 .08) from pharmacy records as having been estrogen/progestin combination 0.50 prescribed non-contraceptive estrogens from (0.28-0.80) 1977 to 1983, Subjects were followed for an overall age-adjusted relative risk 0,81 average of 5.8 years. Cases of Ml within (0.71 -0.92) cohort were identified through a regional “The relative risk tended to decrease hospital inpatient registry. A subcohort of 735 with increased duration of followup, ” women were surveyed in 1980 and 1984 by from a relative risk of 0,96 mailed questionnaire to further characterize (0,44-1 .83) during the first year to a the cohort with respect to lifetime hormone relative risk of 0.76 (0.55-1 .02) exposure and the presence of other risk during the last (6 years later). factors. The incidence of first Ml in the cohort was compared with that in the general population.

KEY: CHD = coronary heart disease, IHD = ischemic heart disease, Ml = myocardial infarction a Estimates are of relative riskunless otherwise specified b 95% confidence intervals are provided in parentheses. C Figures obtained from reanalysis of data in text. Reanalysis of data was presented in M J Stampfer, and G A Colditz, “Estrogen Replacement Therapy and Coronary Heart Disease A Quantitative Assessment of the Epidemiological Evidence,” Preventive Medicine 2047-63, 1991 SOURCE Off Ice of Technology Assessment, 1995 Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpointsa

Hammond (1979) Study subjects were identified through Duke 610 women (49% estrogen CHD Ever users: University Medical Center’s (Durham, NC) users); 58 cases unadjusted odds ratiob inpatient medical record retrieval system and 0.33 (0.1 9-0.56) outpatient office records. All cohort subjects received diagnoses related to hypoestrogenism between 1940 and 1964, who returned for followup for five or more years after diagnosis, and had most recently been seen at Duke after January 1, 1974. Patients referred to Duke were excluded from the sample. Mean age of subjects was 46.3 years at baseline. Lafferty (1985) The cohort was recruited from 173 private 124 women (49% estrogen Ml Ever users: practice patients of the author for a users); 7 cases unadjusted odds ratiob prospective study between 1966-1981, 0.17 (0.03-1 .06) Candidates had been followed for not less than 3 years and had periodic physical exams and laboratory studies. The mean duration of followup was 1.6 years. The mean age of subjects was 53,7 years (range 45 to 60 years) at baseline.

Wilson (1985) Patients considered for inclusion were 1,234 women (14Y0 past users All CVD Ever users: members of the Framingham of estrogen, 1O% current relative risk for all CVD (Massachusetts) Heart Study cohort who users); 194 cases of CVD, 1,76 (p< O.01) adjusted for age and participated in the 12th biennial exam (index) 48 cases of CVD death, and HDL level; between 1970 and 1972 and who were 51 cases of Ml relative risk for CVD death postmenopausal and over 50 years of age at 1.94(p<0.05) adjusted for age and that exam. The cohort was followed for 8 HDL level; years. relative risk for Ml 1.87(p<0.05) adjusted— for age and HDL level Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpoints Eaker (1987) The author reanalyzed the data from the 1,297 women (14% past users, CHD except 50-59 years of age: cohort of the Framingham (Massachusetts) 10% current users) angina relative risk 0.26 (0.06-1 .22) adjusted Heart Study, described above (Wilson, 695 women ages 50 to 59; 35 for age and HDL level 1985).The subjects included women in the cases relative risk 0.4 (p< O.05)adjusted for Framingham Study cohort who were 50-59 602 women ages 60 to 69; 51 multiple risk-factors including HDL years of age or 60-69 years of age upon cases level exam 1 in 1950, exam 6 in 1960, or exam 11 50-69 years of age: in 1970. The cohort was divided on the basis relative risk 1.68 (0.71 -4.00) adjusted of the subject’s age at exam. Duration of for age and HDL level followup was 10 years. relative risk 2.2 (p< O.05) adjusted for multiple risk-factors, including HDL level Bush (1987a) The cohort consisted of 2,270 white women, 2,270 women (26% ever Fatal CVD Ever users: ages 40 to 69 at baseline, who were followed estrogen users); 50 cases age-adjusted relative risk for an average of 8.5 years. All women 0.34 (0.12-0.81) included in the study were participants in the risk-factor adjusted relative risk 0,37 Lipid Research Clinic (LRC) Prevalence Study (0.16-0.88) of CVD, that was conducted in 10 North American clinics between 1972 and 1976. Study was restricted to whites due to the small number of minorities in the LRC study.

Pettiti (1987) The cohort included women aged 18 to 54 6,093 women (44% ever users); Fatal CVD or Ever users: during December 1968 through February 40 cases of AM I fatal Ml age-adjusted relative risk 0.9 (0.2-3.3) 1972 who participated in the Walnut Creek for fatal CVD Contraceptive Drug Study who never used risk-factor adjusted relative risk 0.61 any type of estrogens or used estrogens for (0.3-1 .1) for fatal CVD reasons other than contraception. Duration of age-adjusted relative risk 0.3 (0.1 -1 .3) followup was 10 to 13 years. for fatal Ml

Criqui (1988) Study subjects were followed between 1972 1,868 women (39% ever users), Fatal CHD Ever users: to 1986 when they participated in a 87 cases age-adjusted relative risk 0.75 community survey of homogeneous, white, (0.45-1 24) upper-middle class residents of a planned, risk-factor adjusted relative risk 0.99 small Southern California retirement (0.59-1 .67) community (Rancho Bernardo). Women were 50 to 79 years of age at baseline. Average duration of followup was 12 years. Relationship of hormonal Measured replacement therapy to heart Author Description of study Number of study subjects endpoint disease endpoints

Avila (1 990) The study cohort comprised all female 24,900 women (14% current Ml Current users, members of the Group Health Cooperative of users), 120 cases age-adjusted relative risk 0.07 Puget Sound who were ages 50 to 64 years (0.4-1 .3) upon entry to cohort between 1978 to 1984. risk-factor adjusted relative risk 0,7 Cases were selected from women who were (0.4-1 .4) hospitalized and later discharged with a first occurrence of Ml. Average duration of followup was 5 years.

Henderson (1991) The cohort comprised female residents of 8,853 women (41% past users, Fatal AM I or Current users: Leisure World Retirement Community, Laguna 17.3% current users), 203 fatal IHD age-adjusted relative risk 0.601 Hills, California, who responded to a health cases (p< O.001 ) for fatal AMI questionnaire. The cohort was followed for age-adjusted relative risk of 0.79 7.5 years using death certificate records of (NS) for IHD the local health department. Female residents Duration (for a fatal AM I): were almost uniformly white, moderately <3 years: 0.64 (p<0.05) affluent, and well educated with a median 4-14 years: 0.60 (p<0.05) age of 73 years. >15 years: 0.52 (p15 years, 0,73 (NS) 2-14 years. 0.47 (p

KEY: AMI = acute myocardial infarction, CHD = coronary heart disease, CVD = cardiovascular disease, ERT = estrogen replacement therapy, HDL = high-density lipoprotein cholesterol, IHD = ischemic heart disease, Ml = myocardial infarction a Ninety-five percent confidence intervals are provided in parentheses, unless otherwise specified b Estimates of crude odds ratio derived from reanalysis of data in the text or from meta-analysis by Stampfer and Colditz in M J, Stampfer and G.A Colditz, “Estrogen Replacement Therapy and Coronary Heart Disease’ A Quantitative Assessment of the Epidemiological Evidence, ” Preventive Medicine 20:47-63, 1991 SOURCE: Office of Technology Assessment, 1995. 186 Cost Effectiveness of Screening for Osteoporosis

confounding factors (age, blood pressure, and begun in 1981 to investigate the risks and benefits smoking) did not substantially change the finding of menopausal ERT. A questionnaire was mailed of a reduction in risk of cardiovascular death to all 20,000 residents of this retirement commu- among current estrogen users. nity in 1981, and the two-thirds who completed On the basis of multivariate analysis of the re- and returned the questionnaire became members sults of this study, the investigators concluded that of the study cohort. About 9,000 of these were the beneficial effect of estrogens on cardiovascu- women (77). After 7.5 years of followup, current lar disease risk was substantially mediated users had an age-adjusted relative risk of fatal is- through HDL levels. When the multivariate anal- chemic heart disease of 0.46 compared with non- ysis included HDL, the benefit of cardiovascular users. Adjustment for several CHD risk factors disease mortality among estrogen users compared did not substantially change the results. Hender- with nonusers was reduced and no longer statisti- son also found that the overall mortality rate in cally significant, a finding consistent with the hy- those who had ever used estrogen was 20 percent pothesis that the protective effect of estrogens is lower than lifetime nonusers (95 percent confi- substantially mediated through increased HDL dence intervals 0.7 to 0.87) and the overall mortal- levels. Preliminary data from the 15-year follow ity in current users of estrogen with more than 15 up of patients in the study demonstrated a 65-per- years of estrogen use was 36 percent below that in cent reduction in cardiovascular disease and an nonusers (95 percent confidence intervals 0.51 to approximately 50-percent reduction in all-cause 0.82) (39). mortality (28). Even after adjusting for age, HDL, A cohort of 6,093 women ages 18 to 54 from and LDL, estrogen users continued to have a risk the Kaiser Permanence Medical Program was fol- of 60 percent that of nonusers. lowed for an average of 10 to 13 years (80). The A nationwide study of nurses also found post- mortality rate from heart disease and stroke was menopausal estrogens to have a protective effect slightly lower among estrogen users, with a rela- on major coronary disease (102). The Nurses’ tive risk of 0.9 (95 percent confidence intervals Health Study was established in 1976 with 0.2 to 3.3). After adjustment for a variety of car- 121,700 female nurses ages 30 to 55 years old. By diovascular risk factors, including age, hyperten- 1986,48,470 of these women were postmenopau- sion, obesity, and smoking, the apparent benefit sal. Participants who reported ever using estro- was more marked, with a relative risk of 0.6 but gens in the past had a statistically significant the reduction in risk remained statistically insig- relative risk for nonfatal and fatal coronary heart nificant (95 percent confidence intervals 0.3 to disease of 0.51 after an average of 7 years’ follow- 1.1). up. Adjustments for a variety of cardiac risk fac- In contrast to the other cohort studies, the Fra- tors including high cholesterol, family history of mingham Heart Study3 reported a 50 percent in- heart disease, hypertension, diabetes, obesity, and creased risk for all circulatory disorders in smoking did not substantially change these rela- postmenopausal estrogen users (120). An in- tive risk estimates. creased incidence of MI was observed among es- In an ongoing prospective study in a retirement trogen users, particularly those who smoked community near Los Angeles (Leisure World), cigarettes. Henderson and colleagues found that women who One criticism of the Framingham study’s con- used ERT had a relative risk of fatal acute MI of clusions with respect to postmenopausal ERT and 0.60 compared with nonusers (39). This study was

3 The Framingham Heart Study, named for the Boston suburb where residents have participated since 1948, began with 5,209 healthy men and women ages 30 to 62. In 1971, the ongoing study was expanded to include the offspring of the original participants. The effects on diet, medication, and life-style on health have been assessed every two years (biennial examinations). Appendix I Evidence on HRT and Coronary Heart Disease 187

cardiovascular disease is that a reduction in car- lem with some cohort studies is that estrogen use diovascular disease risk in estrogen users may was often ascertained only at the initiation of the have been obscured by including in the estrogen observation period and not reascertained at a later user group women whose use was remote; in the point in the study ( 103). By failing to update estro- Framingham study, anyone who had used estrogen use status, current and former users may be gen at some time in the eight years before the misclassified, and an underestimate of the effect twelfth biennial examination was counted as an of estrogen may result, particularly because the estrogen user (119). benefits of estrogen use are most pronounced Another criticism of the Framingham study is among current or recent users. that the investigators adjusted the results for HDL levels. Because estrogen’s beneficial effects are ❚ Cross-Sectional Studies thought to be substantially mediated through its effect on HDL level (15), the analysis may have Recently, a number of cross-sectional surveys of underestimated the cardiovascular benefits of estrogen use in women who have had coronary an- postmenopausal ERT. giography (heart catheterization) have been re- The Framingham study has also been criticized ported; these studies have found reduced for the use of subjective measures of cardiovascu- incidence of CHD in estrogen users (table I-5). lar disease (29). In the Framingham study, the rel- Angiographically demonstrated coronary artery ative risk of cardiovascular disease was estimated obstruction is thought to be a more specific end- using a number of endpoints including angina point for the presence of CHD than signs and pectoris (chest pain due to inadequate oxygen- symptoms such as angina pectoris or MI. Coro- ation of the heart), coronary heart disease, inter- nary angiography involves the injection of x-ray mittent claudication (symptom associated with opaque dye into each artery of the heart (78). X- atherosclerosis and other occlusive arterial dis- ray images of the heart (cineangiograms) are re- eases characterized by leg pain with walking and corded, and these images are then reviewed by the relieved by rest), transient ischemic attack (occlu- cardiologist for evidence of obstructions of the sive vascular disease symptom characterized by coronary arteries. brief periods of cerebral dysfunction, with no per- Cross-sectional studies are a subcategory of sistent necrologic deficit), myocardial infarction, case-control studies where the presence of disease congestive heart failure, coronary heart disease and the exposure to the agent are ascertained si- death, and sudden death. Chest pain can be due to multaneously (94). In the studies listed in table a wide variety of causes, some of which can be I-5, the presence of angiographically demon- mistakenly attributed to the presence of coronary strated coronary artery obstruction and the pa- heart disease. In a reanalysis of the Framingham tient’s history of estrogen exposure were data, excluding the nonspecific endpoint of angi- simultaneously ascertained. These studies have na pectoris, Eaker demonstrated a statistically found reduced disease in women who had taken nonsignificant reduction in risk of coronary heart estrogen. For example, Gruchow et al. reports on a disease among younger estrogen users (relative series of 933 women ages 50 to 75 years who un- risk 0.4,95 percent confidence interval 0.1 to 2.3) derwent coronary angiography (36). Estrogen us- and a statistically nonsignificant increase in risk ers were one-half as likely as nonusers to have of coronary heart disease among older estrogen moderate or severe occlusion of the coronary ar- users (relative risk 1.8, 95 percent confidence in- teries. In nonusers, the likelihood of occlusion interval 0.5 to 6.9). creased with age, whereas in users, no age trend Although prospective cohort studies have im- was evident. portant advantages over case-control studies in Hong et al. reported on a series of 90 consecu- avoiding bias from the subject recall of exposure tive women 55 years old or older undergoing diag- and the difficulties in selection of controls, a prob- nostic coronary angiography (41). Only 22 188 Cost Effectiveness of Screening for Osteoporosis

percent of estrogen users had significant obstruc- Estrogen replacement therapy, however, has tion of a major coronary artery (defined as 25 per- been found to prolong survival even in women cent or more luminal diameter narrowing), who are not “healthy” women who already have whereas 68 percent of nonusers had significant significant coronary artery disease. Recent stud- obstruction. ies have compared the later survival of estrogen users versus nonusers with previously docu- ❚ Randomized Clinical Trials mented coronary artery lesions demonstrated by In the only prospective randomized double-blind arteriography. Sullivan et al. recently found all- clinical trial of estrogen use and heart disease, cause mortality over a 10-year period to be lower Nachtigall and colleagues reported on the 10-year in women with coronary artery disease who used followup of eighty-four pairs of chronically ill ERT than in those who never used estrogen (1 07). women in a long-term-care hospital matched for They reported a retrospective analysis of postme- age and diagnosis, who were randomly assigned nopausal estrogen use, coronary artery obstruc- to take estrogen opposed with progestin (PERT) tion (stenosis), and survival in 2,268 women 55 or placebo (73). PERT users had a lower relative years or older who underwent coronary arteriogra- risk of myocardial infarction than nonusers, but phy in the past. They compared overall survival in there were only four myocardial infarctions in the estrogen users and nonusers who initially had var- study and the difference in risk was not statistical- ious degrees of coronary artery obstruction as ly significant. In a study this small, however, one demonstrated by arteriography. Over 10 years of would expect only very large differences in rela- followup, there was no difference in survival be- tive risk of disease to be capable of producing sta- tween estrogen users and nonusers with no initial tistically significant results. evidence of coronary artery obstruction on arterio- Some investigators have argued that much of graphy. But in those with initially mild to moder- the reported heart disease benefit of HRT may be ate coronary artery occlusion (less than 70 percent due to “healthy user” bias—the selection of rela- stenosis), 10-year survival was 85 percent in never tively healthy women with a lower risk of heart users versus 95.6 percent in ever users of estrogen. disease for HRT. These investigators argue that es- And in those who initially had severe occlusion trogen users are generally of higher social class (70 percent or greater stenosis), survival was 60 than nonusers (8), and social class is inversely percent among those who never used estrogen and associated with both heart disease and cancer 97 percent among those who had ever used estro- (51,65). They also argue that the lower heart dis- gen. One implication of these findings is that ERT ease incidence in ERT users maybe because doc- may have beneficial effects on the heart even tors were reluctant to prescribe estrogens to when started in older women with preexisting cor- women with coronary risk factors 10 years ago, onary heart disease. because, at that time, estrogen was contraindi- Barrett-Connor found that, even within a group cated in these women because earlier studies had of women from the same socioeconomic class, found increased risk of thrombosis and heart at- women taking estrogen were different from non- tack in young women taking oral contraceptives users with regard to health promotion and disease and in older men treated with estrogen. prevention measures (6). In order to minimize the One investigator showed that cohort studies bias introduced by differences in socioeconomic that found a reduction of heart disease incidence in status between estrogen users and nonusers, Bar- ERT users also showed a reduction in risk of total rett-Connor evaluated the estrogen use patterns of cancer incidence in ERT users, even though ERT 1,057 postmenopausal women from the same so- would not be expected to have a beneficial effect cioeconomically upper-middle-class community on total cancer incidence (83). in California (6). The women were categorized as Appendix I Evidence on HRT and Coronary Heart Disease 189

