Tobacco Documents Online

h JOUKNAL OI~ WOMEN'S HEALTH Volume 6, Number 1, 1997 Mary Ann Liebert, Inc. B925 XF11B ~9 OOLD J WOHENS HEALTH 97 ~O~AR¥ AR~ LIEEERT IRO PgBL NY 20-Year The Nurses' Health Study: Contribution to the Understanding of Health Among Women GRAHAM A. COLDITZ, M.D., Dr.P.H., JOANN E. MANSON, M.D., Dr.P.H., and SUSAN E. HANKINSON, R.N., Sc.D. ABSTRACT The Nurses' Health Study was designed as a prospective follow-up study to examine rela- tions between contraception and breast cancer. With follow-up questionnaires mailed every 2 years, investigators have added extensive details of lifestyle practices. The study, currently in its 20th year, has maintained high follow-up with >90% of participants responding to each of the follow-up cycles since 1988. The relations between use of hormones, diet, exercise, and other lifestyle practices have been related to the development of a wide range of chronic ill- nesses among women. This review describes the methods used to follow up the study par- ticipants and summarizes the major findings that have been described over the first 20 years of the study. We highlight additional areas added to the study in recent years to address emerging issues in women's health. Special emphasis is placed on the recent findings from the study, including relations between weight gain and heart disease, diabetes, and mortal- ity, the lack of relation between calcium and ogteoporotic fractures, and the positive relation between postmenopausal use of hormones and risk of breast cancer. INTRODUCTION THE NURSES' HEALTH STUDY COHORT initially comprised 121,700 female registered nurses who returned a mailed questionnaire in 1976. The nurses.were 30-55 years of age, mar- fled, and resided in one of 11 U.S. states (California, Connecticut, Florida, Maryland, Massachusetts, Michigan, New Jersey, New York, Ohio, Pennsylvania, or Texas) according to 1972 files provided by the state boards of nursing and the American Nurses' Association. In June 1976, under the direction of Frank E. Speizer, M.D., principal investigator, an intro- ductory letter, a two-page questionnaire, and a prepaid return envelope were sent to each nurse. Identical materials were mailed in September and December 1976 in an attempt to enlist the participation of previous nonrespon- dents. Overall, 70% of those invited to partici- pate in the study returned questionnaires.1 Funded initially to examine relations be- tween the use of oral contraceptives (OCs), cig- arette smoking, and risk of major illnesses in women, the study has been broadened over time to include the evaluation of health conse- Channing Laboratory, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts. This work was supported by grant CA 40356 from the National Institutes of Health. This manuscript expands on work previously summarized in J Am Med Worn Assoc 1995;50:40. THIS ARTICLE IS FOR INDIVIDUAL USE ONLY AND MAY NOT BE FURTHER REPRODUCED OR STORED ELECTRONICALLY WITHOUT WRITTEN PERMISSION FROM THE COPYRIGHT HOLDER. UNAUTHORIZED REPRODUCTION HAY RESULT IN FINAND~AZ A)~DDTHER PENAlTiES. 49

5O COLDITZ ET AL. quences of many lifestyle practices, including diet, physical activity, and specific forms of es- trogen replacement therapy. Although the ma- jor source of funding remains the extramural program of the National Cancer Institute (NCI), the wide range of conditions studied has re- sulted in supplemental funding from the National Heart, Lung, and Blood Institute, the National Institute of Diabetes and Digestive and Kidney Diseases, the National Institute of Arthritis and Musculoskeletal and Skin Diseases, and the National Eye Institute. A more recent addition is funding from the National Institute on Aging to study work stress and quality of life. In addition, funding for pilot studies has been received from several pharmaceutical firms and from the National Dairy Council and the Florida Citrus Commission/Florida Department of Citrus. Thh research effort is coordinated by a group of investigators at Harvard Medical School and Harvard School of Public Health, who meet every 2 weeks to review the progress of the component studies and plan analyses and fur- ther research questions. The study was designed as a prospective co- hort investigation to define the relation between OCs and cancer. Based predominantly on data on women's use of OCs marketed dur~." g the 1960s and 1970s, the cohort provides extensive data on the health effects of these early OC for- mulafions. Nurses were chosen because of the higher accuracy of information that they would report than would a broader sample of women. Further, they were expected to understand some of the issues involved in research studies and so participate more readily than women in general. Information was collected from participants while they were free from disease, thus avoid- ing problems of recall of lifestyle factors that plague retrospective studies. Only participants free from disease are followed to examine dis- ease incidence. In an incidence-based follow-up study, the histories of those who subsequently develop disease are compared with the histories of women the same age who remain free from disease. Because we mail follow-up question- naires to all cohort members, women who have been diagnosed with cancer and other major ill- nesses also provide updated information. In the future, as numbers become sufficiently large, we may be able to examine diet, activity, and other lifestyle factors after diagnosis and their relation to survival. After the cohort was established in the initial grant period, additional funding was obtained from the NCI to follow the women to study hypotheses relating cigarette smoking, hair dyes, and postmenopausal hormones to the risk of a range of cancers and cardiovascular dis- eases. FOLLOW-UP OF PARTICIPANTS Follow-up questiormaires are mailed to all cohort members every 2 years. These ques- tionnaires are mailed along with cover letters and a newsletter that updates participants on the progress of the study. Each follow-up ques- tionnaire inquires about a number of exposures as well as the development of cancer, cardio- vascular disease, and other major medical con- ditions diagnosed since the last follow-up. The first follow-up questionnaire is mailed in June of even-numbered years (1978, 1980, 1982, and so on), and those who do not respond are sent a second mailing in September. On average, 80,000 women respond to the first mailing. .Subsequently, we send a third and fourth ques- tionnaire to those who still have not responded. Finally, a fifth mailing Of a short questionnaire that includes only a few key exposure variables and the list of major illnesses is sent. This fifth mailing, which includes a newsletter to update participants, is sent in June of odd-numbered years. This strategy ensures that any change of address is obtained from the post office (whose usual practice is to keep address forwarding or- ders for only 12 months). Most deaths are reported by the subject's next of kin or by postal authorities. These re- ports are supplemented by searches of the National Death Index for deaths among the nonrespondents. Using these methods, we es- timate that more than 98% of deaths in this co- hort have been identified.2 In 1982, we added a telephone follow-up to reach those women who had not responded to any of the five mailings. More than 14,000 women were successfully contacted and com- pleted a brief telephone interview focused on any newly diagnosed illness. Telephone fol-

