Epidemiology of Breast Cancer

Transcription

1 1480 Epidemiology of Breast Cancer Findings from the Nurses Health Study Graham A. Colditz, M.B.B.S., Dr. PH. Background. The epidemiology of breast cancer was reviewed in the context of hormonal, hereditary, histologic, and dietary risk factors. Methods. Literature review. Results. Late age at menarche and early age at first birth decrease the risk of breast cancer as does an early age at menopause. These risk factors relate to the lifetime exposure of the breast tissue to ovarian hormones. Although an early first birth is associated with a transient increase in the risk of breast cancer, perhaps as a result of the breast s exposure to high levels of hormones before terminal differentiation, in older women, parity is associated with a decreased risk of breast cancer. Among postmenopausal women, obesity is associated with higher levels of estrogens and an increased risk of breast cancer. Within the strata of breast cancer stages at diagnosis, obesity is associated with increased mortality, again supporting the influence of endogenous estrogens on this disease s incidence, recurrence, and survival rates. Consistent with these relationships, current use of estrogen therapy among postmenopausal women is associated with an increased risk of breast cancer. A family history of breast cancer is associated with approximately a twofold increase in the risk of breast cancer, and this risk is greater if the diagnosis was made when the woman s mother was young, although even a diagnosis in an older mother is associated with an increased risk in her daughters. The follow-up of women with a history of benign breast biopsy results shows that atypical hyperplasia is associated with a fourfold increase in risk compared with a biopsy specimen without proliferative changes. Atypia doubles the risk. These data support the concept of atypia as a precursor lesion for breast cancer and may warrant its use as a marker in further studies. Consistent Presented at the National Conference on Gynecologic Cancers, Orlando, Florida, April 2-4, Fromthe Channing Laboratory, Brigham and Women s Hospital; the Harvard Medical School; and the Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts. Supported by research grants CA40356 and CA50385 from the National Cancer Institute, Bethesda, Maryland, and an American Cancer Society faculty research award FRA-398, Atlanta, Georgia. Address for reprints: Graham A. Colditz, M.B.B.S., Dr. PH., Harvard Medical School, 180 Longwood Avenue Boston, MA Accepted for publication September 2, data from retrospective and prospective studies show a positive association between moderate alcohol intake and the risk of breast cancer. This may reflect the increase in estrogen levels observed among women who consume alcohol. Data from prospective studies do not support a relationship between dietary fat intake and the risk of breast cancer either in premenopausal or postmenopausal women. Conclusions. Few of these associations offer the potential for intervention to reduce the breast cancer risk. Cancer 1993; 71: Key words: breast cancer, epidemiology, risk factors, diet, prevention. Breast cancer is a major public health problem of affluent countries. Based on American incidence rates and 1987 mortality rates, 12% of women will have breast cancer diagnosed during their expected lifetimes, and 3.5% will die of this disease. S2 Among American women years of age, breast cancer is the leading cause of death.2 The impact of breast cancer is magnified compared with other cancers because women are at risk beginning in their middle years. As a consequence, the average years of life lost by those with breast cancer (20 years) is higher than the average for all cancers combined (16 years). The rates of breast cancer increase rapidly up to the time of menopause. After menopause, the incidence rates continue to increase with age but at a slower rate of increase than before menopause. In less affluent parts of the world and in the Far East, the same pattern of increase with age is seen,3 but the absolute rates are much lower at each age. For example, in Japan, the overall breast cancer incidence has been only approximately 20% of that in the United state^.^ The incidence of breast cancer has been steadily rising since formal registration of this tumor began in the 1930s. From , the age-standardized incidence rate has risen by an average of 1.2% per year in Connecticut, the state with the longest continuous cancer regi~tration.~ In both industrialized and developing countries, similar long-term increases are being ob-