never users, past users, and current users. After an trogen’s effect on various intermediate endpoints average followup period of 4.4 years, 95 percent for coronary heart disease, such as blood lipid and of these women completed a mailed health survey lipoprotein levels. The first long-term large-scale questionnaire that asked about lifestyle and health controlled clinical trial of HRT using coronary care factors related to good health. In general, heart disease endpoints was begun in fall 1993 as women who never used estrogen were least likely part of the Women’s Health Initiative. This to have implemented healthy behavior changes, 15-year, $625 million study, sponsored by the Na- and were least likely to have had screening evalua- tional Institutes of Health, will examine the effect tions. Seventy percent of the group of current es- of HRT, as well as low fat diets, calcium supple- trogen users had had a mammogram in the last 12 ments, and vitamin D supplements on the inci- months, whereas 45 percent of the never users had dence of heart disease, osteoporosis, and other had one (p< 0.001). diseases. The study includes a clinical trial involv- Other investigators have also argued that users ing 57,000 women ages 50 to 79, and an observa- of ERT are relatively compliant, and that “comp- tional study involving 100,000 women from 45 liance bias” may account for some of the appar- medical centers across the United States. ent benefit of ERT on heart disease (79). To Randomized controlled clinical trials examin- examine the magnitude of “compliance bias,” ing the effect of estrogen on heart disease risk fac- analyses of data from two randomized clinical tors have shown evidence of heart disease benefits trials of drug treatments for heart disease have ex- in users of estrogen. The Postmenopausal Estro- amined total mortality in persons who complied gen/Progestin Interventions (PEPI) Trial involved with the taking of placebo (24,43). In these analy- 875 women, 45 to 64 years old at study entry, who ses, subjects who complied with the taking of a were randomly assigned to either estrogen, one of placebo had significantly lower overall mortality. three estrogen/progestin combinations, or place- The benefit of compliance with placebo was not bo (122). During this three-year multicenter trial, reduced by adjustment for a large number of vari- the women were monitored for changes in a num- ables, both medical and sociodemographic, that ber of heart disease risk factors, including blood might affect mortality. pressure LDL, HDL, and hemostatic factors. At The issue of selection bias will not be com- the end of the three year trial period, women tak- pletely resolved until completion of randomized ing estrogen alone had significant increases in controlled clinical trials of HRT and heart disease. HDL, decreases in LDL, and decreased fibrinogen A number of randomized controlled clinical trials levels changes consistent with a decreased risk of have been performed that have examined the ef- heart disease in estrogen users. fect of HRT on lipids and lipoproteins. In women, levels of high density lipoprotein (HDL) and triglycerides, and to a lesser extent, low density lipo- EVIDENCE ON PERT AND CHD proteins (LDL) predict cardiovascular death in The primary indication for adding progestins to women (9). These studies have demonstrated that the HRT regime is to reduce the risk of estrogen- ERT, and to a lesser extent, PERT, have induced induced irregular bleeding, endometrial hyperpla- favorable changes in lipids and lipoproteins, con- sia (abnormal overgrowth of the inner lining of the sistent with a reduced risk of heart disease in HRT uterus, or endometrium), and endometrial cancer users. (118). (See appendix G for more discussion.) But A controlled clinical trial that uses the endpoint an important unresolved issue is whether the of coronary heart disease symptoms or mortality benefits of PERT in protecting the endometrium would be expensive because of the large number are outweighed by the effect of progestins on the of study participants and the long duration of fol- risk of coronary artery disease. Studies of the rela- lowup that would be required (71). Therefore, tionship of HRT to coronary artery disease have many trials have been conducted that measure es- been largely limited to ERT. The effect of proges- 190 Cost Effectiveness of Screening for Osteoporosis

tin supplementation has not been extensively eva- mostatic factors, and fasting serum concentrations luated because the routine addition of progestins of glucose and insulin. Approximately 63 percent to prevent estrogen-induced endometrial carcino- of the women had never used HRT, 16 percent had ma has been recommended only recently (66). formerly used HRT, and 21 percent currently used Progestins are suspected to have an adverse im- HRT. Among current users of HRT, 83 percent pact on cardiovascular disease risk because pro- were using estrogen alone (primarily conjugated gestins have opposite effects on lipid and equine estrogens (CEE)), and 17 percent were us- lipoprotein metabolism from estrogens (84). Pro- ing PERT (primarily CEE with low dose medrox- gestins decrease HDL levels (40,76,99). Different yprogesterone acetate). types of progestins, however, vary in their impact The investigators found that, after adjusting for on lipids and lipoproteins, with the more andro- differences in other heart disease risk factors, cur- genic progestins, particularly those derived from rent users of estrogen had significantly increased the male hormone testosterone, having a greater levels of HDL and decreased levels of LDL than adverse impact. For example, Hirvonen et al. did nonusers (72). They also found no significant found that the progestins levonorgestrel and nore- difference in levels of HDL and LDL between us- thindrone in large doses (up to 10 mg) substan- ers of ERT and users of PERT. Users of ERT had tially reduced HDL, and the less androgenic significantly higher plasma triglyceride levels progestin medroxyprogesterone acetate (10 mg) than users of PERT. As elevated triglyceride lev- also reduced HDL levels, but not to as great an ex- els are thought to increase heart disease risk, users tent (40). Ottosson found that medroxyprogester- of estrogen alone had a somewhat poorer plasma one acetate lowered HDL-2 cholesterol level, lipid profile than users of estrogen with progestin, negating the increase observed with oral estrogens but both groups of current users had better lipid (76). Some evidence suggests that progestins may profiles than nonusers. Finally, current HRT users also adversely affect vessel-wall physiology (62). had significantly lower levels of lipoprotein(a) There is some evidence that lower doses of the than nonusers, with users of PERT having signifi- less androgenic progestins are sufficient to induce cantly lower levels of lipoprotein(a) than users of endometrial transformation and not substantially estrogen alone. Lipoprotein(a) concentrations attenuate estrogen’s beneficial effect on lipopro- may be inversely related to heart disease risk (98). teins (2,34,45,49, 100,113,116,1 17,123) and other Other changes were observed in the two groups of metabolic changes associated with heart disease current users that would be predicted to lower the risk (69,97). Progestin’s effect on lipoproteins ap- risk of coronary artery disease: a decline in fibrin- pears to be dose dependent, and lower doses of ogen levels (a serum protein involved in coagula- progestins may not substantially reduce estro- tion) and a decrease in glucose and insulin levels. gen’s beneficial effect on HDL (45). Users of ERT had higher levels of coagulation fac- Nabulsi and colleagues found in a cross-sec- tor VII and protein C than users of PERT and non- tional analysis of postmenopausal women that the users. This would suggest that PERT would have a addition of progestins did not attenuate estrogen’s better hemostatic profile than ERT. beneficial effects on heart disease risk factors; us- These findings confirm three other population- ers of estrogen with progestin actually had a better based studies in which HDL levels in women who profile of heart disease risk factors than users of received ERT were similar to those in women who estrogen alone (72). The investigators examined received PERT (8,32,1 14). heart disease risk factors among 4,958 postmeno- Recently, Falkeborn et al. reported the results pausal women, ages 45 to 64, from four regions of of a study of first MI among a cohort of 23,174 the United States, who were participating in the postmenopausal estrogen/progestin users com- Atherosclerosis Risk in Communities study. They pared with postmenopausal women in the com- examined the associations of HRT with blood munity (31). They found an age-adjusted relative pressure, concentrations of plasma lipids and he- Appendix I Evidence on HRT and Coronary Heart Disease 191

risk of first MI among current users of CEE (or es- derson et al. showed that women with a history of tradiol) with progestin of 0.74 (0.61 to 0.81). use showed a decrease in relative risk of fatal acute Results from the PEPI trial have also shown ev- MI and fatal ischemic heart disease with increased idence of heart disease benefits in users of PERT, duration of use (39). Rosenberg et al., in a case- although the benefits are not as great as those in control study, also found a significant trend to- users of ERT (122). At the end of the three year ward decreased risk of first MI with increased trial period, women taking estrogen plus a syn- duration of use of HRT, but only among current thetic progestin (medroxy progesterone) had a 2 users (90). milligram per deciliter (mg/dL) increase in HDL, Studies that examined dose failed to demon- whereas users of estrogen alone or estrogen plus a strate a decreased risk of coronary heart disease natural progestin (micronized progesterone, with greater doses (39,102). But Ross found a available in Europe) had about a 6 milligram per nonsignificant trend toward decreased risk with deciliter (mg/dL) increase in HDL, and the higher doses of conjugated equine estrogens (92). women assigned to the placebo group experienced Studies have not examined whether there are dif- no increase in HDL. Both the ERT group and the ferences in efficacy with different estrogen prepa- PERT group had significantly lower LDL than the rations. Further study is needed on whether dose, placebo group. Both treatment groups experi- length of use, and type of estrogen used affect risk. enced improvements in hemostatic factors and no There is evidence that HRT’s heart disease change in blood pressure compared with the pla- benefits will continue into women’s later years. cebo group. Epidemiologic studies have demonstrated HRT’s heart disease benefits in elderly women (15,39, CONCLUSIONS 88,107,108). The conclusion of authors of several recent re- OTA’s review of the evidence concurs with views of the evidence is that ERT reduces the risk those of other reviewers: there is both a theoretical of coronary heart disease (30,38,59). Both Stam- rationale and empirical evidence to support a re- pfer and Bush, in recent meta-analyses of the data, duced risk of heart disease in women who use es- concluded that the evidence strongly suggests that trogens. women taking estrogen therapy are at a risk for OTA chose a relative risk of 0.5 as the base case coronary heart disease about half that of nonusers estimate of heart disease risk in current users of es- (13,102). Several authors have found that the con- trogen. In formulating this estimate, OTA placed sistency of findings is stronger in the better de- greater emphasis on cohort studies than case-con- signed and analyzed studies (13,56,59,84,93, trol studies, because cohort studies are less prone 102). to bias. In general, cohort studies have demon- Several studies demonstrated that women who strated a greater heart protective effect of ERT currently use estrogen (current users) had a lower than case control studies. Among cohort studies, risk of coronary heart disease than women who 10 of 17 estimated relative risks of heart disease of had used them in the past (past users) (4,39, 0.5 or below, and 13 of 17 were consistent with the 64,81,88,90,91,102,111). Few data are available hypothesis that ERT reduces heart disease risk in about whether dose, length of use, and type of es- current users by half (confidence intervals in- trogen affect risk. One study that examined the ef- cluded 0.5).4 The major disadvantage of cohort fect of estrogen duration on CHD risk failed to studies without internal controls is that the “con- detect any effect of duration (102). However, Hen- trol” group may not be comparable to the clinic

4 0TA relied on the results of the Framingham cohort published by Eaker (29), because the results of the major paper on heart disease in the Framingham cohort did not report the crude or age-adjusted cardiovascular disease rates. o

Number Risk–factor of Type of Percentage of adjusted a Study Patient’s age patients estrogen use estrogen users Age-adjusted relative risk relative riska Sullivan, et al. (1988) Mean age 62.8 2,188 Current use 4.4% 0.44 (0.29-0.67) for 70+ percent 0.58 (0.35-0.97) occlusion vs. no steriosls b Gruchow, et al. (1988) Age range 50 to 75 933 Current use 15,5 0.59 (0.48-0.73) moderate vs. low occlusion score 0.37 (0.29-0.46) severe vs. low occlusion score C McFarland, et al. (1989) Age range 35 to 59 283 Ever use 41 0.5 (0.3-0.8) for 70+ percent 0.50 occlusion vs. no stenosis Hong (1992) Mean age 62.3 90 Current use 20 OR for coronary artery disease = 0.13 (p < 0.001) in estrogen users vs. nonusers.

KEY OR = odds ratio a 95 percent confidence intervals are given in parentheses b Value not provided. c Confidence interval not provided. Adapted from: M.J. Stampfer, and G.A. Colditz, “Estrogen Replacement Therapy and Coronary Heart Disease: A Quantitative Assessment of the Epidemiological Evidence,” Preventive Medicine 20:47-63, 1991.

II Appendix I Evidence on HRT and Coronary Heart Disease 193

population. Among cohort studies with internal users have, on average, estimated a lower risk of controls, seven of 12 reported reductions in heart heart disease than studies of ever users or past us- disease risk greater than 0.5 in ERT users. ers of ERT. To encompass the range of estimates OTA’s base case estimate of heart disease risk from these studies in our sensitivity analysis, OTA in ERT users is also consistent with all of the an- chose a relative risk of 0.2 as a best case estimate giographic studies, which show 50-to 60-percent of heart disease risk in ERT users, and a relative reductions in the amount of coronary artery steno- risk of 0.8 as a worst case estimate. sis in ERT users (table I-5). These studies of an- OTA assumed as a base case that users of PERT giographically defined coronary artery disease would have no heart disease benefit, and as a best should provide more precise estimates of heart case, that estrogen/progestin users would have a disease risk than studies using clinical endpoints 20-percent lower risk of heart disease than nonus- of heart attack or ischemic heart disease symp- ers (relative risk 0.8). Randomized clinical trials toms. This is because many postmenopausal examining estrogen with progestin’s effect on lip- women with significant coronary artery occlu- ids and lipoproteins suggest that the heart disease sions have no symptoms, and these women will be benefits of estrogen would be reduced when pro- misclassified as having no heart disease. This mis- gestins are added, although this reduction maybe classification diminishes the ability of an epide- minimized by using the lowest effective dose of miologic study of ERT users and nonusers to the least androgenic progestins. detect differences in risk of heart disease between OTA’s estimates of the relative risk of heart dis- the groups. ease in PERT users are consistent with recent epi- OTA’s base case estimate of heart disease risk demiologic studies. Because the addition of a in ERT users is consistent with that of the meta- progestin to ERT has become standard medical analyses by Barrett-Connor and Bush (7) (approx- practice only relatively recently, there are few epi- imately 50-percent reduction in risk of heart demiologic studies with sufficient numbers of es- disease in ERT users), Bush (13) (a reduction in trogen/progestin users to estimate its impact on risk of 40- to 50-percent), and Mack (60) (an esti- heart disease risk. mated 50-percent reduction in risk). This is also The evidence is weak to support a protective ef- consistent with the meta-analysis by Stampfer et fect extending beyond the period of use. In the ab- al. of cohort studies with internal controls and sence of such evidence, a reasonably conservative cross-sectional angiographic studies (102). Stam- assumption is that ERT (when not combined with pfer et al. obtained a somewhat higher estimate of progestins) reduces heart disease rates by one- heart disease in ERT users when the results of co- half, but only during the therapy period. Once hort studies without internal controls and case HRT ceases, heart disease rates can be assumed to control studies were also factored in to the esti- return quickly to the rates in the general popula- mate (102). OTA’s estimate of relative risk of heart tion of women of the same age. disease in ERT users was less than the meta-analysis of Grady et al., who calculated a relative risk of heart disease in ERT users of 0.65 (35). Because of the uncertainty about the magnitude REFERENCES of the heart protective effect of ERT, OTA tested 1. Adam, S., Williams, V., and Vessey, M., the sensitivity of the model to a wide range of esti- “Cardiovascular Disease and Hormone Re- mates of heart disease risk in ERT users. Although placement Treatment: A Pilot Case-Control most cohort studies have demonstrated a reduced Study,” British Medical Journal risk of heart disease in ERT users, the range of es- 282:1277-1278, 1981. timates of the relative risk varies widely, to as low 2. Adami, S., Rossini, M., Zamberlan, N., et as 0.17. In addition, cohort studies of current ERT al., “Long-Term Effects of Transdermal and 194 Cost Effectiveness of Screening for Osteoporosis