NURSES' HEALTH STUDY low-up was repeated after the 1986 follow-up cycle. In 1988, we used a series of additional ap- proaches, including sending questionnaires by UPS and certified mail;3 and achieved a re- sponse of 88%. In 1990 and subsequently, using both telephone and certified mail to reach initial nonresponders, responses were received from just over 90% of the women in the study. Overall, participation has been very high, a trib- ute to the dedication of the women in the study. Each year we are notified of more than 4000 address changes. In addition, some mail is re- turned to us as undeliverable. Using mecha- nisms developed over the last 20 years, we trace these women through direct contact with the local postmaster, the state boards of nurs- ing, and a contact person designated by the study participant (contacts were identified by study members in 1978, 1982, 1986, 1988, and again in 1992). Through these approaches, we have successfully located the majority of par- ticipants with whom we have lost contact at some time. CONFIRMATION OF SELF-REPORTED ILLNESSES For any report of cancer (except basal cell skin cancer), we seek written permission from study participants to review their medical records. We telephone nonrespondents to this request to obtain verbal confirmation of the in- formation reported on the follow-up question- naire (asking details of diagnosis and treat- ment, such as chemotherapy). All medical records are reviewed by trained physicians blinded to exposure information previously provided by the study participant. For women reporting a myocardial infarc- tion or stroke, we also seek the medical records pertaining to the initial diagnosis. Myocardial infarction is classified as confirmed if the records meet the criteria of the World Health Organization, including symptoms and either typical electrocardiogram changes or eleva- tions of serum cardiac enzymes.4 Stroke is alas- sifted according to the criteria developed by the National Survey of Stroke.5 On the 1982 questionnaire, we added an item seeking a history .of fracture of the hip or fore- arm and details regarding the diagnosis of gall- stones and cholecystectomy. Diagnostic details of these major medical conditions have been in- duded on subsequent follow-up question- naires. Using a similar approach, we have added documentation .of self-reported colon polyps and a range of eye conditions, includ- ing cataract surgery, macular degeneration, and glaucoma. After the 1984 follow-up questionnaire cycle, we mailed supplementary questionnaires to all women who had ever responded affirmatively to the question "Have you ever been diagnosed as having diabetes mellitus?" on any of the previous questionnaires. This supplementary questionnaire included items on symptoms of diabetes at the time of diagnosis, fasting and random glucose levels, oral glucose tolerance testing, presence of glycosuria or ketonuria, history of ketoacidosis (including hospitaliza- tion), history of diabetes treatment, and gesta- tional diabetes. Earlier cohort studies conducted in the United Kingdom to document the health con- sequences of cigarette smoking used popula- tions of doctors6 to reduce the likelihood of er- ror in the reporting of illnesses and to facilitate follow-up, as the professional register served as a means to trace the physicians. Similarly, in establishing a large cohort of women, a key consideration was the ability of participants to accurately report the diagnosis of major ill- nesses. Because each reported disease must be confirmed, even a small increase in documen- tation due to erroneous reporting would greatly increase the cost of the study. The ex- tremely accurate reporting of major medical conditions by Nurses' Health Study partici- pants has contributed greatly to the cost-effec- tive nature of this large study. After confirming illnesses reported on the • 1978 and 1980 follow-up questionnaires, we re- ported the level of agreement. Overall, almost all self-reported cancers were confirmed by medical record review.7 Application of strict criteria for cardiovascular end points may re- sult in rejection of some true cases and a slight underestimate of the true incidence of disease, but with few false-positive diagnoses. The reliability of reporting of hypertension, high blood cholesterol, fractures, and diabetes

has been confirmed in random samples of women. Agreement between self-report and medical records has been high, more than 98% for those conditions. In contrast, for classic con- nective tissue diseases, we were only able to document <20% of cases when applying stan- dard diagnostic criteria as defined by the American College of Rheumatology to infor- mation contained in medical records. STATISTICAL ANALYSIS We perform statistical analyses on data col- lected prospectively from participants in the Nurses' Health Study. All data are analyzed for statistical purposes only, and the confidential- ity of participants is maintained by storing all questionnaire information by identification number only. Names and addresses are stored on a computer system separate from the com- puter that stores questionnaire response data. We use relative risk as a measure of association between exposure (lifestyle variables) and dis- ease. The relative risk, or rate ratio, is calcu- lated as the rate of disease among women in each category of an exposure (e.g., duration of use of OCs) divided by the rate of disease among women in the reference category (e.g., women who have never used OCs). Relative risks are adjusted for age in 5-year intervals. To control simultaneously for age and other po- tential confounding variables, we use either lo- gistic regression or proportional hazards (Cox) models. LIFESTYLE EXPOSURES The design of the Nurses' Health Study in- cludes several unique features. Among these is the repeated assessment of lifestyle and other exposure variables. Such repeated assessment is needed, at least in part, because of the ques- tions being asked by the study. For example, given a focus on OCs and health, we need to update the status of women who are using these or other exogenous hormones, such as those used in postmenopausal hormone ther- apy. The changing availability of products and their patterns of use, such as the addition of COLDITZ ET