2 Epidemiology of Breast Cancer/Colditz 1481 Age-adjusted breast cancer mortality rates, by contrast with incidence, have been much more stable in the United States. However, the time trends appear to vary depending on the patient's age at diagnosis; mortality rates have increased since 1950 by approximately 15% among women over age 55 years and declined by approximately the same amount among those younger than 45 years.' Since 1975, the mortality rates for black women have increased substantially and, currently, are slightly higher than for white women.9 The relatively constant overall mortality rate, despite increases in incidence, could be the result of more complete reporting of incident cases, increases in a more benign form of disease, earlier detection, or advances in treatment. These factors all appear to contribute to the divergence of the incidence and mortality rates. Whether the increase in breast cancer incidence during recent years is the result of more widespread screening mammography has been examined in several analyses. The initiation of a screening program temporarily will increase the incidence detected by advancing the time of diagnosis, as was noticed nationally in when several prominent women had their breast cancers diagnose. In 1974, the upward trend in the rate of breast cancer was interrupted by a 1-year increase of almost 15%. If screening is not repeated, a deficit of incident cases will ensue. If screening is maintained at a regular interval, a new steady-state incidence will be achieved at a rate close to that which would occur without screening. The number of breast cancers diagnosed in screening programs that eventually would not be recognized clinically appears to be small; there is minimal underdetection of breast cancer in autopsy series. No excess incidence over 10 years was seen in a randomized screening trial," and little increase was seen among women undergoing mammography for routine screening in a national breast cancer detection program." Because screening causes, at most, a transient rise in incidence and was not widespread at least through the early 1980s, its effect can explain little of the long-term increase in breast cancer incidence. Between 1980 and 1987, the age-adjusted breast cancer incidence increased from per 100,000 women." This recent upsurge in breast cancer incidence is almost entirely related to an increase in the detection of tumors less than 2 cm in diameter; the incidence rate of tumors 2 cm or larger has not changed appre~iably.~ These findings and an improved 2-year survival rate are compatible with the substantial increase in screening mammography that occurred during this peri~d.~ To the extent that the recent acceleration in breast cancer incidence represents the transient rise expected in the early stages of a screening program, this eventually will result in the prevention of deaths of breast cancer during this decade. However, the incidence of larger tumors and those with regional or distant metastases at diagnosis has not de~reased,~ which would have been expected if a screening program were implemented and the true incidence were constant. Several attempts to model the changing incidence of breast cancer suggest that approximately 25% of the increase in rates since 1980 cannot be attributed to ~creening,'~,'~ indicating that the underlying long-term increase in breast cancer incidence has continued through the 1980s and suggesting that no major decline in mortality rates should be expected in the near future. Stable mortality rates in the face of an apparently true increase in incidence suggest that the earlier detection of cases in more recent years, and possibly improvements in treatment, have improved survival rates sufficiently to off - set the rising incidence. Although the recent surge in breast cancer incidence may be related largely to increased mammographic screening, the longer term increase over the last half century appears to be real. The breast cancer rate clearly is increasing, especially among postmenopausal women, and will require even greater attention by researchers and clinicians. In particular, specific factors that explain the long-term increase should be sought. Previous Findings From the Nurses' Health Study on Breast Cancer In this section, I review the relationship between relative weight and breast cancer, studies of oral contraceptive agents, postmenopausal hormones, hair dyes, cigarette smoking, and the role of diet as causes of breast cancer. Recent findings relating the histopathologic findings of benign breast biopsy specimens and the risk of breast cancer are discussed, along with models of breast cancer incidence that correlate with the reproductive risk factors to incidence. Finally, some implications of these findings for prevention will be presented. In the Nurses' Health Study cohort, the established reproductive risk factors for breast cancer carry with them the expected associations showing an increased risk for nulliparous women, a maternal history of breast cancer,15 sister history of breast cancer, a personal history of benign breast disease, later age at first birth,i6 and early age at menarche." A family history of breast cancer increases the risk by approximately 80% over that of women without a first-degree relative who has the disease. The risk is greater if the disease is bilateral among family members.'' In our prospective data during 12 years of follow-up, the excess risk declined with advancing age. For a woman whose mother had unilateral disease after the age of 60 years, the excess risk was approximately 40% greater than that in women with no