Oral Estrogen Serum Lipids and Lipopro- 12. Bush, T., “The Lipid Research Clinics Pro- teins in Postmenopausal Women,” Maturi- gram,” Postgraduate Medicine (Suppl.): tus 17(3):191-196, 1993. 45-48, April 1989. 3. Avila, M., Walker, A., and Jick, H., “Use of 13. Bush, T., “Noncontraceptive Estrogen Use Replacement Estrogens and the Risk of and Risk of Cardiovascular Disease: An Myocardial Infarction,” Epidemiology Overview and Critique of the Literature,” 1:128-133, 1990. The Menopause: Biological and Clinical 4. Bain, C., Willett, W., Hennekens, C., et al., Consequences of Ovarian Failure: Evolu- “Use of Postmenopausal Hormones and tion Management, S.G. Korenman (cd.) Risk of Myocardial Infarction,” Circulation (Norwell, MA: Serono Symposia, USA, 64:42-46, 1981. 1990). 5. Bar, J., Tepper, R., Fuchs, J., et al., “The Ef- 14. Bush, T., and Barrett-Connor, E., “Noncon- fect of Estrogen Replacement Therapy on traceptive Estrogen Use and Cardiovascular Platelet Aggregation and Adenosine Tri- Disease,” Epidemiologic Reviews 7:80- 104, phosphate Release in Postmenopausal 1985. Women,” Obstetrics & Gynecology 15. Bush, T., Barrett-Connor, E., Cowan, L., et 81(2):261-264, 1993. al., “Cardiovascular Mortality and Noncon- 6. Barrett-Connor, E., “Postmenopausal Es- traceptive Use of Estrogen in Women: Re- trogen and Prevention Bias,” Annals of In- sults from the Lipid Research Clinics ternal Medicine 115(6):455-456, 1991. Program Follow-Up Study,” Circulation 7. Barrett-Connor, E., Bush, T. L., “Estrogen 75(6):1102-1109, 1987a. and Coronary Heart Disease in Women,” 16. Bush, T., and Miller, V., “Effects of Phar- Journal of the American Medical Associa- macologic Agents Used During Menopause: tion 265(14): 1861-1867, 1991. Impact on Lipids and Lipoproteins,” Meno- 8. Barrett-Connor, E., Wingard, D., and Cri- pause: Physiology and Pharmacology, D.R. qui, M., “Postmenopausal Estrogen Use and Mishell, Jr., (cd.) (Chicago, IL: Year Book, Heart Disease Risk Factors in the 1980s: 1987). Rancho Bernardo, California Revisited,” 17. Byrd, B., Burch, J., and Vaughn, W., “Im- Journal of the American Medical Associa- pact of Long Term Estrogen Support After tion 261:2095-2100, 1989. Hysterectomy,” Annals of Surgery 9. Bass, K., Newschaffer, C., Klag, M., et al., 185(5):574-580, 1977. “Plasma Lipoprotein Levels as Predictors of 18. Caruso, M.G., Berloco, P., Notarnicola, M., Cardiovascular Death in Women,” Archives et al., “Lipoprotein (a) Serum Levels in of Internal Medicine 153:2209-2216, 1993. Post-Menopausal Women Treated with Oral 10. Bauwens, S., “Transdermal Versus Oral Es- Estrogens Administered at Different trogen for Postmenopausal Replacement Times,” Hormonal Metabolism Research Therapy,” Clinical Pharmacology 26(8) ;379-382, 1994. 8:364-366, 1989. 19. Cauley, J. A., Cummings, S. R., Black, 11. Beard, C., Kottke, T., Annegers, J., and Bal- D. M., et al., “Prevalence and Determinants lard, D., “Rochester Coronary Heart Disease of Estrogen Replacement Therapy in Elderly Project: Effect of Cigarette Smoking, Hy- Women,” American Journal of Obstetrics pertension, Diabetes and Steriodal Estrogen and Gynecology 163: 1438-1444, 1990. Use on Coronary Heart Disease Among 40- 20. Chetkowski, R., Meldrum, D., Steingold, to 59-Year-Old Women, 1960 Through K., et al., “Biologic Effects of Transdermal 1982,” Mayo Clinic Proceedings Estradiol,” New England Journal of Medi- 64:1471-1480, 1989. cine 314: 1615-1620, 1986. Appendix I Evidence on HRT and Coronary Heart Disease 195

21. Colditz, G., Willett, W., Stampfer, W., et al., 30. Ernster, V., Bush, T., Huggins, G., et al., “Menopause and the Risk of Coronary Heart “Clinical Perspectives: Benefits and Risks Disease in Women,” New England Journal of Menopausal Estrogen and/or Progestin of Medicine 316: 1105-1110, 1987. Hormone Use,” Preventive Medicine 22. Collins, P., Rosano, G., Jiang, C., et al., 17:201-223, 1988. “Cardiovascular Protection by Oestrogen— 31. Falkeborn, M., Persson, I., Adami, H. O., et A Calcium Antagonist Effect?” Lancet 341: al., “The Risk of Acute Myocardial Infarc- 1264-1265, 1993. tion After Oestrogen and Oestrogen-Proges- 23. Colvin, P., Auerbach, B., Applebaum-Bow- ton Replacement,” British Journal of den, D., et al., “Effect of Estrone Versus 17 Obstetrics and Gynecology Beta-Estradiol on Lipoproteins in Post- 99(10):821-828, 1992. Menopausal Women,” Clinical Research 32. Gambrell, R. D., and Teran, A., “Changes in 36:269A, 1988. Lipids and Lipoproteins with Long-Term 24. Coronary Drug Project, “Influence of Ad- Estrogen Deficiency and Hormone Replace- herence to Treatment and Response of Cho- ment Therapy,” American Journal of Ob- lesterol on Mortality in the Coronary Drug stetrics and Gynecology 165(2):307-317, Project,” New England Journal of Medicine 1991. 303:1038-1041, 1980. 33. Ganger, K.F., Reid, B. A., Crook, D., et al., 25. Creager, M.A., “Estrogen Improves Endo- “Estrogens and Atherosclerotic Disease— thelium-Dependent, Flow-Mediated Vaso- Local Vascular Factors,” Bailliere’s Clinical dilation in Postmenopausal Women,” Endocrinology and Metabolism 7:47-60. Annals of Internal Medicine 121(12): 1993. 936-941, 1994 34. Gibbons, W., Moyer, D., and Lobo, R., 26. Criqui, M., Suarez, L., Barrett-Connor, E., “Biochemical and Histologic Effects of Se- et al., “Postmenopausal Estrogen Use and quential Estrogen/Progestin Therapy on the Mortality, Results from a Prospective Study Endometrium of Postmenopausal Women,” in a Defined, Homogeneous Community,” American Journal of Obstetrics and Gy- American Journal of Epidemiology 128: necology 154:456-461, 1986. 606-614, 1988. 35. Grady, D., Rubin, S. M., Petitti, S. B., et al., 27. Croft, P., and Hannaford, P. C., “Risk Fac- “Hormone Therapy to Prevent Disease and tors for Acute Myocardial Infarction in Prolong Life in Postmenopausal Women,” Women: Evidence from the Royal College Annals of Internal Medicine 117(12): of General Practitioners Oral Contraception 1016-1037, 1992. Study,” British Medical Journal 298: 36. Gruchow, H.W., Anderson, A.J., Barboriak. 165-168, 1989. J.J., et al., “Postmenopausal Use of Estrogen 28. Drug Research Reports, “Cardiovascular, and Occlusion of Coronary Arteries,” Amer- Coronary Heart Disease Mortality Down ican Heart Journal 115(5):954-963, 1988. 65% in Women Taking Estrogen, LRC Data 37. Hammond, C., Jelovsek, F., Lee, L., et al., Shows,” The Blue Sheet p. 2-3, Jan. 16, “Effects of Long-Term Estrogen Replace- 1991. ment Therapy: I. Metabolic Effects,” Ameri- 29. Eaker, E., and Castelli, W., “Coronary Heart can Journal of Obstetrics and Gynecology Disease and Its Risk Factors Among Women 133(5):525-536, 1979. in the Framingham Study,” Coronary Heart 38. Hazzard, M. D., “Estrogen Replacement and Disease in Women (New York, NY: Haymar- Cardiovascular Disease: Serum Lipids and ket Doyma Inc., 1987). Blood Pressure Effects,” American Journal 196 Cost Effectiveness of Screening for Osteoporosis

of Obstetrics and Gynecology 161: 47. Jick, H., Dinan, B., Herman, R., et al., 1847-1853, 1989. “Myocardial Infarction and Other Vascular 39. Henderson, B. E., Paganini-Hill, A., and Diseases in Young Women: Role of Estro- Ross, R. K., “Decreased Mortality in Users gens and Other Factors,” Journal of the of Estrogen Replacement Therapy” (com- American Medical Association 240(23): ment), Archives of Internal Medicine 2548-2552, 1978. 151:75-78, 1991. 48. Jick, H., Dinan, B., and Rothman, K. J., 40. Hirvonen, E., Malkonen, M., and Man- “Noncontraceptive Estrogens and Nonfatal ninen, V., “Effects of Different Progestogens Myocardial Infarction,” Journal of the on Lipoproteins During Postmenopausal American Medical Association 239(14): Replacement Therapy,” New England Jour- 1407-1408, 1978a. nal of Medicine 304:560-563, 1981. 49. Kable, W. T., Gallagher, J. C., Nachtigall, L., 41. Hong, M., Romm, P., Reagan, K., et al., “Ef- et al., “Lipid Changes After Hormone Re- fects of Estrogen Replacement Therapy on placement Therapy for Menopause,” Jour- Serum Lipid Values and Angiographically nal of Reproductive Medicine 35(5): Defined Coronary Artery Disease in Post- 512-518, 1990. menopausal Women,” American Journal of 50. Kannel, W., Hjortland, M., McNamara, P., Cardiology 69(3):176-178, 1992. et al., “Menopause and Risk of Cardiovascu- 42. Horwitz, R. I., and Feinstein, A. R., “Alter- lar Disease: The Framingham Study,” An- native Analytic Methods for Case-Control nals of Internal Medicine 85:447-452, 1976. Studies of Estrogens and Endometrial Can- 51. Kaplan, G., and Keil, J., “Socioeconomic cer,” New England Journal of Medicine Factors and Cardiovascular Disease: A Re- 299:1089-1094, 1978. view of the Literature,” Circulation 43. Horwitz, R. I., Viscoli, C. M., Berkman, L., 88:1973-1995, 1993. et al., “Treatment Adherence and Risk of 52. Karas, R. H., Patterson, B. L., Mendelssohn Death After a Myocardial Infarction,” Lan- M.E., “Human Vascular Smooth Muscle cet 336:542-545, 1990. Cells Contain Functional Estrogen Recep- 44. Hunt, K., Vessey, M., McPherson, K., et al., tor,” Circulation 89(5): 1943-1950, 1994. “Mortality in a Cohort of Long-Term Users 53. Knopp, R., “Cardiovascular Effects of En- of Hormone Replacement Therapy: An Up- dogenous and Exogenous Sex Hormones dated Analysis,” British Journal of Obstet- Over a Woman’s Lifetime,” American Jour- rics and Gynecology 97: 1080-1086, 1990. nal of Obstetrics and Gynecology 158: 45. Jensen, J., Nilas, L., and Christiansen, C., 1630-1643, 1988. “Cyclic Changes in Serum Cholesterol and 54. La Vecchia, C., Franceschi, S., Decarli, A., Lipoproteins Following Different Doses of et al., “Risk Factors for Myocardial Infarc- Combined Postmenopausal Hormone Re- tion in Young Women,” American Journal placement Therapy,” British Journal of Ob- of Epidemiology 125(5):832-843, 1987. stetrics and Gynecology 93:613-618, 55. Lafferty, F., and Helmuth, D., “Post-Meno- 1986. pausal Estrogen Replacement: The Preven- 46. Jensen, J., Riis, B., Strom, V., et al., “Long- tion of Osteoporosis and Systemic Effects,” Term Effects of Percutaneous Estrogens and Maturitas 7: 147-159, 1985. Oral Progesterone on Serum Lipoproteins in 56. Langer, R. D., and Barrett-Connor, E., “Epi- Postmenopausal Women,” American Jour- demiology and Prevention of Cardiovascu- nal of Obstetrics and Gynecology 156: lar Disease in Women,” Contemporary 66-71, 1987. Internal Medicine, Clinical Case Studies— Appendix I Evidence on HRT and Coronary Heart Disease 197

Volume 3, J.M. Bowen (cd.) (New York, NY: 66. Martin, K.A., and Freeman, M.W., “Post- Plenum Medical Books, 1991). menopausal Hormone-Replacement Thera- 57. LaRosa, J., “Effect of Estrogen Replacement py” (editorial comment), New England Therapy on Lipids: Implications for Cardio- Journal of Medicine 328(15):1115-1117, vascular Risk,” Journal of Reproductive 1993. Medicine 35(Suppl.):81 1-813, 1985. 67. McFarland, K., Boniface, M., Hornung, C., 58. Lieberman, E., Gerhard, M., Uehata, A., et et al., “Risk Factors and Noncontraceptive al., “Estrogen Improves Endothelium-De- Estrogen Use in Women With and Without pendent, Flow-Mediated Vasodilation in Coronary Disease,” American Heart Jour- Postmenopausal Women, ’’Archives of lnter- nal 117(6): 1209-1214, 1989. nal Medicine 121:936-941, 1994. 68. McGill, H., Jr., “Sex Steroid Hormone Re- 59. Lobo, R.A., “Cardiovascular Implications ceptors in the Cardiovascular System,” of Estrogen Replacement Therapy,” Obstet- Postgraduate Medicine :64-68, April 1989. rics & Gynecology 75( Suppl. 4): 18S-25S; 69. Mendoza, S., Velazquez, E., Osona, A., et discussion 31S-35S, 1990. al., “Postmenopausal Cyslic Estrogen— 60. Mack, T., and Ross, R., “A Current Percep- Progestin Therapy Lowers Lipoprotein(a),” tion of HRT Risks and Benefits,” Osteopo- Journal of Laboratory and Clinical Medi- rosis: Physiologic Basis, Assessment, and cine 123(6):837-841, 1994 Treatment, H. Deluca and R. Mazess (eds.) 70. Mileikowsky, G., Nadler, J., Huey, F., et al., (New York, NY: Elsevier Science Publish- “Evidence that Smoking Alters Prostacyclin ing Co., Inc., 1990). Formation and Platelet Aggregation in 61. MacMahon, B., “Cardiovascular Disease Women Who Use Oral Contraceptives,” and Noncontraceptive Oestrogen Therapy,” American Journal of Obstetrics and Gy- Coronary Heart Disease in Young Women, necology 159(6): 1547-1552, 1988. M.F. Oliver (cd.) (New York, NY: Churchill 71. Moon, T., “Estrogen and Disease Preven- Livingstone, 1978). tion” (editorial), Archives of Internal Med- 62. Makila, U., Wahlberg, L., Vlinikka, L., et icine 151: 17-18, 1991. al., “Regulation of Prostacyclin and Throm- 72. Nabulsi, A. A., Folsom, A. R., White, A., et boxane by Human Umbilical Vessels: The al., “Association of Hormone-Replacement Effect of Estradiol and Progesterone in a Su- Therapy with Various Cardiovascular Risk perfusion Model,” Prostaglandins Leuko- Factors in Postmenopausal Women,” New trienes and Medicine 8: 115-124, 1982. England Journal of Medicine 328(15): 63. Mann, R. D., Lis, Y., Chukwujindu, J., et al., 1069-1075, 1993. “A Study of the Association Between Hor- 73. Nachtigall, L. E., Nachtigall, R. H., Nachti- mone Replacement Therapy, Smoking and gall, R. D., et al., “Estrogen Replacement the Occurrence of Myocardial Infarction in Therapy II: A Prospective Study in the Rela- Women,” Journal of Clinical Epidemiology tionship to Carcinoma and Cardiovascular 47(3):307-312, 1994. and Metabolic Problems,” Obstetrics & Gy- 64. Manolio, T.A., Furberg, C. D., Shemanski, necology 54(1):74-79, 1979. L., et al., “Associations of Postmenopausal 74. Notelovitz, M., “The Role of the Gynecolo- Estrogen Use with Cardiovascular Disease gist in Osteoporosis Prevention: A Clinical and Its Risk Factor in Older Women,” Cir- Approach,” Clinical Obstetrics and Gy- culation 88(5, part 1):2163-2171, 1993. necology 30(4):871-884, 1987. 65. Marmot, M., and McDowell, M., ’’Mortality 75. Ottosson, U., “Oral Progesterone and Estro- Decline and Widening Social Inequalities,” gen/Progestogen Therapy: Effects of Natu- Lancet ii:274-276, 1986. ral and Synthetic Hormones on Subfractions 198 Cost Effectiveness of Screening for Osteoporosis