AL. progesfins to postmenopausal estrogen and varying number of days per month that these products are used, preclude the application of more controlled research designs to address risks associated with current prescribing pat- terns. Likewise, with > 1% of smokers stopping every year, it is necessary to update smoking status on a regular basis to accurately estimate the relation between current smoking and dis- ease as well as the benefits of quitting smok- ing. The repeated measurement of lifestyle al- lows for the study of many factors as they relate to health. Another benefit of repeated ques- tionnaires has been the ability to add items to the follow-up questionnaires to address new and evolving hypotheses. Among the many ad- ditions have been some of the diseases and conditions discussed and such variables as diet, physical activity, and screening behaviors. Several studies have grown out of the ongo- ing Nurses' Health Study: a study of mortality among spouses who use vasectomy as their form of contraception,8 a new study of younger nurses to address questions that we cannot ad- equately answer with the ongoing study, and the establishment of a cohort of children of the younger participants to examine adolescent diet, activity, and excess weight gain. Each study is made possible by the many compo- nents that already exist, including our data pro- cessing methods, software for data manage- ment and analysis, and, for the study of spouses, a population of women already com- mitted to health research. In 1980, a dietary component was added to the follow-up questionnaire. A food frequency questionnaire was pilot tested among study participants during 1979, and based on the re- sults, 61 food items were selected and included in the follow-up questionnaire mailed to all co- hort members in 1980. For each food, a com- monly used unit or portion (e.g., one egg or slice of bread) was specified, and the women were asked how often, on average over the past year, they had consumed that amount of each food. There were in possible responses, rang- ing from "never" to "six or more times per day." We also inquired in detail about the types of fat used in cooking and at the table and about the use of specific vitamin supplements. Nu- trient intakes were compiled by multiplying [ [ [ [ [ [ [ [ [ [

NURSES' HEALTH STUDY the frequency of consumption of each unit of food by the nutrient content of the specified portions. This food frequency questionnaire has been evaluated extensively for reproducibility and validity. Nutrient intake assessed by this ques- tionnaire was compared with detailed diet records kept by a sample of 194 participants who weighed or measured everything they ate or drank for 4 weeks over the course of a year.9-11 In addition, various nutrients mea- sured in the blood (vitamin E, beta carotene, omega-3 fatty acids) were found to be corre- lated with the questionnaire estimates of in- take.12,13 The instrument's reproducibility was assessed in 1614 women14 and not found to be influenced by obesity or other personal char- acteristics, including cigarette smoking, alcohol intake, or age. These validation studies were crucial in establishing the validity of dietary questionnaires in large-scale studies and re- main the standard for such studies in the field of epidemiology. Since 1980, the food fre- quency questionnaire has been expanded to in- dude approximately 120 individual food items plus vitamin and mineral supplement use that collectively account for >90% of the major nu- trient intakes being measured. This expanded questionnaire was completed by the cohort in 1984, 1986, 1990, and 1994. BIOLOGIC SPECIMENS Toenail samples Because of evidence suggesting that sele- nium may be important in the etiology of can- cer and heart disease, our research group was interested in obtaining selenium exposure lev- els from participants in the cohort. It is not pos- sible to assess selenium intake from a food frequency questionnaire because of high vari- ability of selenium values within specific grains and vegetables as a result of variability in soil selenium content.~5 After validation of the use of toenails as a means to measure body sele- nium levels integrated over an extended pe- riod,16 we invited 92,000 participants to mail a set of 10 toenail clippings following the return of the 1982 follow-up questionnaire. In ~11, 53 68,213 women responded, and their nails are stored in a bank of toenail specimens that have been used in several analyses comparing women who have developed a serious illness during follow-up to a sample of those who have remained free from disease. Blood samples In 1989, the Nurses" Health Study research group was awarded funds to undertake the col- lection of blood specimens from participants in the cohort to address hypotheses related to hor- mone leveis, micronutrients, and risk of breast cancer. The collection of blood specimens was completed over a year-long period. The 1988 questionnaire asked participants if they would be willing to provide a blood sample, and those who indicated yes were sent a blood collection kit. More than .32,000 women participated in this additional phase of the study. Only about 6500 of the participants were premenopausal. Although their blood samples were not col- lected at a specific time in their menstrual cy- cle, the day their current cycle started was recorded on the study questionnaire. The women who provided blood samples were similar to those who did not in terms of both age and body mass index (BMI), although they were slightly less likely to be current cigarette smokers and more likely to be currently using postmenopausal hormones. Samples were sent by overnight delivery to our research labora- tory, where they were centrifuged, labeled, and stored on liquid nitrogen for subsequent nested case-control analyses. We will identify the blood samples from women who subsequently develop breast cancer during follow-up and compare hormone levels with those of controls who have remained free from cancer and are the same age as the women with breast cancer (i.e., a nested case-control analysis). Likewise, when stud3~ing hormones and risk of fractures or antioxidant levels and risk of cataracts or cardiovascular disease, we will identify women who provided blood samples and subsequently developed these conditions and a group of women who remained free from these diseases as a comparison group. Studies among a subset of postmen0pausal participants who provided repeated blood