3 1482 CANCER Supplement February 15, 1993, Volume 71. No. 4 first-degree relatives with breast cancer. Although a range of genetic markers are being pursued among families at very high risk, only a small amount of the incidence of breast cancer can be explained by these genetic factors. In an analysis based on 4 years of follow-up, we noticed a possible variation in the effect of parity according to age, with a stronger protective effect among older women. This effect persisted with additional follow-up, and Dr. Rosner and I currently are exploring this concept using data accrued through In this new analysis, we have extended the Pikez0 model in which risk factors for breast cancer are postulated to influence the rate of breast tissue aging and the risk of breast cancer is expressed as a function of this tissue age. We added terms for births after the first birth and showed that this gives a good fit to the observed data. Specifically, the rate of tissue aging decreased at the time of first birth and with subsequent births. Furthermore, at menopause there is a one-time decrease in breast tissue age. First birth was associated with a onetime increase in breast tissue age, consistent with the reports from other epidemiologic studies and mortality data indicating that married premenopausal women have higher rates of breast cancer than unmarried or nulliparous premenopausal women. Adiposity Obesity has been related to risk of breast cancer in some, but not all, studies. Inverse relations have been reported among premenopausal women,23 and modest positive associations were seen among postmenopausal women.24 Analyzing the relationship between an index of obesity in 1976 (Quetelet s index computed as weight divided by height to the second power) and the risk of breast cancer during the first 4 years of follow-up in the Nurses Health Study, we observed that the incidence of breast cancer among premenopausal women decreased with higher levels of relative weight.25 For women in the highest level of Quetelet s index (mean weight, 80.4 kg) compared with those in the lowest (mean weight, 52.7 kg), the relative risk of breast cancer during 4 years of follow-up was This represents a 34% lower risk among obese women. The trend for premenopausal women, between lower risk and increasing relative weight, was statistically significant (chi-square, -2.82; P = 0.005). Among women who have undergone a natural menopause, we observed no relationship between relative weight and the risk of breast cancer, although when the analysis was limited to women who have never used Hormone-replacement therapy, a positive relationship was observed. None of the established risk factors modified these relationships. Among premenopausal and postmenopausal women, thinner women tended to have smaller tumors. Thus, we considered the possibility that the observed association between relative weight and breast cancer risk might be the result of a bias resulting from easier, and thus earlier, detection in thin women. Considering the potential benefit of mammography and physician examination and allowing for earlier diagnosis, we estimated that the relative risks would not be changed. Overall, our findings suggest that adult adiposity inversely is related to the risk of premenopausal breast cancer, and we could not exclude a modest increase in the risk of breast cancer among postmenopausal women. Data from a large follow-up study of 570,000 Norwegian women indicate that, among postmenopausal women, adiposity is associated with an increased risk of breast cancer incidence.26 Based on more than 8000 cases, he observed a relative risk of 1.18 for women in the top quintile of body mass index compared with those in the bottom quintile. In this study, adiposity at diagnosis was associated with increased mortality, a finding that has been confirmed in other studies that provide detailed information on other breast cancer risk factors. Overall, obese women before diagnosis have a 20% increase in mortality rate, even within the same stratum of breast cancer stage at diagnosis. This again may reflect the impact of estrogen on the recurrence of tumors, although delayed diagnosis of recurrences also might contribute to this increased mortality rate. The importance of estrogen in the cause and progression of breast cancer is supported, however, by the combined results from the trials of tamoxifen therapy (a competitive inhibitor of estrogen binding) that show a 20% reduction in mortality rate among women who have been treated with this form of therapy.28 These data, combined with evidence that reproductive events such as early age at menarche and late age at menopause increase the risk of breast cancer, indicate that the estrogenic milieu is important in the initiation and progression of breast cancer lesions. Oral Contraceptives Although most epidemiologic studies suggest that there is no association between the use of oral contraceptive agents and the risk of breast cancer,29 others suggest that particular subgroups of oral contraceptive users may be at increased risk.30 To examine the possible association between the use of oral contraceptive agents and the risk of breast cancer further, we analyzed our data at the end of 4 years of follow-up and, again, after 10 years of follow-up.