of HDL Cholesterol and Liver Proteins,” 84. Psaty, B., Heckbert, S., Atkins, D., et al., “A Acta Obstetrician et Gynecological Scandi- Review of the Association of Estrogens and navia 127 (Suppl.): 1-37, 1984. Progestins with Cardiovascular Disease in 76. Ottosson, U., Johansson, B., and von Postmenopausal Women, ’’Archives of lnter- Schoultz, B., “Subfractions of High-Density nal Medicine 153:1421-1427, 1993. Lipoprotein Cholesterol During Estrogen 85. Riedel, M., Raffenbeul, W., and Lichtlen, P., Replacement Therapy: A Comparison Be- “Ovarian Sex Steroids and Atherosclerosis,” tween Progestins and Natural Progester- Clinical Investigation 71:406-412, 1993. one,” American Journal of Obstetrics and 86. Riis, B., Johansen, J., and Christiansen, C., Gynecology 151:746-750, 1985. “Continuous Oestrogen-Progestogen Treat- 77. Paganini-Hill, A., Ross, R. K., Henderson, ment and Bone Metabolism in Post-Meno- B.E., et al., “Endometrial Cancer and Pat- pausal Women,“ Maturitas 10:51-88, 1988. terns of Use of Oestrogen Replacement 87. Rosano, G. M., Sarrel, P. M., Poole-Wilson, Therapy: A Cohort Study,” British Journal P. A., et al., “Beneficial Effect of Oestrogen of Cancer 59:445-447, 1989. on Exercise-Induced Myocardial Ischaemia 78. Peterson, K., and Ross, J., Jr., “Cardiac Ca- in Women with Coronary Artery Disease,” theterization and Angiography,” Harrison’s Lancet 342:133-136, 1993. Principles of Internal Medicine, 1lth Ed., E. 88. Rosenberg, L., Armstrong, B., and Jick, H., Braunwald, K.J. Isselbacher, R.G. Peters- “Myocardial Infarction and Estrogen Thera- dorf, et al., (eds.) (New York, NY: Little py in Post-Menopausal Women,” New Eng- Brown, 1987). land Journal of Medicine 294:1256-1259, 79. Petitti, D. B., “Coronary Heart Disease and 1976. Estrogen Replacement Therapy: Can Com- 89. Rosenberg, L., Hennekens, C., Rosner, B., pliance Bias Explain the Results of Observa- et al., “Early Menopause and the Risk of tional Studies?” Annals of Epidemiology Myocardial Infarction,” American Journal 4(2): 115-118, 1994. of Obstetrics and Gynecology 139:47-51, 80. Petitti, D. B., Perlman, J. A., and Sidney, S., 1981. “Noncontraceptive Estrogens and Mortal- 90. Rosenberg, L., Palmer, J. R., and Shapiro, ity: Long-Term Follow-Up of Women in the S., “A Case-Control Study of Myocardial Walnut Creek Study,” Obstetrics & Gy- Infarction in Relation to Use of Estrogen necology 70:289-293, 1987. Supplements,” American Journal of Epide- 81. Pfeffer, R.I., Whipple, G. H., Kurosaki, T.T., miology 137(1):54-63, 1993. et al., “Coronary Risk and Estrogen Use in 91. Rosenberg, L., Slone, D., Shapiro, S., et al., Postmenopausal Women,” American Jour- “Noncontraceptive Estrogens and Myocar- nal of Epidemiology 107(6):479-497, 1978. dial Infarction in Young Women,” Journal 82. Pitt, B., Shea, M. J., Romson, J.L., et al., of the American Medical Association “Prostaglandins and Prostaglandin Inhibi- 244(4):339-342, 1980. tors in Ischemic Heart Disease,” Annals of 92. Ross, R., Mack, T., Paganini-Hill, A., et al., Internal Medicine 99:83-92, 1983. “Menopausal Oestrogen Therapy and 83. Posthuma, W., Westendorp, R., Vanden- Protection from Death from Ischemic Heart broucke, J., “Cardioprotective Effect of Disease,” Lancet :858-860, 1981. Hormone Replacement Therapy in Post- 93. Ross, R., Paganini-Hill, A., Mack, T., et al., menopausal Women: Is the Evidence “Estrogen Use and Cardiovascular Dis- Biased?” British Medical Journal 308: ease,” Menopause, Physiology and Phar- 1268-1269, 1994. Appendix I Evidence on HRT and Coronary Heart Disease 199

macology, D.R. Mishell (cd.) (Chicago, IL: of the Epidemiologic Evidence,” Preventive Year Book, 1987). Medicine 20:47-63, 1991. 94. Rothman, K., Modern Epidemiology (Bos- 103. Stampfer, M.J., Willett, W.C., Colditz, ton, MA: Little, Brown and Co., 1986). G. A., et al., “Past Use of Oral Contracep- 95. Sacks, F., McPherson, R., and Walsh, B. W., tives and Cardiovascular Disease: A Meta- “Effect of Postmenopausal Estrogen Re- Analysis in the Context of the Nurses’ placement on Plasma Lp(a) Lipoprotein Health Study,” American Journal of Obstet- Concentrations,” Archives of Internal Medi- rics and Gynecology 163(1 Pt 2):285-291, cine 154(10): 1106-1110, 1994. 1990. 96. Sarrell, P., Lindsay, D., Rosano, G., et al., 104. Stampfer, M.J., Willett, W. C., Colditz, “Angina and Normal Coronary Arteries in G. A., et al., “A Prospective Study of Post- Women: Gynecologic Findings,” American menopausal Estrogen Therapy and Coro- Journal of Obstetrics and Gynecology nary Heart Disease,” New England Journal 167:467-472, 1992. of Medicine 313(17): 1044-1049, 1985. 97. Scarabin, P. Y., Plu-Bureau, G., Bara, L., et 105. Stanczyk, F., Shoupe, D., Nunez, V., et al., al., “Hemostatic Variables and Menopausal “A Randomized Comparison of Nonoral Status: Influence of Hormone Replacement Estradiol Delivery in Postmenopausal Therapy,” Thrombosis and Haemostasis Women,” American Journal of Obstetrics 70(4):584-587, 1993. and Gynecology 159(6): 1540-1546, 1988. 98. Schaefer, E.J., Lamon-Fava, S., Jenner, J.L., 106. Stevenson, J., Crook, D., Godsland, I., et al., et al., “Lipoprotein[a] Levels and Risk of “Hormone Replacement Therapy and the Coronary Heart Disease in Men. The Lipid Cardiovascular System: Nonlipid Effects,” Research Clinics Coronary Primary Preven- Drugs 47(Suppl. 2):35-41, 1994. tion Trial,” Journal of the American Medical 107. Sullivan, J., Vander Zwaag, R., Hughes, J., Association 271(13):999-1003, 1994. et al., “Estrogen Replacement and Coronary 99. Silferstolpe, G., Gustafson, A., Samsioe, Artery Disease. Effect on Survival in Post- G., et al., “Lipid Metabolic Studies in Oo- menopausal Women,” Archives of Internal phorectomized Women: Effect of Three Dif- Medicine 150:2557-2562, 1990. ferent Progestins,” Acta Obstetrician et 108. Sullivan, J., Vander Zwaag, R., Lemp, G., et Gynecological Scandinavia 88(suppl.): al., “Postmenopausal Estrogen Use and Cor- 89-95, 1979. onary Atherosclerosis,” Annals of Internal 100. Soma, M., Osnago-Gadda, I., Paoletti, R., et Medicine 108(3):358-363, 1988. al., “The Lowering of Lipoprotein[a] In- 109. Szklo, M., Tonascia, J., Gordis, L., et al., duced by Estrogen Plus Progesterone Re- “Estrogen Use and Myocardial Infarction placement Therapy in Postmenopausal Risk: A Case Control Study,” Preventive Women,” Archives of Internal Medicine Medicine 13:510-516, 1984. 153(12): 1462-1468, 1993. 110. Talbott, E., Kuller, L., Detre, K., et al., “Bio- 101. Sporrong, T., Hellgren, M., Samsioe, G., et logic and Psychosocial Risk Factors of Sud- al., “Comparison of Four Continuously Ad- den Death from Coronary Disease in White ministered Progestogen Plus Estradiol Women,” American Journal of Cardiology Combinations for Climacteric Complaints,” 39:858-864, 1977. British Journal of Obstetrics and Gynaecol- 111. Thompson, S., Meade, T., and Greenberg, ogy 95: 1042-1048, 1988. G., “Use of Hormonal Replacement Therapy 102. Stampfer, M.J., and Colditz, G. A., “Estro- and the Risk of Stroke and Myocardial In- gen Replacement Therapy and Coronary farction in Women, ’’Journal of Epidemiolo- Heart Disease: A Quantitative Assessment 200 Cost Effectiveness of Screening for Osteoporosis

gy and Community Health 43:173-178, 9S-75S, 1990. 1989. 119. Wilson, P., “Prospective Studies: The Fra- 112. Tikkanen, M.J., “Mechanisms of Cardio- mingham Study, ” Postgraduate Medicine vascular Protection by Postmenopausal (Suppl.):51-53, April 1989. Hormone Replacement Therapy,” Cardio- 120. Wilson, P.W.F., Garrison, R.J., and Castelli, vascular Risk Factors 3:138-143, 1993. W.P., “Postmenopausal Estrogen Use, Ciga- 113. Van der Mooren, M.J., Leuven, J. A., Rol- rette Smoking, and Cardiovascular Morbid- land, R., “Effect of Conjugated Estrogens ity in Women Over 50: The Framingham With and Without Medrogestone: A Pro- Study,” New England Journal of Medicine spective Study,” Maturitas 19(1):33-42, 313(17):1038-1043, 1985. 1994. 121. Wolf, P. H., Madans, J. H., Finucane, F. F., et 114. Vaziri, S., Evans, J., Larson, M., et al., “The al., “Reduction of Cardiovascular Disease- Impact of Female Hormone Usage on the Related Mortality Among Postmenopausal Lipid Profile: The Framingham Offspring Women Who Use Hormones: Evidence Study,” Archives of Internal Medicine from a National Cohort,” American Journal 153:2200-2206, 1993. of Obstetrics and Gynecology 164(2): 115. Walsh, B.W., Schiff, I., Rosner, B., et al., 489-494, 1991. “Effects of Postmenopausal Estrogen Re- 122. Writing Group for the PEPI Trial, “Effects placement on the Concentrations and Me- of Estrogen or Estrogen/Progestin Regi- tabolism of Plasma Lipoproteins,” New mens on Heart Disease Risk Factors in England Journal of Medicine 325(17): Postmenopausal Estrogen/Progestin Inter- 1196-1204, 1991. ventions (PEPI) Trial,” Journal of the Amer- 116. Webber, C.E., Blake, J. M., Chambers, L.F., ican Medical Association 273(3): 199-208, et al., “Effects of 2 Years of Hormone Re- 1995. placement on Bone Mass, Serum Lipids, and 123. Yancey, M.K., Hannan, C.J., Plymate, S.R., Lipoproteins,” Maturitas 19(1):13-23, et al., “Serum Lipids and Lipoproteins in 1994. Continuous or Cyclic Medroxy Progester- 117. Weinstein, L., Bewtra, C., and Gallagher, one Acetate Treatment in Postmenopausal J. C., “Evaluation of a Continuous Com- Women Treated with Conjugated Estrogen,” bined Low-Dose Regimen of Estrogen-Pro- Fertility and Sterility 54(5):778-772, 1990. gestin for Treatment of the Menopausal 124. Ylikorkala, O., Puolakka, J., and Viinikka, Patient,” American Journal of Obstetrics L., “Vasoconstrictory Thromboxane A2 and and Gynecology 162: 1534-1542, 1990. Vasodilatory Prostacyclin in Climacteric 118. Whitehead, M. I., Hillard, T. C., and Crook, Women: Effect of Oestrogen-Progestogen D., “The Role and Use of Progestogens,” Therapy,” Maturitas 5:201-205, 1984. Obstetrics & Gynecology 75( Suppl. 4): Appendix J: Methods for Estimating costs J

his appendix describes the methods and rently covered by Medicare (18). It is also the sources used in estimating cost parameters method used in OTA’s model to estimate the dis- for the OTA model. The components of tribution of BMD levels at various ages. Other T cost required for the model include those BMD measurement technologies are available, of bone mineral density (BMD) measurement, generally at higher cost than SPA. The estimate of hormone replacement therapy (HRT), heart dis- $100 used for this analysis is consistent with esti- ease, hip fractures, gall bladder disease, endome- mates from a variety of sources. Table J- 1 summa- trial cancer, and breast cancer. rizes the cost estimates used in various studies of The OTA model considers only those health BMD screening. care costs that are directly attributable to the specific conditions whose incidence or severity are HRT COST affected by osteoporosis prevention strategies. They do not include any health care costs unre- The components of the treatment costs are sum- lated to these conditions. When an osteoporosis marized in table J-2. HRT regimens vary, some- prevention strategy increases life spans, people times including estrogen (ERT) alone and are likely to use more health care simply because sometimes estrogen in combination with (or fol- they are living longer. These increases in “unre- lowed by) progestin (PERT). lated costs” are not included in OTA’s analysis, be- ❚ cause they are assumed to be part of the portfolio Estrogen-Only Therapy of both costs and benefits embodied in the effec- The annual cost of ERT, $75, was based on the re- tiveness measure (years of life gained). The basis tail price of PremarinTM (0.625 mg daily for 273 for OTA’s assumptions concerning each compo- days per year). To arrive at this estimate OTA sur- nent is described below. veyed retail pharmacy costs in the Washington, DC area in 1991. BMD SCREENING COST The American College of Obstetricians and The cost of screening for BMD was based on Gynecologists recommends annual mammo- the cost of single photon absorptiometry (SPA). grams and endometrial biopsies for all women on This is the only method of bone densitometry cur- ERT ( 1). However, annual mammograms are gen- 201 Cost of screening Data Constant Author/article year SPA DPA QCT Other dollar year Source Comments Health Care Financing 1990 $1 18a $97a 1990 Medicare a Average submitted Administration, 1993 71b 69b charges b Average allowed charges National Osteoporosis 1990 75.50C 75.50C 1990 Estimate based on C Average Medicare Foundation, 1990 Medicare fee screens reimbursement for SPA and DPA

d d SXA (single photon Vogel et al., 1991 1990 50-150 150-300 150-400 $ 50-$150 1990 Based on information 150-300 e obtained from hospital, x-ray absorptiometry) 75-00 f clinic, and private e DXA (dual energy x-ray 50-1509 settings absorptiometry f Lumbar spine series g Radiogrammetry

Cummings and Black, 1984 40-120 150-300 100-300 — NG American College of 1986 Physicians

American Osteoporosis 1987 75 78 86 — 1987-1988 HCFA Medicare average Alliance, 1990 1988 75 68 81 — allowable amounts.

h h Health Insurance Claim .51 54 — — 1989-1990 Health Insurance ‘Professional Association of America, dates: 116i 159 i — — Association of America’s Component 1991 11/1/89 to Medical Prevailing ‘Total Component 1 0/31/90 Healthcare Charges System

ABBREVIATIONS: DPA = dual photon absorptiometry, NG = not given, QCT = quantitative computed tomography,SPA = single photon absorptiometry SOURCES National Osteoporosis Foundation,Medicare Reimbursement for Bone Mass Measurement:Costsand savings executive summary (Washington, DC August 1990); J Petrie, U S Department of Health and Human Services, Health Care Financing Administration, personal communication, July 1993, J M Vogel, P D Ross, J W Davis, et al , Hawaii Osteoporosis Center, Honolulu, Hl, “Technologies to Detect Osteoporosis” Final Report, unpublished contractor report prepared for the Office of Technology Assessment, U S Congress, Washington, DC, Mar 25, 1991 S Cummings and D Black, “Should Perimenopausal Women Be Screened for Osteoporosis?’’ Annals of Internal Medicine 104(6) 817-823, 1986, American Osteoporo- sis Alliance memorandum to U S House Select Committee on Aging, Subcommittee on Human Services, Washington, DC, 1990, Health Insurance Association of America, B L Harris, Associate Director, Washington, DC, letter to R C Herdman, Office of Technology Assessment, U S Congress, Washington, DC, Aug 30, 1991 Appendix J: Methods for Estimating Costs 203

erally recommended for all women over 50, so endometrial biopsy is the only cost recognized in the model. The cost of this procedure was estimated ERT a,b from average submitted physician charges in 40 Prescription drug $ 98 Blue Cross/Blue Shield plans (10). Endometrial biopsy 129d e In a certain percentage of women on ERT, ade- Dilation and curettage 39 nomatous or atypical endometrial hyperplasia Additional endometrial biopsy 3 f would be detected, requiring dilatation and curet- Total $269 tage (D&C) of the uterus. Weinstein and Schiff es- PERT

a,b timated that 7.5 percent of estrogen users would Estrogen $ 98 be diagnosed with endometrial hyperplasia (19). Progestin 1 19b,c The average submitted physician charge for D&C Doctor visit 38g in 40 Blue Cross/Blue Shield plans was $526 (10); Endometrial biopsy 3 d,h therefore, the expected additional annual cost Total $258 associated with this procedure for estrogen thera- KEY: ERT = estrogen replacement therapy; HRT = hormone replace- py users would be $39. ment therapy; PERT = progestin/estrogen replacement therapy

a Most women who experience unscheduled Drug cost IS from a survey conducted in 1993 by OTA of retail phar- bleeding while on estrogen replacement therapy macies in the Washington, DC area TM would be given an additional endometrial biopsy. bBased on Premarin (0.625 mg daily) for 273 days/year. c TM Weinstein and Schiff have estimated the incidence Based on Provera (1 O mg daily) for 156 days/year. dAverage submitted physician charges from 40 Blue Cross & Blue of bleeding in estrogen users at 2.5 percent (19). Shield plans. Therefore, $3 per year is added to the yearly treat- e Weinstein and Schiff estimated that 7.5 percent of patients on estro- ment cost to account for this procedure. gen alone would need a dilation and curettage The cost of $526 for this procedure was obtained from Blue Cross & Blue Shield average On the basis of all of these estimated compo- submitted physician charges. nents of HRT cost, the total annual cost of ERT in f 2.5 percent of Patients on estrogen alone are assumed to need an the OTA model is $269. additional endometrial biopsy due to increased risk of unscheduled bleeding. g Charge is based on the 1993 Medicare average submitted charge ❚ Estrogen-Progestin Therapy for an established patient requiring limited service (CPT code 90050 in Physicians’ Current Procedural Terminology). In PERT, a progestin is added to the estrogen regi- h Unscheduled bleeding in estrogen-progestin therapy Patients men. The price of the progestin used in our analy- would require an endometrial biopsy. The incidence of bleeding in sis was based on Provera ( 10 mg daily dosage) for this group is assumed to be 2,5 percent. 12 days per four-week cycle. From a survey of re- SOURCES: Office of Technology Assessment, 1995. R. Lapp, Man- ager for Provider Strategy, Blue Cross&Blue Shield Association, Chi- tail pharmacy costs in the Washington, DC area, cago, IL, personal communications, July 18, 1993, and July 23, 1993. we estimated the annual prescription drug cost of M.C, Weinstein and I. Schiff, “Cost-Effectiveness of Hormone Re- PERT at $119. placement Therapy in the Menopause,” Obstetrical and Gynecologi- An annual endometrial biopsy is not usually re- cal Survey 38(8):445-455, 1983. quired for women on PERT, but they would have a yearly physician visit. The cost of the doctor visit, colleagues recently compared the frequency of $38, was based on the Medicare charge for an es- abnormal vaginal bleeding in post-menopausal tablished patient who receives limited service. women taking cyclic PERT with a cohort of Most observers believe that women on PERT women not on HRT of any kind (8). They found a are not at increased risk of endometrial hyperpla- relative risk of abnormal bleeding of approxi- sia. Therefore, extra D&C procedures would not mately 3.1 in women on therapy. The frequency of occur. PERT does result in increased frequency of abnormal bleeding declined with age, however. abnormal vaginal bleeding, however. Ettinger and About 60 extra events per 1,000 patient-years 204 Cost Effectiveness of Screening for Osteoporosis

Fatal acute myocardial infarction $14,470-31 ,397a

a,b Nonfatal acute myocardial infarction 74,217

c Ratio of nonfatal to fatal acute myocardial infarction 2.6/1 Total heart attack costs per fatal heart attack $207 , 4 3 4-$ 224,361

a E. Wittels, J Hay, and A. Gotto, "Medical Costs of Coronary Artery Disease in the United States, ” American Journal of Cardiology 65 ”432-440, 1990, b Flve_year costs discounted to their net present value at the time of the heart attack at an annual rate Of 5 perCent.