54 samples over 3 years have shown that a single blood sample, as obtained from the more than 30,000 women, is a good indicator of blood hor- mone levels over at least the previous 3 years.17 Analyses of hormone levels among post- menopausal participants in this study indicate that with increasing levels of obesity (BMI), higher levels of estrogens are present. For es- trone and estrone sulfate, the correlation was 0.37, and for bioavailable estradiol, it was 0.67.is Prolactin was the only hormone analyzed that was unassociated with BMI. Height was unas- sociated with plasma estrogens or prolactin. Alcohol intake was positively associated with estrone sulfate concentration (r = 0.17). These data suggest that the associations of BMI and al- cohol intake with subsequent breast cancer risk might be mediated, at least in part, through in- fluence on postmenopausal estrogen levels. The possibility of exploring the genetic basis for cancer, as well as other diseases, by assess- ing the stored blood for candidate genes is be- ing undertaken. This is a rapidly changing field, and, at present, it is not clear what the full extent of using blood samples from the co- hort will be. QUALITY OF LIFE AND SOCIAL NETWORKS Over the years, participants have indicated to us through letters returned with question- naires that they have concerns beyond the more directly biologic exposures (such as cigarette smoking and menopause) that we routinely record. In response to their concern, in the fall of 1991 we identified a series of questions to as- sess quality of life using the Medical Outcomes Study SF-36,19 work-related stress (including job demands and control),2° caregiving outside of work, retirement, and other measures of so- cial support or social networks.21 In 1992, we included these questions in the initial June mailing to all participants in the study. More than 70,000 women completed and returned questionnaires with these additional items re- lated to quality of life, thus forming the basis for detailed analyses relating these measures to changes over the life course as these women are followed through the next decade. COLDITZ ET AL. Comparing the mean score for each subscale on the SF-36 against the National Opinion Social Survey-General Social Survey,22 we ob- serve that for working women, the mean scores for nurses were quite comparable to U.S. work- ing women in general. The main difference ap- pears to be that the Nurses' Health Study par- ticipants reported ~gher levels of physical functioning compared with the general popu- lation of working women (Table 1). Initial analyses of the work stress questions show that there is substantial variability even within a single occupation. Inpatient and operating room nurses were more likely to be in high- strain jobs. Outpatient nurses in passive jobs and nurse educators were more likely to be in low-strain jobs or active jobs. Thus, although job demands and control do not measure unique aspects of nursing work, these data sug- gest that they are reasonably differentiating nursing work in expected ways. We also note that when we compare role functioning and vi- tality as measured by the SF-36 across cate- gories of job strain, women with high strain have substantially and significantly lower role physical and role emotional functioning and also lower vitality than women in active jobs. MAJOR FINDINGS AND CONTRIBUTIONS TO WOMEN'S HEALTH The major disease-related findings from the study over the first 18 years of follow-up are summarized in Table 2. Here we set forth the TABLE 1. COMPARISON OF SF-36 SCORES ON THE NATIONAL OPINION RESEARCH CEN'rER'S GENERAL SOCIAL SURVEY aND Ttm NURSES" HEALTH STUDY NORC-General Nurses" Health SI~-36 Subscale Social Survey Study Vitality 63.9 (60.7-67.1) 62.8 (62.5-63.0) Role emotional functioning 85.9 (81.2-90.5) 83.9 (83.5-84.2) Mental health 77.6 (75.0-80.3) 75.7 (75.3-75.9) Bodily pain 74.7 (71.2-78.2) 76.6 (76.3-76.8) Physical functioning 85.3 (82.5-88.1) 89.4 (89.2-89.6) Role physical functioning 87.1 (82.1-92.2) 82.0 (81.6-82.4)

NURSES' HEALTH STUDY major lifestyle factors and their relations to ma- jor illnesses among women. For each associa- tion, a citation to the full published report is in- cluded. Of note, the study has also made major contributions to the methods of assessing diet and other lifestyle variables that are now in- corporated into many of the more recently cre- ated cohort studies, both in the United States and elsewhere. Many of these have been sum- marized previously.23 The major new findings reported over the past 2 years include a lack of association be- tween dietary calcium intake among post- menopausal women and risk of osteoporotic fractures. Higher intake of calcium from di- etary sources was not protective against frac- tures of the hip or wrist. In addition, a positive relation was observed between protein intake and risk of fractures. Dairy products high in protein and calcium were not protective against fractures. However, we observed a trend toward lower risks among women who consumed higher levels of milk during adoles- cence. We also reported that calcium intake does not protect against risk of colon cancer24 and that the risk of pancreatic cancer falls rapidly after cessation from cigarette smoking.25 Women who have used OCs for ->5 years have under half the risk of ovarian cancer compared with women who never used OCs.26 Impor- tantly, we have made major contributions to the framing of the revised dietary guidelines for Americans. A series of articles addressed the adverse effects of weight gain during adult life. Women who gained substantial weight af- ter age 18 are at significantly increased risk of coronary heart disease (CHD),27 noninsulin-de- pendent diabetes mellitus,2s and total mortal- ity29 compared with women who remained within 5 pounds of their weight at age 18. Based on these results and an extensive body of liter- ature showing physiologic changes with weight gain, the dietary guidelines now place greater emphasis on avoiding weight gain and state "Balance the food you eat with physical activity. Maintain or improve your weight."3° With regard to the use of postmenopausal hormones, we observed that longer use of hor- mones (->5 years) was associated with in- creased risk of breast cancer incidence and 55 mortality.31 Also, we reported that the addition of progestins to estrogen therapy did not re- duce the risk of breast cancer. Consistent with many other studies, early menopause is asso- ciated with substantially lower risk of breast cancer among women who do not take post- menopausal hormones. Current use of post- menopausal hormones continues to protect women against CHD.32 Within this cohort of women up to age 71, almost three cases of breast cancer are diagnosed for every heart at- tack. Other major findings that may lead to greater prevention of chronic illnesses include a de- crease in risk of colon cancer with longer du- rations of use of aspirin.33 The risk reduction was substantial after ->10 years of use. A de- crease in risk of colon cancer with moderate levels of physical activity also offers an impor- tant avenue for prevention. We reported that coffee drinking is not related to risk of CHD. Women consuming ->6 cups of coffee per day had a relative risk of CHD that was 0.95 (95% CI 