4 Epidemiology of Breast CancerlCo1dit.z 1483 Details on prior use of oral contraceptives were included on the 1976 questionnaire. Based on these responses, women were categorized as never users, current users, and past users. Within these categories of use, the total duration of use was calculated from the data provided by the participants in the Nurses Health Based on almost 1 million person-years of follow-up, we observed that past use of oral contraceptive agents was not related to an increased risk among premenopausal or postmenopausal women. However, current use was associated with a significant (53%) increase in risk. Because of concerns regarding long-term use before first pregnancy and the possible increase in the risk of breast cancer, we specifically addressed this possibility using the data available to us from the Nurses Health Study. Among parous women, 10.3% had received oral contraceptive agents before their first full-term pregnancy. Overall, we observed no increased risk with the first use of these drugs before or after their first pregnancy lasting 6 months or more among premenopausal and postmenopausal parous women. For premenopausal parous women, the age-adjusted relative risk of ever using oral contraceptives before the first birth was 1.16 (95% confidence interval, ). For postmenopausal parous women, it was 0.58 (95% confidence interval, ). We observed no evidence of increased breast cancer risk with increasing duration of oral contraceptive use before the first pregnancy among parous premenopausal women. For women receiving oral contraceptives for 36 or more months before their first pregnancy, the relative risk was 0.95 (95% confidence interval, ) compared with parous never users. These analyses, based on small numbers, have broad confidence intervals and do not exclude the modest positive associations reported in several recent casecontrol studies. To address this relationship with more confidence, we recruited a new cohort of 116,000 women years of age in These women provided extensive details of their contraceptive history, including their age at use, the brand of each form of oral contraceptive agent, and their age at each pregnancy lasting less than 6 months or 6 months or more. In a metaanalysis, past use of oral contraceptive agents was not related to the risk of breast cancer.32 However, the combined data from published studies indicate that the risk of breast cancer increases with the duration of use before the woman s first pregnancy, and a significant increase in risk was observed among premenopausal women with longer durations of use. Combined data from four studies suggest that this increased risk might be as great as 70% for 4 years or more of use before first pregnancy (relative risk, = 1.72; 95% confidence interval, ). Postmenopausal Hormones Long-term use of postmenopausal hormones were suggested to be associated with an increased risk of breast cancer.29 In a metaanalysis limited to data from casecontrol studies, it was estimated that, after 15 years of use, a 30% increase in the risk of breast cancer would be observed.33 This result contrasted with the prospective data from the Nurses Health Study where we observed no increase in risk with increasing duration of use among current users of hormone-replacement therapy among 1050 patients with invasive breast cancer documented during 12 years of follow-up. Furthermore, we observed no increase in risk among past users, even those women who had used hormone therapy for 10 or more years. Among current users, we observed a 36% increase in risk compared with never users.34 This risk was found among women with and without a family history of breast cancer and was not altered materially by other breast cancer risk factors. Because it has been postulated that the addition of progestins to estrogen therapy may reduce the risk of breast cancer among postmenopausal women,35 we specifically addressed this relationship and observed that the risk was not reduced among women who used this combination therapy. Women who used conjugated estrogens had a relative risk of 1.42 (95% confidence interval, ), and women using estrogen and progestin had a relative risk of 1.54 (95% confidence interval, ). Permanent Hair Dyes The aromatic amines contained in permanent hair dyes can be absorbed percutaneously and are mutagenic and carcinogenic in some laboratory studies.36 Because of such evidence and some case-control studies among hairdressers, concern has been raised whether the use of hair dyes is associated with the development of cancer in women. We examined this possible association using data collected prospectively from participants in the Nurses Health Study who provided information on use of permanent hair dyes in 1976, 1978, and Incident cases of breast cancer diagnosed during 6 years of follow-up were included in the analysis. Among women who had used permanent hair dyes, breast cancer developed in 353 during 246,848 personyears of follow-up, and 505 cases occurred during 397,460 person-years of follow-up among never users (age-adjusted relative risk, 1.l; 95% confidence interval, ). The relative risk for current users was 1.1 as was the relative risk for past users. The risk of breast cancer did not increase with more frequent use, longer duration of use, or interval since first use. Compared with never users, women using permanent hair dyes