C M. Stampfer, G. Colditz, W. Willett, et al , “Postmenopausal Estrogen Therapy and Cardiovascular Disease Ten-Year Follow-Up from the Nurses’ Health Study, ” New England Journal of Medicine 325(11):756-762, 1991 SOURCES” Based on data from sources cited in the footnotes

(6%) occurred in women, aged 50-54; 40 extra COST OF HEART DISEASE events per 1,000 patient-years (470) occurred in The major cardiovascular impact of HRT is a po- women aged 55-59; and about 5 extra events per tential reduction in the number of acute myocardi- 1,000 patient-years (0.5%) occurred in women al infarctions (AMIs) (i.e., heart attacks). The cost aged 60-67. In an early cost-effectiveness study, of AMI used in our analysis, shown in table J-3, Weinstein and Schiff estimated that PERT would reflects that of both fatal and nonfatal heart at- induce abnormal bleeding in approximately 2.5 tacks. percent of patients (19). This estimate is similar to The model assumes that there is a fixed ratio of the data from the Ettinger study, if a single esti- fatal to nonfatal heart attacks. Thus, for every mate of relative risk is used for all ages.5 We there- heart attack death assigned by the model, costs fore adopted 2.5 percent as an estimate of the would accrue both for that woman and for a given number of cases of abnormal bleeding that would number of additional women who would have require an endometrial biopsy. Thus, $3 is added nonfatal heart attacks. to the annual cost of HRT with progestin therapy. OTA used data from the Nurses’ Health Study The total cost of estrogen/progestin treatment is to estimate the ratio of fatal to nonfatal heart at- therefore $258. tacks (13). That study reported the incidence of The components of HRT (i.e., drugs, doctor both fatal and nonfatal (confirmed and probable) visits, tests) used in this model are similar to those AMIs in 48,470 postmenopausal nurses followed used in other analyses of HRT and osteoporosis for 10 years. The observed ratio of nonfatal to fatal (5,15,16,19). Some analyses included only drug AMIs in that sample was 2.6. costs (5,15) or used drug costs plus the costs of Data from the Framingham Heart Study, a large physician visits to monitor treatment without ongoing study of heart disease incidence and out- itemizing the costs of specific diagnostic proce- come, gave cost estimates for fatal and nonfatal dures included in the treatment (16). Our cost heart attacks. The study categorizes coronary components and total cost estimates for HRT— artery disease into five events: acute myocardial $269 per year for ERT and $258 for PERT—are infarction, angina pectoris, unstable angina pecto- consistent with these cost estimates. ris, sudden death, and nonsudden death. Using these outcomes, Wittels and colleagues analyzed

1 If Ettinger’s estimates for the 60-67 year-old population is assumed to hold for older women as well, the average frequency across all ages would be approximately 16 per 1000 patient-years ( 1.6%). But as the population ages, fewer women in the cohort are alive, so this simple average underestimates the true average. Appendix J: Methods for Estimating Costs 205

Medicare cost data, a pharmaceutical price survey, resource used in the Framingham study (based on and Houston area surveys to estimate five-year a cohort of Massachusetts residents) may not rep- costs including those for hospitalization, emer- resent the national population of coronary artery gency room care, monitoring and testing, heart ca- disease patients. theterization, and thrombolytic therapy resulting from acute myocardial infarction (20). The final COST OF HIP FRACTURE cost estimates were based on these charges, An analysis of the health care costs attributed to weighted by their frequency of occurrence. hip fractures is provided in a separate OTA back- In the Framingham study, sudden death was de- ground paper (18). The total cost includes in-hos- fined as a change within a l-hour period from a pital and post-hospital expenses as shown in table stable clinical status to death. Nonsudden death J-4. OTA used the inpatient hospital costs for pa- occurs when a patient admitted with a diagnosis of tients between 50 and 64 years of age as the basis AMI dies in the hospital. We used the costs of sud- for our estimate, because it is higher than the cost den and nonsudden death as the endpoints of the for older women and therefore gives an optimistic range of estimates for the cost of a fatal AMI. In estimate of the cost savings associated with reduc- 1993 dollars, the costs of sudden and nonsudden tions in hip fractures. The total estimated cost (in- death from AMI were $14,470 and $31,397, re- cluding long-term care costs) is $21,189 in 1990 spectively. dollars, or $22,914 in 1993 dollars (after adjusting AMI patients in the Framingham Heart Study for inflation using the CPI-U for all items). who survived hospitalization had a five-year cost (in 1993 dollars) of $81,630. This estimate was COST OF CHOLECYSTECTOMY not discounted over the five-year period, and it also did not include costs incurred in years beyond The cost of a cholecystectomy was obtained from the five-year period of study. To better estimate the rates (physician and hospital costs) allowed by costs spread out over five years, the total was di- Blue Cross and Blue Shield of the Washington, vided into five equal amounts, which were then DC region (1 1). Open cholecystectomy, which discounted to their present value at the time of the costs $11,160, was used as the basis for procedure heart attack at an annual discount rate of 5 percent. cost estimation. Laparoscopic cholecystectomy is We used this discounted cost—$74,217—as the an increasingly frequent procedure costing estimate of the cost of a nonfatal AMI. approximately $9,000. This estimate is consistent Together, these estimates imply that for every with a previous analysis of gallstone treatments, heart attack death, a cost of between $207,434 and which calculated the costs for a cholecystectomy - $224,361 is incurred in treating heart attacks, both based on a 45-year-old female patient at between fatal and nonfatal. These estimates are consistent $6,024 and $18,072 in 1993 dollars (3). with the few analyses that identify costs of AMI by fatal and nonfatal outcomes (9). BREAST CANCER COSTS The costs calculated for the Framingham study There are no accurate data on the lifetime cost of included those for both men and women. If the treating breast cancer. The existing literature pro- cost of treating a female AMI patient is different vides estimates of some parts of cancer treatment, from the cost of treating a male AMI patient, then but for this analysis we constructed age- and the estimates used in our model may be inaccu- stage-specific lifetime breast cancer cost esti- rate. Also, the expected frequencies of events or mates. 206 Cost Effectiveness of Screening for Osteoporosis

Hospital services $ 7,732 In-hospital physician services 1,946 Anesthesia services 576 Radiologic services 298 Physical therapy 785 Total $11,337 In-hospital for persons age 65 and over Hospital services 7,623 In-hospital physician services 1,236 Anesthesia services 319 Radiologic services 116 Physical therapy 28 Total $9,322 Post-hospital for persons of all ages Nursing home care 7,054 Care in a rehabilitation facility or other short-stay hospital 742 Readmission to a short-stay hospital 440 Home health care 453 Nonmedical home care 329 Outpatient physician services 550 Emergency room and ambulance services 284 Total $ 9,852 Total cost of hip fracture for patients 50 to 64 $21,189 Total cost of hip fracture for patients 65 and over $19,174

SOURCE U.S. Congress, Office of Technology Assessment, Hip Fracture Outcomes in People Age Fifty and Over-Background Paper, OTA-BP- H-120 (Washington, DC U S. Government Printing Off Ice, July 1994).

The expected lifetime cost of treating breast years, 8 years, and 13 years (the midpoints of the cancer varies with the number of years of survival surviv-al intervals available in the SEER data). after detection, which is itself a function of both Women who survive for at least 15 years were as- age and stage at detection. Age-and stage-specific sumed to live 15 years plus the average life expec- l-year, 5-year, 10-year, and 15-year all-cause sur- tancy of U.S. women who reach an age equal to vival rates were provided to OTA by the National their age of detection plus 15 years. For example, a Cancer Institute from its Surveillance, Epide- 65-year-old who survives 15 years to age 80 is as- miol-ogy, and End Results (SEER) tumor registry sumed to live the average life expectancy of other data6 (12). These data were used to estimate the 80-year-olds. proportion of women who would live for 1 year, 3

2 The SEER data on cancer incidence, stage, and survival is based on tumor registries maintained in 10 cities. Appendix J: Methods for Estimating Costs 207

Treatment Unadjusted cost Comorbidity costa Net attributable cost Initial care local 13,500 1,369 12,131 regional 14,470 1,369 13,101 distant 14,470 1,369 13,101 Continuing care 10,620/yr. 5,477/yr. 5,143/yr. Terminal care 27,744 2,739 25,005 aComorbidity cost refers to the cost of treating patients for diseases unrelated to breast cancer. SOURCES: M. Baker, L. Kessler, N. Urban, et al., “Estimating the Treatment Costs of Breast and Lung Cancer, ” Medical Care 29(1) 40-49, 1991 U.S. Congress, Office of Technology Assessment, Breast Cancer Screening for Medicare Beneficiaries: Effectiveness, Costs to Medicare and Medical Resources Required (Washington, DC U.S. Government Printing Off Ice, November 1987)

For each combination of age and stage of can- The resulting age- and stage-specific lifetime cer at detection, we assumed that a patient would discounted costs are shown in table J-6. incur an initial cost in the year of cancer onset, continuing care costs in each remaining year of ENDOMETRIAL CANCER COSTS life, and terminal care costs in the final year of life. The osteoporosis prevention strategies tested in Initial costs were assumed to vary with the stage of this analysis are relevant to women with intact cancer at detection, whereas continuing and termi- uteri who are therefore at risk of endometrial can- nal care costs were assumed constant across al- cer. The risk, prognosis, and treatments vary, de- lages and stages. The age- and stage-specific pending on whether the woman is currently on schedule of costs was constructed by combining ERT or PERT at the time of diagnosis. these cost parameters with the age- and stage-spe- OTA assumed that endometrial cancer found cific survival times and discounting the costs in- during HRT would have no excess associated curred over time to their present value in the year mortality. The woman would undergo a hysterec- of cancer detection at an annual rate of 5 percent. tomy and face no permanent loss of vitality. The Estimates of the cost of initial, continuing, and procedure would affect costs only, not the length terminal breast cancer care are available from a of life. (See appendix G for the rationale underly- study conducted by researchers at the National ing this assumption.) Cancer Institute (2). That study used data on a The cost of a hysterectomy was estimated from sample of Medicare patients to estimate the net several published articles shown in table J-7. We health care costs of each kind (initial, continuing, used a mid-range estimate (updated to 1993 dol- and terminal) attributable to breast cancer. lars) of $6,000 to include all costs (physician fees The costs of initial care were not estimated by and hospitalization) associated with a hysterecto- stage in that study, however, so OTA broke down my. the initial care cost into stage-specific costs using If a woman is not on HRT at the time of diagno- information on initial care costs by stage provided sis, the mortality risk depends on age and stage at in another study (7,17). The resulting estimates of diagnosis. The costs for this scenario were esti- initial, continuing, and terminal care costs used in mated in a manner similar to that used for breast the construction of the age- and stage-specific cancer. breast cancer costs are shown in table J-5. 208 I Cost Effectiveness of Screening for Osteoporosis

Age Stage A Stage B Stage C Age Stage A Stage B Stage C 50 $78,153 $67,274 $45,043 71 $50,849 $48,598 $41,285 51 76,671 66,228 44,855 72 49,250 47,527 41,173 52 75,385 65,237 44,823 73 47,340 46,161 40,893 53 73,861 63,966 44,490 74 45,605 44,911 40,679 54 73,181 63,338 44,357 75 43,951 43,818 40,619 55 71,727 62,183 44,176 76 42,293 42,685 40,375 56 70,049 61,107 43,985 77 40,825 41,824 40,352 57 69,310 60,503 43,996 78 39,127 40,514 39,778 58 67,765 59,525 43,720 79 37,542 38,638 39,071 59 66,299 58,668 43,545 80 35,414 37,169 38,497 60 65,717 58,209 43,394 81 33,403 35,869 37,960 61 64,300 57,394 43,220 82 31,637 34,781 37,433 62 63,810 57,135 43,163 83 29,572 33,515 36,655 63 62,019 56,132 42,814 84 27,703 30,667 35,044 64 60,296 55,256 42,653 85 24,941 28,835 34,364 65 59,417 54,851 42,516 86 22,334 27,156 33,643 66 57,742 53,863 42,336 87 20,056 25,791 33,089 67 57,028 53,356 42,268 88 18,209 24,528 32,533 68 54,947 52,001 41,951 89 16,673 19,099 27,503 69 53,831 51,277 41,856 90 12,616 15,837 26,230 70 51,885 50,091 41,601

aCosts recurred in years after detection are discounted to their present value at the age of detection at a rate of 5 percent per annum SOURCE: Office of Technology Assessment, 1995

The initial, continuing, and terminal care costs cedures was about $600, and assuming that the used in the estimation procedure are summarized cost of radiation therapy increases with increasing in table J-8. We used the endometrial cancer treat- dosage, we estimated that the initial treatment cost ment guidelines described in a cancer textbook(6) would be $1,200. The continuing cost was the to determine the components of the cost model. annual cost of radiation therapy, $600, for no more The initial cost was the cost of a hysterectomy if than eight years of treatment, and the cost of annu- the cancer was detected at stage I or stage 11. We al doctor visits, $200. After consulting with an used $6,000 for the cost of a hysterectomy as de- analyst at the National Cancer Institute (4), we as- scribed above. Endometrial cancers found at more sumed that the terminal care cost for endometrial advanced stages (III or IV) are usually treated ini- cancer is the same as the terminal care cost of tially with intense radiation therapy. This cost was breast cancer, or $25,005. obtained from Blue Cross & Blue Shield average These endometrial cancer costs were used in es- submitted charges for several clinical brachyther- timating discounted lifetime cancer costs by age apy treatments (CPT codes 77750,77761,77762, and stage of onset in a cost model similar to the 77776, 77777) (10). The average cost of these pro- Appendix J: Methods for Estimating Costs 209

Source Treatment costs Data Summitt et al., 1992 Average hospital charge for a laparoscopy- Cost data collected from women undergoing assisted vaginal hysterectomy: $7,905 vaginal hysterectomies at the University of Average hospital vaginal hysterectomy Tennessee, Memphis Gynecological Clinic. charge: $2,831.05 Kovac et al., 1991 Average hospital abdominal hysterectomy Based on data collected from all patients charge: $3,584.82 undergoing hysterectomies in a St. Louis Average hospital vaginal hysterectomy Missouri hospital between January 1, 1986 charge: $2,831.05 and December 31, 1986. Health Care Average physician charge for an abdominal Medicare Part B data, Office of Research and Financing hysterectomy: $2,020 (submitted) $890 Demonstrations Administration, 1991 (allowed) Blue Cross & Blue Average submitted charges for physician Based on data gathered from 40 Blue Cross Shield, 1990 (Lapp, services for a total abdominal hysterectomy: plans (New York and California not included.) 1 993) $1,906 Tosteson et al., 1990 Abdominal hysterectomy: $4,900 (based on Physician fees based on Medicare prevailing stage 1, charge includes hospital care and charges or charges at selected Boston-area professional fees) teaching hospitals