0.73-1.26) compared with women who did not consume coffee.34 Suicide is less likely among women as level of coffee intake in- creases.3s Also with regard to risk of CHD, we observed that women who had worked rotat- ing shifts for ->6 years were at increased risk. One important feature of the prospective co- hort design is the ability to study total mortal- ity. With this outcome, we can begin to balance the risks and benefits of lifestyle choices, such as use of OCs, smoking, alcohol consumption, and body weight. Other studies that focus on one disease at a time are typically not able to address these important (and, from an indi- vidual perspective, often difficult) tradeoffs. Recent analyses have shown that use of OCs is not related to any overall increase in mortal- ity,36 that increasing body weight is associated with increased risk of mortality from all causes and separately from CHD and from cancer,29 that smoking is associated with increased mor- tality and that risk is reduced after stopping smoking,37 that alcohol is associated with in- creased mortality among women under age 40, and for women over age 50, light to moderate alcohol intake was associated with significant reduction in mortality.3s Importantly/ death from breast cancer was elevated among women

TABLE 2. MAJOR FINDINGS FROM THE NURSES' HEALTH STUDY, 1976-1996 Breast Ca CHD~/stroke Colon Ca Fracture Diabetes Other diseases Cigarette smoking No relation with Smoking dominant Current Increased risk Increased risk Strong predictor current or past cause of CHD; strong smoking of hip of NIDDMa4a of lung cancer smoking43 dose-response related to fracture suicide,49 and relation44 polyps; strong cataractsSO; relation with risk of total Risk of CHD reduced cancer after mortality for by 14% within 2 30-year latent ex-smokers years of period42 approaches that stopping4s of never smoker after 10-14 Strong relation with years37 ' stroke46 reduced after stopping Smoking smoking~7 cessation Oral contraceptives Current use Current use No increases risks2 increases risksa association54 Past use~little Past use---little association relation Postmenopause hormones Current use for Current use reduces Suggestive >5 years risk of CHD57 decrease in increases risk31 risk of colon cancers4 associated with modest weight gain of about 6 pounds51 No relation with total mortalitya6 No relation with rheumatiod arthritiss6 Decreased risk of ovarian cancer26 No relation with rheumatoid arthritis Progestins added to estrogen therapy do not reduce risk31 Obesity Weak positive relation with incidence among postmenopausal women61 Strong relation, even average weight women at increased risk of CI-ID~2 Weight gain after age 18 associated with increased risk27 Increased risk63 Not examined Reduces risk of hip fracture Strong protection against hip fracture No associationss Not related to risk of NIDDMsa Strong dose- response relation; average weight women at significantly increased risk6~ Increased risk of endometrial cancer Increases risk of systemic lupus erythematosus5~ Increasing risk of cholecystectomy with increasing duration of uses9 Strong relation with gall stones~ and total mortality29

Alcohol Increasing risk Strong inverse Moderate with increasing relation for CHD; intake drinks per positive relation for increases risk day66 subarachnoid of polyps68 hemorrhage67 Diet Low vitamin A Vitamin E protects Red meat intake associated against CHD73 intake with increased increases risk risk7° but no Trans-fatty acids of cancer7s relation for increase risk of CHD74 vitamin C or E Folate intake associated No relation for Coffee consumption with reduced total fat not related to risk risk of polyps68 intake7~ of CHD34 Monounsaturated Calcium intake fat intake not related to inversely related risk of polyps76 or to colon to risk of breast cancer24 cancer No relation with selenium72 Other exposures Atypical 1-6 aspirin per Aspirin use (>-20 hyperplasia week reduces risk years) associated with of CHEPa reduces risk33 increased risk82 Family history accounts for 6% of breast cancer~3 Use of hair dyes not related to risks4 First pregnancy increases risk of breast cancer in short term but decreases risk long term; closer spacing of births associated with lower riskss History of adult- onset diabetes increases risk of CHD and strokes7 Taller women have lower riskss Rotating shift work increases risk of CHIY9 L~98~9890G ,Family history increases risk up to 4-fold among women <509o Leisure time physical activity reduces risk of colon cancer91 Increased risk of hip fracture with moderate intake6~ No reduction in risk with higher dietary calcium intake~V; dietary protein associated with increased risk78 Caffeine intake positively related to risk of hip and forearm fracture~9 Taller women more likely to have hip fractures92 Weight gain after age 18 significantly increases risk2~ Strong inverse dose-resj~onse relation°~ Magnesium intake inversely related to risk79 No relation with fat intake or total carbohydrate Vigorous activity at least once per week reduces risk93 Reduced risk of total mortality among older women~8 Dietary vitamin A intake associated with reduced risk of cataractsso Antioxidant supplements-- no important relation to asthma81 Coffee intake inversely related to suicide35 Tubal ligation halves risk of ovarian cancer93 Number of blistering sun- burns before age 20 positively related to risk of melanoma9s Breast implants not related to risk of connective tissue disease Use of hair dyes not related to risk of hematopoietic cancerssl ~CHD, coronary heart disease; NIDDM, noninsulin-dependent diabetes mellitus.

consuming more than a drink per day, and death from cardiovascular disease was reduced among women with this level of intake. The contribution of genetics to most major chronic diseases remains small. For breast can- cer, for example, perhaps 6%-10% can be at- tributed to inherited genetic factors.39 A simi- lar estimate may prevail for colon cancer and for heart disease. Thus, the study of lifestyle factors acting in the broader population is more useful in identifying areas for prevention in the general population than merely focusing on the high-risk subgroups for specific diseases. With additional follow-up, the numbers of cases available for study has increased, allow- ing application of new biomathematical mod- els to the analysis of breast and lung cancer. These analyses allow us to better understand the interrelationships between particular life- st3fle habits or exposures, the timing of these exposures, and the subsequent risk of cancer.4° Among the methodologic advances made in the Nurses" Health Study, the repeated mea- sures of diet, hormone use, physical activity, body weight, and cigarette smoking have be- come the standard for modern studies among women. Of particular note is the need for re- peated measures when studying behaviors, such as postmenopausal hormone use, with changing