5 1484 CANCER Supplement February 25, 2993, Volume 71, No. 4 once every 1-4 weeks had a relative risk of 1.2 (95% confidence interval, ). For those using permanent hair dyes once every weeks, the relative risk was 1.2 (95% confidence interval, ). For less frequent use, the relative risk was 0.9 (95% confidence interval, ). Cigarette Smoking Cigarette smoking reduced the age at menopause as discussed previously. Furthermore, it was associated with altered metabolism of estrogen^.^' Thus, a protective effect on the risk of breast cancer has been proposed. The results reported from previous studies showed an increased risk among premenopausal women but no overall relationship with breast cancer. In an analysis that included 1799 incident cases of breast cancer diagnosed between the time of return of the 1976 questionnaire and June 1, 1986, we observed no relationship between cigarette smoking and the risk of either premenopausal or postmenopausal breast cancer.39 Tumor size and the presence of nodal metastases at the time of diagnosis were unrelated to cigarette smoking. These results suggest that smoking and breast cancer are not related materially. Breast Feeding An association between lactation and protection from breast cancer has been postulated for a long time. We examined the relationship between lactation and the risk of breast cancer among 89,413 parous women who provided details on their total duration of lactation in From ,1262 cases of breast cancer occurred during 785,958 person-years of follow-up. We observed no independent association between lactation and the risk of breast cancer.40 After adjusting for age and parity, compared with women who never lactated, the relative risks (and 95% confident intervals) were: 0.95 ( ) for less than 7 months of lactation, 0.87 ( ) for 7-11 months, 0.94 ( ) for months, and 0.98 ( ) for 24 months or more (test for trend, -0.80; P = 0.42). The association did not differ according to age or menopausal status. Our results do not confirm reports that relatively short durations of lactation reduce the risk of breast cancer among younger women. Dietary Fat Large international differences in the rates of breast cancer and striking increases among populations migrating from low to high incidence areas suggest that environmental factors, probably diet, may have a strong influence on the occurrence of this disease. In particular, the high correlation between per capita fat consumption and breast cancer incidence, together with data from laboratory animal experiments, suggest that a high intake of dietary fat may be an important risk factor for breast cancer. In 1980, 89,538 American participants in the Nurses Health Study, who were years of age and had no history of cancer, completed a validated dietary questionnaire designed to measure their individual consumption of total fat, saturated fat, linoleic acid, cholesterol, and other nutrients. During 4 years of follow-up, 601 cases of breast cancer were diagnosed among the 89,538 participants in the study. After adjusting for known determinants of breast cancer, the relative risk among women in the highest quintile of total fat intake, adjusted for total energy intake, compared with women in the lowest quintile, was 0.82 (95%, ). The corresponding relative risk for women in the highest quintile of saturated fat intake was 0.84; for those in the highest quintile of linoleic acid intake, 0.88; and for cholesterol, Although no suggestion of a positive relationship between fat intake and the risk of breast cancer was observed overall, we considered the possibility that associations might exist in subgroups defined by known risk factors. However, no statistically significant positive associations were found within categories of age at first birth, parity, age at menarche, maternal history of breast cancer, history of breast cancer in a sister, previous benign breast disease, or relative weight. Furthermore, we found no evidence of any positive association between dietary fat consumption and breast cancer among either premenopausal or postmenopausal women. In a subsample of 173 participants studied in detail, those in the highest quintile of fat intake consumed a mean of 44% of calories from fat compared with 32% for those in the lowest quintile. These findings suggest that a reduction in total fat intake of approximately 25% by women, which corresponded to the difference between the highest and lowest quintile of intake in this population, was unlikely to cause a substantial decrease in the incidence of breast cancer. These results persisted when we extended the followup through Based on extreme deciles of fat intake in 1984, expressed as a percentage of energy intake, we compared women consuming more than 40% of energy from fat with those consuming less than 25% and observed a relative risk of 0.83 (95% confidence interval, ). These findings for breast cancer contrasted with the strong relationship seen in the same population when we examined diet and the risk of colonic cancer. Specifically, women consuming a high-fat diet had a significant increase in the risk of colonic cancer compared