SOURCES: R.L. Summitt, Jr., T.G. Stovell, G.H. Lipscomb, et al., “Randomized Comparison of Laparoscopy-Assisted Vaginal Hysterectomy with Standard Vaginal Hysterectomy in an Outpatient Setting,” Obstetrics and Gynecology 80(6):895-901, 1992, S R Kovac, S.J. Christie and G.A. Bindbeutel, “Abdominal Versus Vaginal Hysterectomy’ A Statistical Model for Determining Physician Decision Making and Patient Outcome, ” Med- ical Decision Making 11(1):19-28, 1991, J. Petrie, U S Department of Health and Human Services, Health Care Financing Administration, personal communication, July 1993; R. Lapp, Manager for Provider Strategy, Blue Cross&Blue Shield Association, Center for Health Economics and POIicy Research, Chicago, IL, personal communications, July 18, 1993, and July 23, 1993; A.N.A. Tosteson, D.I. Rosenthal, J Melton, Ill, et al., “Cost Effectiveness of Screening Perimenopausal White Women for Osteoporosis: Bone Densitometry and Hormone Replacement Therapy, ” Anna/s of Internal Medicine 113(8):594-603, 1990.

one for breast cancer. The lifetime costs due to endometrial cancer were discounted by 5 percent per year to the age of onset and weighted by the proba- Initial care: bility of surviving for different lengths of time, All stages based on age- and stage-specific all-cause survival $6,000 (hysterectomy) rates provided to OTA by the National Cancer Stages Ill and IV (radiation treatment) 1,200 Continuing care: Institute (12). These costs are shown in table J-9 Annual doctor visits 200 (see page 210). Annual radiation therapy 600 Terminal care 25,005

SOURCE: Office of Technology Assessment, 1995 210 I Cost Effectiveness of Screening for Osteoporosis

Age Stage A Stage B Stage C Age Stage A Stage B Stage C 50 $15,702 $20,203 $21,552 71 $15,178 $21,482 $22,964 51 15,796 20,467 21,679 72 15,297 21,677 23,064 52 15,925 20,635 21,804 73 15,418 21,891 23,166 53 16,018 20,888 21,964 74 15,599 21,993 23,220 54 16,053 20,985 22,027 75 15,780 21,177 23,322 55 16,207 21,241 22,227 76 15,837 22,119 23,151 56 16,286 21,370 22,317 77 15,941 22,136 23,005 57 16,317 21,509 22,376 78 16,048 22,167 22,863 58 16,394 21,681 22,521 79 14,949 21,015 21,970 59 16,491 21,862 22,623 80 15,114 21,178 21,927 60 16,519 21,932 22,713 81 15,254 21,291 21,882 61 16,581 22,138 22,810 82 15,401 21,474 21,848 62 16,610 22,154 22,860 83 15,742 21,588 21,806 63 16,670 22,351 22,988 84 13,643 19,375 21,028 64 16,733 22,473 23,055 85 13,605 19,438 20,993 65 16,781 22,549 23,165 86 13,525 19,443 20,935 66 16,807 22,660 23,158 87 13,505 19,509 20,892 67 16,831 22,726 23,177 88 13,517 19,648 20,880 68 16,898 22,851 23,224 89 8,990 13,546 15,996 69 16,920 22,907 23,238 90 8,635 13,418 15,890 70 17,014 23,021 23,292 — aCosts incurred in years after detection are discounted to their present value at the age of detection at a rate of 5 percent per annum SOURCE: Office of Technology Assessment, 1995

REFERENCES Institute, Bethesda, MD, personal commu- 1. American College of Obstetricians and Gy- nication, July 19, 1993. 5. Clark, A. P., and necologists, “Hormone Replacement Thera- Schuttinga, J. A., “Targeted Estrogen/Proges- py,” ACOG Technical Bulletin, No. 166 terone Replacement Therapy for Osteoporo- (Washington, DC: American College of Ob- sis: Calculation of Health Care Cost stetricians and Gynecologists, April 1992). Savings,” Osteoporosis International 2(4): 2. Baker, M., Kessler, L., Urban, N., et al., “Esti- 195-200, 1992. mating the Treatment Costs of Breast and 6. DeVita, V.T., Hellman, S., and Rosenberg, S. Lung Cancer,” Medical Care 29(1):40-49, (eds.), Cancer Principles & Practice of On- 1991. cology (Philadelphia, PA: J.B. Lippincott 3. Bass, E. B., Steinberg, E. P., Pitt, H. A., et al., Co., 1989). “Cost-Effectiveness of Extracorporeal Shock- 7. Eddy, D., “Screening for Breast Cancer,” An- Wave Lithotripsy Versus Cholecystectomy nals of Internal Medicine 111(5):389-399, for Symptomatic Gallstones,” Gastroenterol- 1989. ogy 101(1): 189-199, 1991. 8. Ettinger, B., Selby, J. V., Citron, J. T., et al., 4. Brown, M., U.S. Department of Health and “Gynecologic Complications of Cyclic Estro- Human Services, Public Health Service, Na- gen Progestin Therapy,” Maturitas 17(3): tional Institutes of Health, National Cancer 197-204, 1993. Appendix J: Methods for Estimating Costs 211

9. Grimm, R., “Epidemiological and Cost Im- 15. Technology Management Group, Osteoporo- plications of Antihypertensive Treatment for sis Markets--A Worldwide Study on Treat- the Prevention of Cardiovascular Disease,” ment and Prevention (New Haven, CT: Journal of Human Hypertension 3 (Suppl. Technology Management Group, 1989). 2):55-61, 1989. 16. Tosteson, A. N. A., Rosenthal, D. I., Melton, J., 10. Lapp, R., Manager for Provider Strategy, Blue III, et al., “Cost Effectiveness of Screening Cross & Blue Shield Association, Center for Perimenopausal White Women for Osteopo- Health Economics and Policy Research, Chi- rosis: Bone Densitometry and Hormone Re- cago, IL, personal communications, July 18, placement Therapy,” Annals of Internal 1993, and July 23, 1993. Medicine 113(8):594-603, 1990. 11. NIH Consensus Development Conference, 17. U.S. Congress, Office of Technology Assess- “Gallstones and Laparoscopic Cholecystecto- ment, Breast Cancer Screening for Medicare my,” Abstracts of Clinical Care Guidelines Beneficiaries: Effectiveness, Costs to Medi- 5(1):12-15, 1993. care and Medical Resources Required 12. Potosky, A., U.S. Department of Health and (Washington, DC: U.S. Government Printing Human Services, Public Health Service, Na- Office, November 1987). tional Institutes of Health, National Cancer 18. U.S. Congress, Office of Technology Assess- Institute, Bethesda, MD, personal commu- ment, Hip Fracture Outcomes in People Age nication, September 1993. Fifty and Over—Background Paper, OTA- 13. Stampfer, M. J., Colditz, G. A., Willett, W. C., BP-H-120 (Washington, DC: U.S. Govern- et al., “Postmenopausal Estrogen Therapy ment Printing Office, July 1994). and Cardiovascular Disease: Ten-Year Fol- 19. Weinstein, M. C., and Schiff, I., “Cost-Effec- low-Up from the Nurses’ Health Study,” New tiveness of Hormone Replacement Therapy in England Journal of Medicine 325(11): the Menopause,” Obstetrical and Gynecolog- 756-762, 1991. ical Survey 38(8):445-455, 1983. 14. Tanouye, E., “Estrogen Drug Revives Ameri- 20. Wittels, E. H., Hay, J.W., and Gotto, A. M., Jr., can Home Products Sales: Premarin is Poised “Medical Costs of Coronary Artery Disease in to Treat Boomers’ Menopausal Symptoms, the United States,” American Journal of Car- Heart Disease,” Wall Street Journal, p. B4, diology 65(7):432-440, 1990. Aug. 10, 1992. Appendix K: Abbreviations and Glossary K

low-density lipoprotein 1,25 OH2D3 calcitriol or 1,25 LDL

dihydroxyvitamin D3 cholesterol ACOG American College of MI myocardial infarction Obstetricians and Gynecologists MPA medroxyprogesterone acetate AHCPR Agency for Health Care Policy MRI magnetic resonance imaging and Research (DHHS) NCI National Cancer Institute (NIH) AMI acute myocardial infarction NHLBI National Heart, Lung, and Blood BMC bone mineral content Institute (NIH) BMD bone mineral density NIA National Institute on Aging (NIH) CEA cost-effectiveness analysis NIAMS National Institute of Arthritis, CHD coronary heart disease Musculoskeletal and Skin CVA cerebrovascular accident Diseases (NIH) CVD cardiovascular disease NIDDK National Institute of Diabetes and DES diethylstilbestrol Digestive and Kidney Diseases DEXA dual energy x-ray absorptiometry (NIH) DHHS U.S. Department of Health and NIH National Institutes of Health Human Services (DHHS) DPA dual photon absorptiometry NOF National Osteoporosis Foundation ERT estrogen replacement therapy OB-GYN obstetrics and gynecology FDA Food and Drug Administration OHTA Office of Health Technology (DHHS) Assessment (AHCPR) HCFA Health Care Financing OTA Office of Technology Assessment Administration (DHHS) PEPI Postmenopausal Estrogen/ HDL high-density lipoprotein Progestin Interventions Trial cholesterol PERT combined estrogen/progestin HRT hormone replacement therapy replacement therapy IHD ischemic heart disease PHS Public Health Service IU international units QALY quality-adjusted life year IV intravenous 1213 214 Cost Effectiveness of Screening for Osteoporosis

QCT quantitative computed USP DI U.S. Pharmacopoeia Dispensing tomography Information SC subcutaneous VLDL very low-density lipoprotein SPA single photon absorptiometry cholesterol SXA single photon x-ray absorptiometry

19-nortestosterone Anovulation A form of progestin. The absence of ovulation (development and release of an ovum from the ovary). Amenorrhea The absence or abnormal cessation of menstrua- Appendicular skeleton tion. The bones of the extremities, including all of the bones of the limb girdles. Anabolic steroid Any of a group of synthetic derivatives of the an- Arteriography drogen testosterone (a sex steroid or hormone) A diagnostic procedure that allows blood vessels having pronounced anabolic properties and rela- to be seen on X-ray film after the injection of con- tively weak androgenic properties. Anabolic ste- trast material into the bloodstream; used to detect roids such as dromostanolone, ethylestrenol, abnormalities such as obstructions, aneurysms, nandrolone, oxandrolone, oxymetholone, and sta- clots, tumors, and injured organs. nozolol are used clinically mainly to promote Arteriosclerosis growth and repair of body tissues in senility, de- A general term that describes thickened and hard- bilitating illness, and convalescence. ened arteries, the condition that leads to most Androgen (or androgenic hormone) cases of heart disease and a significant proportion of cerebrovascular disease in the United States. Naturally occurring male sex hormones (e.g., testosterone, androsterone, and dehydroepiandros- Atherogenesis terone) and substances that exert biological effects The formulation of patchy plaques of fatty or lipid characteristic of these hormones (e.g., the synthet- material on the inner lining of the arteries, restrict- ic compound methyltestosterone). Androgens in- ing blood flow and encouraging the development hibit bone resorption and increase calcium of blood clots; can result in sudden stoppage of absorption in the intestine, but also have serious blood flow to the heart. side effects. Atherosclerosis Androgenic A descriptive term for thickened and hardened lip- Any substance, e.g., androsterone and testoster- id-rich lesions of the medium and large muscular one, that stimulates male characteristics. arteries; classified into two forms: early lesions, consisting of fatty streaks, and advanced lesions, Angina consisting of fibrous plaques; commonly occurs Any spasmodic, choking, or suffocating pain. The in arteriosclerosis, in which deposits of fibrous term is often used to denote angina pectoris—a and cellular tissue, cholesterol, and fat accumulate condition characterized by severe, transient chest in large and medium-sized arteries, impeding pain, accompanied by a feeling of suffocation, is blood flow; responsible for the majority of cases due to a deficiency in blood supply to the heart. of myocardial and cerebral infarction. Angiography Axial skeleton See arteriography. The spine, ribs, sternum, and skull. Appendix K Abbreviations and Glossary 1215

Bilateral oophorectomy Bone Surgical removal of both ovaries. A specialized connective tissue in which a matrix Bioavailability consisting of collagen fibers, a large variety of The degree to which a drug or other substance be- other proteins and ground substances is impreg- nated with a solid mineral. Bone is the dynamic comes available to the target tissue after adminis- and complex tissue of which the bones in the adult tration. skeleton of humans and other vertebrates is large- Bioequivalence ly composed. The skeleton is composed of two The requirement that a generic product include the kinds of bone: an outer, dense shell of cortical (or same therapeutic ingredient, and that its rate and compact or haversian) bone and an inner, open, extent of absorption be the same as the innovative sponge-like region of cancellous (or trabecular) product. bone. About 80 percent of the mass of the skeleton Bisphosphonates is cortical bone, and 20 percent is cancellous bone. Chemical compounds developed during the past Bone densitometry 20 years for treatment of various bone diseases. A term used to refer to a range of noninvasive Basically, they are carbon-substituted analogues techniques that use a densitometer to measure the of pyrophosphate (an endogenous physiologic in- density of bone (e.g., SPA, DEXA, DPA, SXA) hibitor of bone mineralization). and is used to detect osteoporosis. Several bisphosphonates are under investigation as therapeutic agents for osteoporosis, among Bone density them etidronate, clodronate, tiludronate, pa- The mass of bone substance per unit volume 3 midronate, risedronate, and alendronate. Bis- (g/cm ). phosphonates are considered experimental Bone mineral content (BMC) (investigational) in the prevention and treatment The mass of bone divided by the one dimensional of osteoporosis in the United States. length of bone measured expressed as grams per Blinding cm. A technique used in a randomized clinical trial Bone mineral density (BMD) (RCT) to prevent bias by preventing the patients The mass of bone divided by the two dimensional and/or investigators involved in the trial from projected area of the bone measured, expressed as knowing which participants are receiving which mass per unit area (g/cm2). treatment. In a single-blind RCT, the patients in the trial are “blind” as to which individuals in the Calcaneus study are receiving the experimental or control The heel bone. treatment. In a double-blind RCT, both the in- Calcitonin, human vestigators and the patients are “blind” as to which One of the three calcium-regulating hormones in individuals in the study are receiving the exper- humans, the others being human parathyroid hor- imental treatment and which are receiving control mone (hPTH) and calcitriol. treatment. Additional layers of blinding can be added, as, for example, when a third individual Calcitriol (1,25 OH2D3, 1,25-dihydroxy-

(usually the evaluator of outcomes, the individual vitamin D3 or 1,25-dihydroxychole- analyzing the data) also is unaware of treatment calciferol) assignments. One of several vitamin D metabolizes. Body mass index (BMI) Calcium Weight in kilograms divided by height in meters The most abundant mineral element in the human squared. A measure used to define normal ranges body. In humans, calcium is an essential nutrient of body weight. not only for the normal mineralization of bones 216 Cost Effectiveness of Screening for Osteoporosis

and teeth but also for regulating intracellular Cerebral infarction events in most, if not all body tissues. An area of dead tissue in the cerebrum caused by an interruption of blood circulation due to func- Cancellous (or trabecular) bone tional constriction or actual obstruction of a blood The inner, open, sponge-like region of bone prev- vessel, hemorrhage, etc. Also known as a stroke alent in the vertebrae and in the pelvis, the main or cerebrovascular accident (CVA). sites of osteoporotic fractures. About 20 percent of the mass of the skeleton is cancellous bone. Cerebrovascular disease Compare cortical bone. Any disease of the blood vessels supplying blood to the brain or of the brain’s covering membranes Carcinogen (meninges), characterized by rupture of the blood An agent that causes cancer (e.g., certain chemi- vessels or inadequacy of blood to the brain. Com- cals, ionizing radiation, tobacco smoke, asbestos mon causes include atheroma, hypertension, cere- fibers, and estrogen). bral thrombosis, or embolism. Carcinoma Cholecystectomy A cancer arising from epithelial cells, including Surgical removal of the gall bladder. the external epitheliums (mainly skin and linings Cholesterol of the gastrointestinal tract, lungs, and cervix) and A sterol present in animal tissues (e.g., cell mem- the internal epitheliums that lines various glands branes, blood plasma, and lipoproteins), involved (e.g., breast, pancreas, uterus, and thyroid). in physiological processes, such as the manufac- Cardiac arrhythmias ture of bile acids, sex hormones, and adrenocorti- Variations from the normal rate or rhythm of heart coid hormones; also involved in the development beats. of pathological processes such as atherosclerosis. See also high-density lipoprotein cholesterol and Cardiac catheterization low-density lipoprotein cholesterol. Passage of a small catheter through a vein or artery Climacteric into the heart for the purpose of securing blood The syndrome of endocrine, somatic, and psychic samples, determining intracardiac pressures, and changes occurring at the end of the female repro- detecting cardiac anomalies. ductive period (menopause). Cardiovascular disease (CVD) Clinical trial (also called therapeutic Any of a diverse group of diseases affecting the trial) heart, blood vessels, and/or blood circulation. A research activity that involves the administra- CVD includes diseases of the heart muscle itself, tion of an experimental prophylactic, diagnostic, ischemic heart disease, hypertension, cerebrovas- or therapeutic agent, device, regimen, procedure, cular diseases, and various other conditions. etc. to humans to evaluate its safety and effective- Case-control study ness. The term is subject to wide variation in A type of observational study, where the frequen- usage, from the first use in humans without any cy of a suspected causative factor, such as estrogen control treatment to a rigorously designed and use, is compared in a group of people who have a executed experiment involving test and control disease (cases) and those who do not (controls). If treatments and randomization. See also random- this factor is found with greater frequency in those ized clinical trial (RCT). with disease, a causal association may be sus- Coagulation pected. Results of the comparison may be ex- The process of certain particles joining together to pressed as the relative risk. See relative risk. form larger masses. Appendix K Abbreviations and Glossary 1217