products and patterns of use that pre- clude the application of more controlled re- search designs to address risks associated with current prescribing patterns. One concern when interpreting the results from the Nurses' Health Study is their internal validity. This point has been addressed, as noted, through extensive validation of reported lifestyle measures and careful documentation of disease outcomes. Once internal validity is established, issues of generalizability must be considered. The participants are predomi- nantly white women, reflecting the ethnic back- ground of women who trained as registered nurses through the 1960s.41 At entry, their level of cigarette smoking was comparable to U.S. national data for women, and their use of OCs is comparable to that of their birth cohorts.23 Their experiences (and age distribution) of menopause could not conceivably be altered by their training, as registered nurses. Further, their reproductive histories are similar to na- COLDITZ ET AL. tional census data, and correspondingly, their rates of breast cancer are very close to those ex- pected based on the national Surveillance, Epidemiology, and End Results age-specific rates. Data on occupational work stress is com- parable to that of other studies, and the qual- ity of life measures reported in 1992 reflect the patterns observed in the national reference study. Based on data such as these, we con- clude that the cohort reflects the relations be- tween lifestyle and health of white women in general. Thus, although they come from a pro- fessional group, their training in large part serves as an advantage, in all likelihood re- moving socioeconomic and other barriers to ac- cess to health care. If such barriers existed, it would be possible that women with specific lifestyle characteristics may be less likely to be diagnosed with d.isease, not because of the lifestyle but because of their access to care. This factor could then substantially distort relations, giving biased results. Women from ethnic groups other than Caucasian are not well rep- resented in this cohort. Thus, when findings are expected to vary due to some underlying bio- logic difference among ethnic groups, which will be rare, those women will need to be specifically studied. The Nurses' Health Study, the largest and longest ongoing cohort study of lifestyle and health that is focused on women, is a tribute to the commitment of its participants over the 20 years of the study to date and the foresight of Frank Speizer, M.D., who initiated pilot stud- ies for the cohort >24 years ago. Components of the study, such as the collection of blood samples, have been possible only because of the education and professional experience of its participants, in addition to their willingness to give of their time to the research effort. (As a token of appreciation and in recognition of their excellent record of participation, Harvard Medical School and the Brigham and Women's Hospital jointly awarded to each participant a 20th Anniversary Certificate of Appreciation in June 1996.) Long-term follow-up with high participa- tion, as exemplified by the Nurses' Health Study, is essential to providing valid estimates of the relations between lifestyle and. risk of chronic diseases. For cancer, behaviors may act

NURSES' HEALTH STUDY as initiators (cigarette smoking and colon can- cer risk),42 promoters (red meat for colon can- cer), or proliferators (estrogens for breast can- cer). Repeated measures and long-term follow-up permit a detailed understandin~ of these relations. The detailed history of use of vitamin supplements, at different doses and over varying durations, adds a richness to the data that cannot be obtained through a ran- domized study in which the dose and duration are predetermined (and are perhaps more or less than ideal). This unique feature will allow the identification of the dietary and other lifestyle factors that are most beneficial to women in terms of morbidity, mortality, and quality of life as the study continues through the coming years. Through the Nurses' Health Study, many im- portant advances have been made in under- standing the etiology and prevention of major illnesses among women. Before the Women's Health Initiative began, the Nurses" Health Study study was the largest and most compre- hensive study of health among women. It re- mains the most detailed study of diet and ma- jor illnesses, providing details on many components of lifestyle updated over the years. This unique study will continue to shed light on the causes and prevention of disease and the fea- tures of healthy aging over the coming years. Needless to say, this all reflects the enormous contribution made by over 120,000 registered nurses who entered the study 20 years ago. ACKNOWLEDGMENTS The continuing commitment of the study participants is g(atefully acknowledged. The investigators on the Nurses' Health Study have made major contributions over the years. Gur- rent members of the study team include Cath- erine Berkey, Celia Byrne, Carlos Camargo, Vincent Carey, Gary Chase, Graham Colditz, Karen Corsano, Gary Curhan, Barbara Egan, Diane Feskanich, Lindsay Frazier, Charles Fuchs, Edward Giovannucci, Francine Grodstein, Susan Hankinson, Charles Hennekens, David Hunter, JoAnn Manson, Stefanie Parker, Cathy Rexrode, Janet Rich-Edwards, Bernard Rosner, Caren Solomon, Meir Stampfer, Harry Taplin, 59 Waiter Willett, and Frank E. Speizer (principal investigator). REFERENCES 1. Barton J, Bain C, Hennekens CH, et al. Characteristics of respondents and non-respondents to a mailed questionnaire. Am J Public Health 1980;70:823. 2. Stampfer M-J, Willett WC, FE, et al. Test of the National Death Index. Am J Epidemiol 1984;119:837. 3. Rimm EB, Stampfer MJ, Colditz G, Giovannucci E, Willett WC. Effectiveness of various marling strate- gies among nonrespondents in a prospective cohort study. Am J Epidemiol 1990;131:1068. 4. Rose GA, Blackburn H. Cardiovascular survey meth- ods. WHO Monograph Series No. 58. Geneva: World Health Organization, 1982. 5. Walker AE, Robins M, Weinfeld FD. The National Survey of Stroke. Clinical findings. Stroke 1981;12(2 Part 2 Suppl 1):I13. 6. Doll R, Hill A. Lung cancer and other causes of death in relation to smoking. A second report on the mor- tality of British doctors. Br Med J 1956 (November 10):1071. 7. Colditz GA, Martin P, Stampfer MJ, et al. Validation of questionnaire information on risk factors and dis- ease outcomes in a prospective cohort study of women. Am J Epidemiol 1986;123:894. 8. Giovannucci E, Tosteson T, Speizer F, Vessey M, Colditz G. A long-term study of mortality in men who have undergone vasectomy. N Engl J Med 1992;326: 1392. 9. Willett WC, Sampson L, Stampfer MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985; 122:51. 10. Hunter DJ, Sampson L, Stampfer M~, Colditz GA, Rosner B, Willett WC. Variability in portion sizes of commonly consumed foods among a population of women in the United States. Am J Epidemio11988;127: 1240. 11. Salvini S, Hunter DJ, Sampson L, et al. Food-based validation of a dietary questionnaire: The effects of week-to-week variation in food consumption. Int J Epidemiol 1989;18:858. 12. Wiliett WC, Stampfer MJ, Underwood BA, Speizer FE, .Rosner B, Hennekens CH. Validation of a dietary questionnaire with plasma carotenoid and alpha-to- copherol levels. Am J Clin Nutr 1983;38:631. 13. Willett WC. Nutritional epidemiology. New York: Oxford University Press, 1990. 14. Colditz GA, Willett WC, Stampfer MJ, et al. The in- fluence of age, relative weight, smoking, and alcohol intake on the reproducibility of a dietary question- naire. Int J Epidemiol 1987;16:392. 15. Ullrey DE. Selenium in the soil-plant-food chain. In: Spallhotz JE, et al., eds. Selenium in biology and med- icine. Westport, CT: AVI Publishing, 1981:176.