Coagulation proteins trogens have pharmacologic effects similar to Proteins found in the plasma that aid in the coagu- those of endogenous estrogens. lation process of blood (e.g., Factor VIII, Factor Consensus conference IX. antihemophilic factor, or prothrombin). Also A meeting of scientists, medical practitioners, and called “clotting factors.” informed lay people to review scientific in- Cohort study formation about biomedical technologies and to A type of observational study, where the investi- develop a consensus statement on the clinical ap- gator begins with a group of subjects (the “CO- plication of current medical findings. hort”), some or all of whom are exposed to a Coronary artery bypass graft (CABG) suspected causative factor, and follows this cohort surgery over time for development of a disease. Compari- A surgical procedure in which a vein or an artery son is made with a control group composed of un- is used to bypass a constricted portion of one or exposed members of the cohort (internal more coronary arteries. This procedure has be- controls), or to subjects outside the cohort who come the primary surgical approach to the treat- are similar to members of the cohort, but who have ment of coronary artery disease. not been exposed to the suspect factor (population, community, or external controls). Coronary artery disease (CAD) Narrowing or blockage of the coronary arteries, Cones fracture which usually results in reduced blood flow to the Fracture of the wrist. heart muscle. Compact bone Coronary perfusion See cortical bone The pumping of a fluid through the heart by way Comparison group of an artery. In a cohort study, the group of unexposed mem- Cortical (or compact or haversian) bone bers to which the exposed members are compared. The dense outer shell of bone. One of the two gen- In case-control studies, the group of subjects with- eral structural categories of the bone tissue mak- out disease to which the subjects with disease (the ing up the skeleton, cortical bone consists of cases) are compared. tightly packed layers of bone. It forms the outer Confidence interval shell of all bones and is prevalent in the shafts of Conventionally, a 95-percent confidence interval the long bones of the arms and legs. About 80 per- is used, which implies that there is a 95-percent cent of the mass of the skeleton is cortical bone. chance that the true relative risk being measured Compare cancellous bone. falls within the interval, and a 5-percent risk that Cost-effectiveness analysis (CEA) it does not. An analytical technique that compares the costs of Confounding variables a project or of alternative projects to the resultant Variables related to both the disease and the expo- benefits, with costs and benefits/effectiveness ex- sure under study that can explain or alter all or part pressed by different measures. Costs are usually of an observed association. expressed in dollars, but benefits/effectiveness are Conjugated estrogens ordinarily expressed in terms such as “lives An amorphous preparation of naturally occurring, saved,” “disability avoided,” “quality-adjusted water-soluble, conjugated forms of mixed estro- life-years saved,” or any other relevant objectives. gens, chiefly sodium estrone sulfate, extracted Cross-sectional study from the urine of pregnant mares; suitable for pa- In epidemiology, an observational study that ex- renteral, oral, and topical administration and used amines the relationship between diseases (or other in estrogen hormone therapy. The conjugated es- health-related characteristics) and other variables 218 Cost Effectiveness of Screening for Osteoporosis

of interest as they exist in a defined population at or spine. DPA is sometimes used to measure bone one particular time. The temporal sequence of mass in the whole body and can be used to mea- cause and effect cannot necessarily be determined sure bone mass in the forearm, distal radius in a cross-sectional study. (wrist), and calcaneus (heel). Cyclic regimen Duration of exposure Interrupted episodes with ongoing medication. In The length of time a person or test animal is ex- cyclic regimens of estrogen and progestin, estro- posed to a chemical. gen is usually given for 21 days each month, and Dysmenorrhea a progestin is given with estrogen for the last 7 to Difficult and painful menstruation. 14 days of those 21 days. Dyspareunia Densitometry Difficult or painful coitus/intercourse in women. See bone densitometry Dysuria Diethylstilbestrol (DES) Painful or difficult urination. A synthetic estrogenic compound used to treat menopausal symptoms, vaginitis, and suppressed Effectiveness lactation. The same as efficacy (see below) except that it re- Discounting fers to “. . . average or actual conditions of use.” A procedure used in economic analysis to express Efficacy as “present values” those costs and benefits that The probability of benefit to individuals in a de- will occur in future years. Discounting is based on fined population from a medical technology ap- two premises: 1) individuals prefer to receive plied for a given medical problem under ideal benefits today rather than in the future; and 2) re- conditions of use. Efficacy is generally evaluated sources invested today in alternative programs in controlled trials of an experimental therapy and could earn a return over time. a control condition. Compare to effectiveness. Double blind Endogenous See randomized clinical trial (R CT) and blinding. Produced within or caused by factors within the Dual energy x-ray absorptiometry organism. (DEXA) Endometrial biopsy A bone densitometry technique similar to DPA The microscopic examination of a sample of cells, but uses an x-ray machine rather than a radioactive obtained from the lining of the uterus, in order to material to produce a dual-energy radiation beam. evaluate ovulatory function and/or to detect the Different manufacturers use different terms for presence of hyperplasia, dysplasia, or cancer. this technology (e.g., quantitative digital radiog- Endometrium raphy (QDR), dual-energy radiography (DER), The tissue lining the inner uterus, the thickness dual energy radiographic absorptiometry (DRA), and structure of which vary with the phase of the and dual x-ray absorptiometry (DXA). menstrual cycle. Dual photon absorptiometry (DPA) Endothelium A bone densitometry technique that uses a dual- The layer of epithelial cells that lines the cavities energy radioactive material (usually gadolini- of the heart and of the blood and lymph vessels, um- 153) as the source of a dual-energy radiation and the serious cavities of the body. beam; the dual-energy beam allows measurement of bone and soft tissue without the necessity of en- Epidemiological studies casement in water required in SPA or SXA. The Studies concerned with the relationships of vari- site of measurement for DPA is typically the hip ous factors determining the frequency and dis- Appendix K Abbreviations and Glossary 219

tribution of specific diseases in a human approved by FDA for osteoporosis. As of 1995, community. etidronate was the only bisphosphonate available Epidemiology on the U.S. market. Neither etidronate nor any The scientific study of the distribution and occur- other bisphosphonates have been approved by rence of human diseases and health conditions, FDA for indications related to osteoporosis. and their determinants. Etiology Equine estrogen The cause or origin of disease. Estrogen pertaining to, characteristic of, or Exogenous estrogen derived from the horse. See conjugated estrogens. Estrogen that is not produced within the body but Erythema is provided by other means, e.g., tablets, injection, Redness of skin due to congestion of the capil- cream. laries. External controls Esterified estrogen In a cohort study, individuals not part of the cohort A mixture of the sodium salts of sulfate esters of and who have not been exposed, with which the estrogenic substances; used for oral estrogen ther- exposed members of the cohort are compared. In apy. a clinical trial, individuals not formally enrolled Estradiol in the trial who have had an alternative treatment, with which the experimentally treated group is The most potent naturally occurring estrogen in compared. External controls may be historical or humans. concurrent. Estrogen A generic term for estrus-producing compounds; External validity the naturally occurring female sex hormones, in- A measure of the extent to which study results can cluding estradiol, estriol, and estrone. The term be generalized to the population that is repre- also refers to substances occurring in plants or sented by individuals in the study, assuming that made synthetically (as benzestrol or diethylstil- the characteristics of that population are accurate- bestrol) that have biologic activity similar to that ly specified. of estrogens produced in the ovaries of female Femur mammals. The thigh bone, the bone that extends from the hip Estrogen replacement therapy (ERT) to the knee. It is the longest and largest bone in the The use of estrogen for the relief of menopausal body. symptoms, e.g., hot flashes, the prevention of Fibrin heart disease, and the prevention of osteoporosis. A white insoluble protein formed at the site of an Estrone injury from fibrinogen that becomes the founda- An estrogen isolated from pregnancy urine, the tion of a blood clot. human placenta, and palm kernel oil, and also pre- Fibrinogen pared synthetically. A soluble plasma protein synthesized in the liver, Etidronate which is involved in blood coagulation as the pre-

A bisphosphonate (C2H6NA207P2) that was pat- cursor of fibrin. Also called “Factor I.” ented as a therapeutic agent for calcium disorders Fibroblast in 1972 by Procter& Gamble, parent company of A connective tissue cell, found in the skin. Norwich Eaton Pharmaceuticals, Inc. Oral etidronate is indicated for the treatment of Paget’s dis- First-pass hepatic effect ease and some other conditions, but it has not been See hepatic effect. 220 Cost Effectiveness of Screening for Osteoporosis

Functional impairment uals are compared. In a cohort study, unexposed A deficit in an individual’s ability to function in- individuals outside of the cohort, at some time dependently. Functional impairments in elderly previous to the cohort observation period, against people are often described in terms of deficits in which the exposed members of the cohort are activities of daily living (ADLs) and instrumental compared. activities of daily living (IADLs). Hormone Grade A specific organic product of living cells that, The histological appearance of a cancer cell. In transported by body fluids, produces a specific ef- oncology, the classification of cancer according to fect on the activity of cells remote from its point the degree of differentiation of the cancer cell. of origin (e.g., metabolism, growth, and the devel- More differentiated cell types are generally less opment of secondary sex characteristics, such as malignant. breasts and facial hair). Examples of such hormones include insulin, estrogen, progestin, tes- Haversian bone tosterone, adrenaline, and thyroxine. Also a See cortical bone. synthetic substance that resembles a naturally oc- Healthy user effect curring hormone in producing a specific biologi- A phenomenon in epidemiologic studies in which cal effect. subject participants exhibit lower incidence of Hormone replacement therapy (HRT) morbidity or mortality than the general population This term describes either estrogen replacement because they are generally in good health while therapy or combined estrogen and progestin re- the less healthy either choose not to participate in placement therapy when a distinction is not neces- the study or are excluded. sary. Hemostatic Hospital-based case-control studies The arrest of bleeding, whether by the physiologi- In this type of study, all cases diagnosed with the cal properties of vasoconstriction and coagulation disease under study in one or more hospitals are or by surgical means. compared with patients in the same hospitals who Hepatic effect do not have the disease. Pertaining to the liver, the metabolism of estrogen Hot flash by the liver. Sudden sensations of heat and flushing of the face Hepatobiliary and torso, associated with menopause. Related to the gallbladder. Hypermenorrhea High-density lipoprotein cholesterol Excessive menstrual bleeding, but occurring at (HDL) regular intervals and being of usual duration. A class of cholesterol; low levels of HDL are Hyperplasia associated with an increased risk of heart attack. Abnormal increase in the number of normal cells Histology in normal arrangement in an organ or tissue, Microscopic anatomy. The study of the minute which increases its volume. anatomical structure, composition, and function Hypertension of the tissues. Elevated pressure, usually referring to high blood Historical controls pressure—a common and significant cardiovas- In nonrandomized clinical trials, individuals cular disorder characterized by persistently high treated with a “control treatment” outside the arterial blood pressure, usually greater that study proper, at some time previous to the trial, 140mm Hg systolic and 90mm Hg diastolic pres- against which the experimentally treated individ- sure. Appendix K Abbreviations and Glossary 122I

Hypertriglyceridemia Ischemic heart disease (IHD) An excess of triglycerides in the blood. A spectrum of conditions caused by insufficient Hysterectomy oxygen supply to the heart muscle, and the leading Surgical removal of the uterus, in some cases also cause of death in the United states. The most com- including the cervix, ovaries, oviducts, and pelvic mon manifestations of IHD are angina. acute myocardial infarction (heart attack), and sudden lymph nodes. death. In vitro Latency Literally “in glass”; pertaining to a biological pro- Time since first exposure to a suspected causative cess or reaction taking place in an artificial envi- factor. ronment, usually a laboratory. Latent effect In vivo A reaction to a substance that is not immediately Literally “in the living”; pertaining to a biological evident but that appears later in life; also referred process or reaction taking place in a living cell or to as a silent effect. organism. Life expectancy Incidence An expected number of years of life based on sta- In epidemiology, the number of new cases of dis- tistical probability. ease, infection, or some other event having their onset during a prescribed period of time in relation Lipids to the unit of population in which they occur. The A group of organic compounds, classified into incidence rate is the number of new cases of spe- complex lipids (e.g., fatty acids, phospholipids, cified disease divided by the number of people in cholesterol) and simple lipids (e.g., steroids). Lip- a population over a specified period of time, usu- ids are central to a wide variety of metabolic and ally 1 year. structural functions in the body, such as energy storage, formation of hormones and bile acids, Infarction and structure of cell membranes. Necrosis (death) of tissue, resulting from the interruption of blood supply (e.g., as in a heart attack Lipoprotein (myocardial infarction)). Compounds consisting of lipids (fatty substances such as cholesterol) and proteins, the form in Internal controls which lipids are transported in the blood and A control group composed of unexposed mem- lymph fluid. Lipoproteins form the main structur- bers of a cohort. See cohort study. al components of cell membranes and cell organ- Internal validity elles. They are classified as very low-density A measure of the extent to which study results re- (VLDL), low-density (LDL), and high-density flect the true relationship of a “risk factor” (e.g., (HDL) lipoprotein cholesterol. treatment or technology) to the outcome of inter- Low-density lipoprotein cholesterol est in study subjects. (LDL) Ischemia A class of cholesterol; high levels of LDL are Insufficient blood supply to meet the full physio- associated with a greater risk of heart attack. logic needs of the tissue for oxygen (but short of Magnetic resonance imaging (MRI) the degree of ischemia that results in necrosis), A diagnostic technique that produces cross-sec- usually due to atherosclerosis, but also due to inju- tional images of organs and structures in the body ry to blood vessels, muscle spasm, or inefficient by measuring the reaction of nuclei (typically of pumping of the heart. hydrogen protons) in magnetic fields to radio fre- 222 Cost Effectiveness of Screening for Osteoporosis

quency waves. MRI has also been used to measure Menopausal syndrome bone density. Symptoms associated with menopause, e.g., hot Matching flashes, vaginal dryness, osteoporosis. A method of minimizing the impact of confound- Menopause ing factors. Controls may be matched to cases to Cessation of menstruation; the immediate postre- try to minimize other differences between groups. productive phase of a woman’s life, when Matching may be done with groups of subjects menstrual function ceases due to failure to form (stratification) or with individuals, and may be ovarian follicles and ova. Menopause occurs natu- done with highly specific characteristics, such as rally around the age of 50. Menopause is also a age, age at menopause, family history of the dis- secondary consequence of surgical removal of the ease, etc. ovaries, and of certain illnesses (e.g., premature ovarian failure). Medicare A nationwide, Federally administered health in- Menorrhagia surance program authorized by Title XVIII of the Excessive menstruation. Social Security Act in 1965 to cover the cost of Menses hospitalization, medical care, and some related The monthly flow of blood from the female geni- services for eligible persons over age 65, persons tal tract. receiving Social Security Disability Insurance Meta-analysis payments for 2 years, and persons with end-stage A statistical process used to pool results from a renal disease. Medicare consists of two separate number of studies (e.g., from many small random- but coordinated programs—hospital insurance ized clinical trials) to enable the demonstration of (Part A) and supplementary medical insurance statistically significant differences when the re- (Part B). Health insurance protection is available sults are combined. to insured persons without regard to income. Metastasis MEDLINE database The process by which malignant cells spread to The original, largest, and most utilized database in distant body sites via the lymphatic circulation the National Library of Medicine’s computerized of the bloodstream; also, a secondary malignant retrieval and technical processing system. MED- tumor. LINE contains references to biomedical and other Metrorrhagia literature relevant to health and health services. Uterine bleeding, usually of normal amount, oc- Medroxyprogesterone acetate (MPA) curring at completely irregular intervals, the peri- A form of progestin. Also known as Provera od of flow sometimes being prolonged. (Wyeth-Ayerst). In the United States, MPA is the Morbidity most commonly used progestin in combined es- The condition of being ill or otherwise afflicted trogen/progestin replacement therapy. with an unhealthful condition. Membership bias Morbidity rate Bias introduced when the group or cohort being The rate of illness in a population, calculated as studied is unrepresentative of the population at the number of people ill during a time period di- large, so that comparisons and extrapolations to vided by the number of people in the total popula- the population group are unjustified. tion; used to refer to incidence or prevalence rates of disease. Menarche Mortality rate The onset of menses at puberty. The death rate, often made explicit for a particular characteristic; e.g., age, sex, or specific cause of Appendix K Abbreviations and Glossary 223

death. A mortality rate contains three essential (e.g., case-control studies, cross-sectional studies, elements: 1 ) the number of people in a population and cohort studies). Such studies are the tradition- group exposed to the risk of death (the denomina- al source of information on suggestive associa- tor); 2) a time factor; and 3) the number of deaths tions in epidemiology. occurring in the exposed population during a cer- Occlusion tain time period (the numerator). In the context of the vascular system, the blocking Myocardial infarction (Ml) off or obstruction of blood flow through a vessel. Heart attack. Sudden necrosis (death) of tissue in Odds ratio the myocardium (heart muscle) characterized by A measure of association closely related to rela- severe, unremitting chest pain. leading to arrhyth- tive risk; the ratio of the odds of a disease occur- mias and/or heart failure; in most cases, caused by ring in individuals exposed to the risk compared coronary atherosclerosis (obstruction of coronary to those unexposed. For large samples, the odds vessels, leading to insufficient blood supply to the ratio is essentially equal to the relative risk. heart muscle). Oligomenorrhea Myocardium Abnormally infrequent menstruation. Muscle of the heart. Natural estrogen Oophorectomy An estrogen derived from natural sources, (i.e., Excision of one or both ovaries. not synthetic), such as conjugated equine estro- Opposed estrogen gens, estradiol, or estriol. Estrogen used in conjunction with progestin. Natural menopause Osteoblast Menopause that occurs as a natural part of the ag- A cell arising from a fibroblast, which, as it ma- ing process; not surgically induced. tures, is associated with bone production. Naturally occurring estrogenic Osteopenia hormones A reduction in the amount of bone mass, leading Female sex hormones produced by the ovaries, the to fractures after only minimal trauma. placenta, testes, and possibly the adrenal cortex. Osteoporosis Neoplasm A systemic skeletal disease characterized by low Uncontrolled and progressive growth of tissue, ei- bone mass and microarchitectural deterioration of ther benign or malignant; a tumor. bone tissue, with a consequent increase in bone Nested case-control studies fragility and susceptibility to fracture. Case-control studies conducted, or “nested,” P value within a cohort group. In epidemiologic studies, the probability of con- Nonischemic heart disease cluding that a statistical association exists be- Heart disease from causes other than coronary tween, for instance, a risk factor and a health artery disease (e.g., congenital heart disease, myo- endpoint, when, in fact, there is no real associa- cardiopathy). tion; the likelihood that an observed association in Norethidrone acetate a study is due to the play of chance. Also called .4 progestational agent similar in action to proges- “Type I error” or “alpha,” and commonly called terone. the level of significance. See significance level. Observational study An epidemiologic study in which the experiences of the groups being compared are simply observed 224 Cost Effectiveness of Screening for Osteoporosis

Parathyroid hormone (PTH) or human the blood vessels) of large and medium-sized parathyroid hormone (hPTH) veins. A polypeptide hormone which regulates the con- Platelets centration of extracellular fluid calcium. Disk-shaped tissue, found in the blood of mam- Parenteral mals, which responds to injury elsewhere in the Administration through routes other than the ali- body. Platelets are known for their role in blood mentary canal. Parenteral administration includes coagulation (clotting). Also called “thrombo- intravenous. subcutaneous, transdermal, intraoc- cytes.” ular, and intranasal administration. Population-based case-control studies Pathophysiology In this type of study, all cases diagnosed in the The physiology of disordered function. community or in a sample of the general popula- Percutaneous tion are compared with controls selected from the Literally, “through the skin”; refers to a surgical community or a sample of the general population. procedure that requires only a very small incision, Postmenopause such as a biopsy, or aspiration of fluid beneath the The period of time after the menopause. skin using a needle, catheter, or syringe. Power Perimenopause The power of the study refers to the chance of find- The time around the menopause. ing a true difference in risk and labeling it as statis- Peripheral conversion tically significant. Thus the power of the study is Conversion of estrogen outside of the liver, in pe- equal to 0.80 when the sensitivity level is 0.20 ripheral tissues. (i.e., 1- 0.20= 0.80). Peripheral nervous system Predictive test The autonomic nervous system, the cranial A medical test generally applied to asymptomatic nerves, and the spinal nerves including associated individuals to provide information regarding the receptors. future occurrence of disease. Pharmacodynamics Premarin The study of the actions of drugs on living sys- Wyeth-Ayerst’s brand of conjugated estrogen. tems. The most commonly used estrogen for HRT in the Pharmacokinetics United States. See conjugated estrogens. The rate of change in a physical or chemical sys- Premature ovarian failure tem, specifically in relation to drugs. Condition characterized by the failure to ovulate Placebo before the normal age of menopause. A drug or procedure with no intrinsic therapeutic Premenopause value. In a placebo-controlled randomized clini- The stage of life before menstruation stops. cal trial, a placebo is given to patients in control groups as a means to blind investigators and pa- Premenstrual syndrome tients as to whether an individual is receiving the The pattern of symptoms related to the menstrual experimental or control treatment. See random- cycle. ized clinical trial (RCT). Preovulation Plaques The first 14 days of a woman’s menstrual cycle, Yellowish fatty deposits formed within the intima when estrogen levels are rising before ovulation and inner media (innermost and middle coats of takes place. Appendix K Abbreviations and Glossary 225

Prevalence prevention of osteoporosis; progestin opposes the A measure of the number of individuals in a given carcinogenic effects of estrogen on the endome- population who have a specific disease or other trium. condition at a designated time (or during a particu- Progestogen lar period). See progestin. Primary prevention A category of health and/or related interventions Prospective study that aim to eliminate a disease or disordered state An epidemiologic study in which data are gath- before it can occur. Compare secondary preven- ered after a hypothesis has been generated and the tion. study approved. In a prospective study, the investigator first identifies the subjects and then fol- Progestational lows them over time for development of disease. Favoring pregnancy; conducive to gestation; hav- Compare retrospective study. ing a stimulating effect upon the uterine changes essential for the implantation and growth of the Provera fertilized ovum. Referring to progesterone, or to Wyeth-Ayerst’s brand of the progestin medroxy- a drug with progesterone-like properties. progesterone acetate (MPA). The most commonly used progestin for HRT in the United States. See Progesterone (also called progestational progestin, progesterone. hormone) An antiestrogenic female sex hormone secreted by Quality-adjusted life-year (QALY) the ovaries (specifically by the corpus lutetium, In cost-effectiveness analysis, a measure of health formed immediately after ovulation), by the pla- impact used to compare the benefit or effective- centa during pregnancy to prepare the inner lining ness of alternative health interventions, such as of the uterus for implantation of an ovum, and in the value of an extra year of life gained through small amounts by the adrenal glands and testes, kidney transplantation versus dialysis for patients and prepares the inner lining of the uterus for preg- with end stage renal disease; involves some de- nancy. Receptors for progesterone have been iden- gree of arbitrary valuation and weighting of differ- tified on osteoblasts, and although the data are not ent conditions. entirely clear, it appears that progesterone may Quantitative computerized tomography stimulate bone formation. Progesterone and other (QCT) agents capable of producing some or all of the bio- A bone densitometry technique similar to DEXA logical effects similar to those of progesterone are which measures the central, trabecular portion of called progestins (or progestogens). the vertebral body, the spongiosa. Techniques Progestin have been developed to allow for quantitative Originally, the crude hormone of the corpus lute- measurement on most commercially available CT tium of the ovary; it has since been isolated in pure scanners. form and is now known as progesterone. The ge- Radiographic densitometry neric term for any substance, natural or synthetic, Noninvasive measurement of bone mass in vivo that effects some or all of the biological changes was first performed by quantifying radiographs produced by the hormone progesterone. See pro- with an optical densitometer. The film densitv gesterone. . . over the bone is compared to that over the soft tis- Progestin/estrogen replacement therapy sue, and the resultant absorption is related to that (PERT) obtained over a series of standards with known The use of estrogen combined with progestin for mineral content. Because of technical errors re- the treatment of menopausal symptoms, e.g., hot lated to the polychromate nature of the radiation flashes, the prevention of heart disease, and/or the source and differences in x-ray film characteris- 226 Cost Effectiveness of Screening for Osteoporosis

tics, this technique has many errors and is rarely measuring the effects of several factors concur- used clinically today. rently. Radiogrammetry Relative risk X-rays of the hand and radius, using fine grain in- A measure of a relationship, defined as the chance dustrial film, have been used to measure the corti- of an outcome, such as breast cancer, among a cal thickness of these tubular bones as an indicator group of persons having a suspected causative fac- of cortical bone mass. Although the proximal ra- tor, divided by the chance of this outcome among dius and ulna and the six middle metacarpal of a similar group without this suspect factor. both hands have been used, the mid-shaft of the Reliability second metacarpal is the usual measurement site. The reproducibility of results over repeated mea- Radius surements, and relates to the lack of random error The bone on the outer or thumb side of the fore- over these repeated measurements. Reliability is arm. The distal radius is the end of the radius a prerequisite to validity. bone adjacent to the wrist. The proximal radius is the end of the radius adjacent to the elbow. Retrospective study An epidemiologic study in which data that are al- Random allocation ready available are analyzed to test a hypothesis In a randomized clinical trial, allocation of indi- (e.g., inferences about exposure to a possible viduals to treatment groups such that each individ- causal factor are derived from data on subjects ual has an equal probability of being assigned to who already have the disease in question, any group. compared to other subjects who do not have the Randomized clinical trial (RCT) (also disease). called randomized controlled clinical Risk factor trial or controlled clinical trial) An aspect of personal behavior or lifestyle, an en- In epidemiology, a clinical trial of a prophylactic, vironmental exposure, or an inborn or inheritance diagnostic, or therapeutic agent, device, regimen, characteristic, which on the basis of epidemiolog- procedure, etc. in which human (or animal) sub- ic evidence is known to be associated with health jects are randomly allocated into groups, usually related conditions considered important to pre- called “the experimental group” (in which sub- vent. jects receive the treatment being studied) and “the control group” (in which subjects do not receive Route of administration the treatment being studied), and outcomes are In pharmacology, refers to the means by which a compared. drug is administered: namely intravenous (in- Recall bias jected into the bloodstream), inhalation (through Bias caused by differences in abilities of two the lungs), oral (through ingestion), and dermal groups to remember exposure to suspected caus- (through the skin). ative factors. Secondary preventions Recency An intervention that strives to shorten the course Time since last exposure to a suspected causative of an illness by early identification and rapid inter- factor. vention. Regression analysis Selection bias A statistical procedure for determining the best A distortion in the estimate of effect resulting approximation of the relationship between vari- from the manner in which subjects are selected for ables. Multiple regression analysis is a method for a study population. Appendix K Abbreviations and Glossary 227

Sensitivity by the sum of true-negatives plus false-positives). The percentage of all those who actually have the Compare sensitivity. condition being tested who are currently identi- Statistically significant fied as positive by the test. The likelihood that an observed association is not Sensitivity analysis due to chance. See P value. In cost-effectiveness and cost-benefit analysis, an Steroid hormone analysis of the effect of changes in key assump- Any of numerous hormones characterized by ste- tions or uncertainties on the findings and outcome roid structure (i.e., four carbon rings interlocked of the overall study. to forma hydrogenated cyclopentophenanthrene- Serum lipid profiles ring system). Steroid hormones include the sex A quantitative representation of the level of serum hormones estrogen, progestin, and androgen; they lipids. also include cortisone and adrenocortical hormones. Serum triglycerides Esters formed from glycerol and one to three fatty Steroids acids; fats and oils are triglycerides. Any of a class of compounds characterized by a polycyclic structure like that of the sterols and that Significance level usually include the sterols (e.g., cholesterol) and The significance level, or “p” value, is the proba- vitamin D, as well as many other naturally occur- bility of concluding that a relative risk is different ring compounds (e.g., bile acids). from 1.0 when it is not. By convention, a difference in risk is said to be statistically significant if Stratification there is less than a five percent chance of making In randomized clinical trials, the categorization of this type of error (i.e., p < 0.05). The significance individuals for the purpose of adjusting the groups can also be expressed as a confidence interval. to take into account unequal distribution of characteristics of prognostic importance. Stratifica- Single photon absorptiometry (SPA) tion may be used during patient allocation, A bone densitometry technique that employs a creating subgroups within which individuals are sealed source of radioactive material that emits a randomized to treatments; or stratification maybe single energy radiation (photon), and a detector applied during data analysis to statistically adjust that measures the amount of photons transmitted for differences between the groups. through bone and soft tissue. The radius and calcareous can be measured with this device. Stress incontinence See urinary stress incontinence. Single photon x-ray absorptiometry (SXA) Stroke A bone densitometry technique similar to SPA but Loss of sensation, movement, or function caused using an x-ray source rather than a radioactive by a sudden interruption of the blood supply or a source. Both the radius and calcareous can be leakage of blood in the brain. This can be caused measured with this device. by heart failure, blockage of arteries (cerebral thrombosis or cerebral embolism), or hemorrhage Specificity in the brain. One measure of the validity (or accuracy) of a diagnostic or screening test: the percentage of all Subcutaneous those who do not have the condition being tested Beneath the skin. or who are correctly identified as negative by the Subgroup analysis test. Operationally, it is the number of negative A separate analysis performed on a subgroup of test results divided by the number of patients that the population being studied to identify important actually have the disease (true-negatives divided differences in risk. Such differences in risk maybe 228 Cost Effectiveness of Screening for Osteoporosis

due to interactions of the purported causative fac- artery or vein, causing thrombophlebitis or deep tor with other risk factors for the disease. vein thrombosis. Factors contributing to thrombosis include atherosclerosis, an increase in coag- Surgical menopause Menopause following the surgical removal of the ulation factors, or a deficiency of anticlotting ovaries. factors in the blood. Symptomatology Total cumulative exposure The combined symptoms of a disease. The total dose, which is related both to the level and to the duration of exposure. For example, for Synthetic calcitonin estrogen replacement therapy, total dose is related A synthetically produced/manufactured form of to how much and for how long the estrogen is the hormone calcitonin, a 32 amino acid polypep- given. tide, one of the three calcium regulating hormones found in humans. The FDA has granted qualified Trabecular bone approval for both natural and synthetic calcitonin See cancellous bone. for the treatment but not the prevention of osteo- Transdermal porosis. Through the skin. Synthetic estrogen Transmenopausal A synthetically produced/manufactured estrogen Occurring across the time period of the meno- product. pause. Systemic circulation Treatment group Channels through which nutrient fluids of the In a randomized clinical trial, the group receiving body flow: often restricted to the vessels convey- the treatment being evaluated for safety and effi- ing blood. cacy. Also known as the experimental group. See Testosterone randomized clinical trial (RCT). An androgen or steroid hormone secreted by the interstitial cells of the testes, which functions in Tumor the induction and maintenance of male secondary A new growth of tissue in which the multiplica- sex characteristics and affects sperm production; tion of cells is uncontrolled and progressive. Also testosterone and its cypionate, enanthate, and pro- called neoplasm. pionate esters are used in palliative therapy in in- Ultrasound operable carcinoma of the female breast and Predictor of fracture risk using sound velocity and certain gynecologic conditions. sound attenuation to measure both bone mass and Thromboembolitic disease altered bone architecture. Disease related to blood vessel obstruction. Unopposed estrogen Thrombosis Estrogen used alone, without a progestin. Also The abnormal development of a blood clot known as estrogen replacement therapy (ERT). (thrombus) inside an intact blood vessel, which Unopposed progestin can be life-threatening if it obstructs the blood Progestin used alone. Progestins alone have been supply to the brain (leading to stroke), heart (lead- used to relieve menopausal symptoms. ing to myocardial infarction), the lungs (leading to pulmonary embolism), or other organs (leading to Urinary bladder tissue damage or loss of function); the presence of The hollow, muscular organ that collects urine such clots also raises the risk that part of the clot from the ureters and stores it until the urine is dis- (an embolus) may break off and travel to a distant charged through the urethra during urination. Appendix K Abbreviations and Glossary 229

Urinary incontinence Vagina An involuntary loss of urine sufficient in quantity The canal in the female, from the vulva to the cer- and/or frequency to be a social or health problem. vix uteri, that receives the penis in copulation and Urinary stress incontinence is the birth canal. Involuntary escape of urine due to strain on the Vaginal atrophy orifice of the bladder, as in coughing or sneezing. The wasting or diminution in size of the vagina. Vaginismus Urodynamics A process that evaluates characteristics of the Painful, involuntary contraction or spasm of the muscles around the outer third of the vagina, inter- urine stream and the pelvic musculature, and the fering with sexual intercourse. activity of the bladder. Validity Uterine prolapse A measure of the extent to which an observed situ- Descension of the uterus down into the vagina, ation reflects the “true” situation. caused by weakening of the support ligaments and muscles that hold the uterus in place. Vertebra In the human body, any of the 33 bones of the spi- Uterus nal column, comprising the 7 cervical, 12 thorac- The hollow muscular organ in the female in which ic, 5 lumbar, 5 sacral, and 4 coccygeal vertebrae. the fertilized ovum normally becomes embedded Withdrawal bleeding and in which the developing embryo and fetus are Bleeding, associated with hormone replacement nourished. Its cavity opens into the vagina below therapy, caused by the sloughing of the endome- and into a fallopian tube on either side. trium due to withdrawal of estrogen stimulation.

*U.S. G.P. O. : 1995-387-789:47402