6O 16. Morris JS, Stampfer MJ, Willett WC. Dietary selenium in humans: Toenails as an indicator. Boil Trace Ele- ment Res 1983;5:529. 17. Hankinson S, Manson J, Speigelman D, Willett W, Longcope C, Speizer F. Reproduci.bility of plasma hormone levels in postmenopausal women over a 2-3 year period. Cancer Epidemiol Biol Prey 1995;4:649. 18. Hankinson SE, Willett WC, Manson JE, et al. Alcohol, height, and adiposity in relation to estrogen and pro- lactin levels in postmenopausal women. J Nat Cancer Inst 1995:1297. 19. Ware J, Snow K, Kosinski M, Gandek B. SF-36. Manual and interpretation guide. Boston: The Health Institute, New England Medical Center, 1993. 20. Karasek R, Theorell T. Health work: Stress, produc- tivity, and the reconstruction of working life. New York: Basic Books, 1990. 21. Berkman LF, Syme SL. Social networks, host resis- tance and mortality: A nine-year follow-up study of Alameda County residents. Am J Epidemio11979;109: 186. 22. Thalji L, Haggerty CC, Rubin R, Berckmans TR, .Pardee BL. National survey of junctional health sta- tus: Final report. Chicago: National Opinion Research Center, 1991. 23. Colditz GA. The Nurses' Health Study: Findings dur- ing 10 years of follow-up of a cohort of US women. Curt Probl Obstet Gynecol Fertil 1990;13:129. 24. Martinez M, Giovannucci E, Colditz G, et al. Calcium, vitamin D, and the occurrence of colorectal cancer among women. J Natl Cancer Inst 1996;88:1375. 25. Fuchs C, Colditz G, Stampfer M, et al. A prospective study of cigarette smoking and the risk of pancreatic cancer. Arch Intern Med 1996;156:2255. 26. Hankinson S, Colditz G, Hunter D, et al. A prospec- tive study of reproductive factors and risk of epithe- lial ovarian cancer. Cancer 1995;76:284. 27. Willett WC, Manson JE, Stampfer MJ, et al. Weight, weight change, and coronary heart disease in women: Risk within the "normal" weight range. JAMA 1995; 273:461. 28. Colditz GA, Willett WC, Rotnitzky A, Manson JE. Weight gain as a risk factor for clinical diabetes in women. Ann Intern Med 1995;122:481. 29. Manson JE, Willett WC, Stampfer MJ, et al. Body weight and mortality among women. N Engl J Med 1995;333:677. 30. U.S. Department of Agriculture. Repo.rt of the dietary guidelines advisory committee on the dietary guide- lines for Americans, 1995. To the Secretary of Health and Human Services and the Secretary of Agriculture. Washington, DC: U.S. Department of Agriculture, 1995. 31. Colditz GA, Hankinson SE, Hunter DJ, et al. The use of estrogens and progestins and the risk of breast can- cer in postmenopausal women. N Engl J Med t995; 332:1589. 32. Grodstein F, Stampfer M, Manson J, et al. Post- menopausal estrogen and progestin use and the risk of cardiovascular disease. N Engl J Med 1996;335:453. COLDITZ ET AL. 33. Giovannucci E, Egan KM, Hunter DJ, et al. Aspirin and the risk of colorectal cancer in women. N Engi J Med 1995;333:609. 34. Willett W, Stampfer M, Manson J, et al. Coffee con- sumption and coronary heart disease in women: A ten-year follow-up. JAMA 1996;275:458. 35. Kawachi I, Willett W; Colditz G, Stampfer M, Speizer F. A prospective study of coffee drinking and suicide in women. Arch Intern Med 1996;156:521. 36. Colditz G, for the NHS Research Group. Oral contra- ceptive use and mortality during twelve years of fol- low-up: The Nurses' Health Study. Ann Intern Med 1994;120:821. 37. Kawachi I, Colditz GA, Stampfer MJ, et al. Smoking cessation in relation to total mortality rates in women: A prospective cohort study. Ann Intern Med 1993; 119:992. 38. Fuchs CS, Stampfer M-J, Colditz GA, et al. Alcohol consumption and mortality among women. N Engl J Med 1995;332:1245. 39. King M-C, Rowell S, Love SM. Inherited breast and ovarian cancer. What are the risks? What are the choices? JAMA 1993;269:1975. 40. Rosner B, Colditz G. Extended mathematical model of breast cancer inddence 'in the Nurses" Health Study. J Natl Cancer Inst 1996;88:359. 41. Roth A, Graham D, Schmittling G. National sample survey of registered nurses (1977). A report on the nurse population and factors affecting their supply. Hyattsville: Public Health Service, 1977. 42. Giovannucd E, Colditz GA, Stampfer MJ, et al. A prospective study of cigarette smoking and risk of col- orectal adenoma and colorectal cancer in U.S. women. J Natl Cancer Inst 1994;86:192. 43. London SJ, Colditz GA, Stampfer MJ, Willett WC, Rosner BA, Speizer FE. Prospective study of smoking and the risk of breast cancer. J Natl Cancer Inst 1989; 81:1625. 44. Willett WC, Green A, Stampfer MJ, et al. Relative and absolute excess risks of coronary heart disease among women who smoke cigarettes. N Engl J Med 1987;317: 1303. 45. Kawachi I, Colditz G, Stampfer M, et al. Smoking ces- sation and time course of decreased risk of coronary heart disease in women. Arch Intern Med 1993; 154:169. 46. Colditz GA, Bonita R, Stampfer MJ, et al. Cigarette smoking and risk of stroke in middle-aged women. N Engl J Med 1988;318:937. 47. Kawachi I, Colditz G, Stampfer M, et al. Smoking ces- sation and decreased risk of stroke in women. JAMA 1993;269:232. 48. Rimm EB, Manson JE, Stampfer MJ, et al. Cigarette smoking and the risk of diabetes in wombn. Am J Public Health 1993;83:211. 49. Hemenway D, Sondick S, Colditz G. Smoking, sui- cide, and nurses. Am J Public Health 1993;83:249. 50. Hankinson SE, Willett WC, Colditz GA, et al. A prospective study of smoking and risk of. cataract surgery in women. JAMA 1992;268:994. cO O~ 0

NURSES' HEALTH STUDY 51. Kawachi I, Troisi R, Rotnitzky A, Coakley E, Speizer F, Colditz G. A prospective study of smoking cessa- tion and weight change in women. J Smoking Rel Dis 1994;5(Suppl 1):91. 52. Romieu L Willett WC, Colditz GA, et al. A prospec- tive study of oral contraceptive use and the risk of breast cancer in women. J Natl Cancer Inst 1989;81: 1313. 53. Stampfer MJ, Willett WC, Colditz GA, Speizer ICE, Hennekens CH. A prospective study of past use of oral contraceptive agents and risk of cardiovascular diseases. N Engl J Med 1988;319:1313. 54. Chute CG, Willett WC, Colditz GA, Stampfer MJ, Rosner B, Speizer IrE. A prospective study of repro- ductive history and exogenous estrogens on the risk of colorectal cancer in women. Epidemiology 1991;2: 201. 55. Rimm E, Manson J, Stampfer M, et al. Oral contra- ceptive use and the risk of non-insulin-dependent di- abetes mellitus in a large prospective study of women. Diabetologia 1992;35:967. 56. Hernandez-Avfla M, Liang MH, Willett WC, et al. Exogenous sex hormones and the risk of rheumatoid arthritis. Arthritis Rheum 1990;33:947. 57. Stampfer MJ, Colditz GA, Willett WC, et al. Postmenopausal estrogen therapy and cardiovascular disease. N Engl J Med 1991;325:756. 58. Manson J, Rimm E, Colditz G, et al. A prospective study of postmenopausal estrogen therapy and sub- sequent incidence of noninsulin-dependent diabetes mellitus. Ann Epidemiol 1992;2:665. 59. Sanchez-Guerrero J, Liang M, Karlson E, Hunter D, Colditz G. Postmenopausal estrogen therapy and risk of developing systemic lupus erythematosus. Ann Intern Med 1995;122:430. 60. Grodstein F, Colditz G, Stampfer M. Postmenopausal hormone use and cholecystectomy in a large prospec- tive study. Obstet Gynecol 1994;83:5. 61. London SJ, Colditz GA, Stampfer MJ, Willett WC, Rosner B, Speizer FE. Prospective study of relative weight, height and the risk of breast cancer. JAMA 1989;262:2853. 62. Manson JE, Colditz GA, Stampfer MJ, et al. A prospec- tive study of obesity and risk of coronary heart dis- ease in women. N Engl J Med 1990;332:882. 63. Chute C, Willett W, Colditz G, et al. A prospective study of body mass, height, and smoking on the risk of colorectal cancer in women. Cancer Causes Control 1991;2:117. 64. Colditz Ga, Willett WC, Stampfer MJ, et al. Weight as a risk factor for clinical diabetes in women. Am J Epidemiol 1990;132:501. 65. Maclure KM,-Hayes KC, Colditz GA, Stampfer MJ, Speizer FE, Willett WC. Weight, diet and risk of symp- tomatic gallstones in middle-aged women. N Engl J Med 1989;321:563. 66. Willett WC, Stampfer MJ, Colditz GA, Rosner BA, Hennekens CH, Speizer FE. Moderate alcohol con- sumption and the risk of breast cancer. N Engl J Med 1987;316:I174. 61 67. Stampfer MJ, Colditz GA, Willett WC, et al. A prospective study of moderate alcohol drinking and risk of diabetes in women. Am J Epidemiol 1988; 128:549. 68. Giovannucci E, Stampfer MJ, Colditz GA, et al. Folate, methionine and alcohol intake and risk of colorectal adenoma. J Natl Cancer Inst 1993;85:875. 69. Hernandez-Avila M, Colditz GA, Stampfer MJ, Rosner B, Speizer FE, Willett WC. Caffeine, moderate alcohol intake and risk of fractures of the hip and fore- arm among middle-aged women. Am J Clin Nutr 1991;54:157. 70. Hunter DJ, Manson JE, Colditz GA, et al. A prospec- tive study of intake of vitamins C, E and A and risk of breast cancer. N Engl J Med 1993;329:234. 71. Willett WC, Hunter DJ, Stampfer MJ, et al. Dietary fat and fiber in relation to risk of breast cancer. An eight- year follow-up. JAMA 1992;268:2037. 72. Hunter DJ, Morris JS, Stampfer MJ, Colditz GA, Speizer FE, Willett WC. A prospective study of sele- nium status and breast cancer risk. JAMA 1990; 264:1128. 73. Stampfer MJ, Hennekens CH, Manson JE, Colditz GA, Rosner B, Willett WC. A prospective study of vitamin E consumption and risk of coronary disease in women. N Engl J Med 1993;328:1444. 74. Willett WC, Stampfer MJ, Manson JE, et al. Trans-fatty acid intake in relation to risk of coronary heart dis- ease among women. Lancet 1993;341:581. 75. Willett WC, Stampfer MJ, Colditz GA, Rosner BA, Speizer irE. Relation of meat, fat and fiber intake to colon cancer risk in a prospective study among women. N Engl J Med 1990;323;1664. 76. Kampman E, Giovannucci E, van't Veer P, et al. Calcium, vitamin D, dairy foods and the occurrence of colorectal adenomas among men and women in two prospective studies. Am J Epidemiol 1994;139:16. 77. Feskanich D, Colditz G, Stampfer M, Willett W. Dietary calcium and bone fractures in middle-aged women. [Abstract] Am J Epidemiol 1994;139:$35. 78. Feskanich D, Willett W, Stampfer M, Colditz G. Protein consumption and bone fractures in women. Am J Epidemiol 1996:143:472. 79. Colditz G, Manson J, Stampfer M, Rosner B, Willett W, Speizer F. Diet and risk of clinical diabetes in women. Am J Clin Nutr 1992;55:1018. 80. Hankinson SE, Stampfer MJ, Seddon JM, et al. Nutrient intake and cataract extraction in women: A prospective study. Br Med J 1992:305:335. 81. Troisi R, Willett W, Weiss S, Trichopoulos D, Rosner B, Speizer F. A prospective study of diet and adult- onset asthma. Am J Respir Crit Care Med 1995;151: 1401. 82. London SJ, Connolly JL, Schnitt SJ, Colditz GA. A prospective study of benign breast disease and the risk of breast cancer. JAMA 1992:267:91. 83. Colditz GA, Willett WC, Hunter DJ, et al. Family his- tory, age and risk of breast cancer: Prospective data from the Nurses' Health Study. JAMA 1993;270:338. " 84. Green A, Willett WC, Colditz GA, et al. Use of per-

62 manent hair dyes and risk of breast cancer. J Natl Cancer Inst 1987;79:253. 85. Rosner B, Colditz GA, Willett WC. Reproductive risk factors in a prospective study of breast cancer: The Nurses" Health Study. Am ] Epidemiol 1994;139:819. 86. Manson J, Stampfer M, Colditz G, et al. A prospec- tive study of aspirin use and primary prevention of cardiovascular disease in women. JAMA 1991;266: 521. 87. Manson JE, Colditz GA, Stampfer MJ, et al. A prospec- tive study of maturity-onset diabetes mellitus and risk of coronary heart disease and stroke in women. Arch Intern Med 1991;151:1141. 88. Rich-Edwards J, Manson J, Stampfer M, et al. Height and the risk of cardiovascular disease in women. Am ] Epidemiol 1995;142:909. 89. Kawachi I, Colditz G, Stampfer M, et al. Prospective study of shift work and risk of coronary heart disease in women. Circulation 1995;92:3178. 90. Fuchs CS, Giovannucci EL, Colditz GA, Hunter DJ, Speizer FE, Willett WC. A prospective study of fam- ily history, age, and diet and colorectal cancer. N Engl J Med 1994;331:1669. 9~. Martinez ME, Giovannucci E, Spiegelman D, et al. COLDITZ ET AL. Physical activity, body size, and colorectal cancer in women. Am J Epidemiol 1996;143:$73. 92. Hemenway D, Feskanich D, Colditz G. Body height and hip fracture: A cohort of 90,000 women. Int J Epidemiol 1995;24:783. 93. Manson JE, Rimm EB, Stampfer MJ, et al. A prospec- tive study of physical activity and the incidence of non-insulin-dependent diabetes mellitus in women. Lancet 1991;338:774. 94. Hankinson SE, Hunter DJ, Colditz GA, et al. Tubal ligation, hysterectomy and risk of ovarian cancer: A prospective study. JAMA 1993;270:2813. 95. Weinstock MA, Colditz GA, Willett WC, et al. Non- familial cutaneous melanoma incidence in women is associated with sun exposure before 20 years of age. Pediatrics 1989;84:199. Address reprint requests to: Graham A. Colditz, M.D. Nurses" Health Study Channing Laboratory 181 Longwood Avenue Boston, MA 02115