6 Epidemiology of Breast Cancer/Colditz 1485 with those consuming a lower fat diet.42 Furthermore, data from other prospective studies of diet and breast cancer were consistent with our results and showed little evidence of an increase in risk with increasing intake of fat. Using data from a 24-hour dietary recall administered to the National Health and Nutrition Examination Survey population, a significant inverse association was found between fat intake and the risk of breast cancer.43 Others observed 519 cases of breast cancer during 5 years of follow-up of a cohort of Canadian women.44 The risk of breast cancer did not increase with increasing fat intake. Specifically, the lowest relative risk was observed among women in the second quartile of fat intake. These data, taken together, suggest that modification of dietary fat intake among adult women is unlikely to reduce the risk of breast cancer. Alcohol In 1977, the association between alcohol and breast cancer was suggested by an analysis of data obtained from the Third National Cancer Survey.45 Several subsequent case-control studies confirmed this positive association, but no relationship was found in others.46 None of these studies included a comprehensive assessment of dietary factors that might have confounded the association with alcohol. Thus, we undertook a prospective analysis based on the data obtained from the 1980 dietary component of the Nurses Health Study.47 Alcohol intake was validated as part of the overall validation study of the food frequency questionnaire. The mean alcohol intake as assessed by four 1-week diet records completed by 194 participants in the Boston area was 9.0 g/day, which was similar to the intake computed from information on the 1980 questionnaire (mean intake, 8.2 g/day, Spearman r = 0.86).48 Among the women consuming 5-14 g/day of alcohol (approximately three to nine drinks per week), the relative risk of breast cancer was 1.3 (95% confidence interval, ). Consumption of 15 g/day of alcohol or more was associated with a relative risk of 1.6 (95% confidence interval, ). Adjustment for known breast cancer risk factors and various nutritional variables did not alter materially this 60% increase in risk. When alcohol intake was expressed as the average number of drinks per day regardless of the type of beverage, the relative risks compared with nondrinkers were 1.0 for less than 25% of a drink per day, 0.8 for 25% of a drink per day, 1.3 for 50% a drink per day, 1.4 for 1 drink per day, and 1.5 for more than 1 drink per day. No increased risk was observed among past drinkers compared with other women reporting no alcohol intake during the previous year (relative risk, 1.0; 95% confidence interval, ). Since epidemiologic associations often are seen most clearly among persons at low risk of disease, we defined such a group as women who were younger than 55 years of age, who had a full-term pregnancy before the age of 26 years, and who had no personal history of benign breast disease or maternal or sibling history of breast cancer. Among the 20,589 low-risk women, 95 cases of breast cancer were diagnosed. Within this group, the relative risk for those consuming 15 g/day of alcohol or more was 2.5 (95% confidence interval, ) compared with women consuming no alcohol. The corresponding relative risk among women not in the lowest group was 1.5 (95% confidence interval, ). These data, taken together with results from other epidemiologic and animal studies,49 indicate that alcohol intake may be related causally to the onset of breast cancer. When data were combined from more than 20 studies, a relative risk of 1.4 (95% confidence interval, ) was observed in case-control studies and 1.7 (95% confidence interval, ) in prospective cohort studies for women consuming alcohol compared with women who never drank. A detailed analysis of data from six case-control studies noted the absence of a relationship between alcohol consumption and the risk of breast cancer for less than 40 g/day of ethanol; there was an elevated risk with intake above this amount.51 These six studies draw on data from diverse populations with different patterns of alcohol consumption. Strong evidence in support of this causal association comes from a feeding trial in which 34 premenopausal women, aged years, were randomized to a controlled diet, which included 29.9 g/day of alcohol for three menstrual cycles, or an alternate diet, which included no alcoh01.~ Hormone levels then were compared between the alcohol-consumption and alcoholfree diets. While receiving the alcohol-containing diet, these women had a significant 27.7% elevation in their estradiol levels during the periovulatory phase and a 21.1% increase in estrone. This was followed by increases in urinary estrogens during the luteal phase. Further evidence regarding the carcinogenicity of alcohol comes from studies of animals and from the depletion of vitamin A from the liver of animals fed alcohol. Despite this potentially causal relationship between alcohol intake and the risk of breast cancer, the association between moderate alcohol consumption and a decreased risk of coronary heart diseases3 and gallstones54 make recommendations on the safe limit for alcohol intake particularly difficult in women. Vitamin A Based on prospective analysis of women in the Nurses Health Study who were followed for 8 years, we ob-

7 1486 CANCER Supplement February 25, 2993, Volume 71, No. 4 served a significant decrease in risk with increasing consumption of preformed vitamin A (including supplement~).~~ Women in the highest quintile of intake had a relative risk of 0.78 (95% confidence interval, ) compared with those in the lowest quintile of intake. Carotene consumption was not related significantly to risk, although the trend was inverse (P = 0.09). The intake of vitamin C was not related to risk, and vitamin E consumption was related inversely to risk (P trend = 0.05 for intake with supplements) in this 8-year analysis. Selenium Selenium is an essential trace element in human nutrition, and selenium deficiency or excess has been implicated in several diseases. Because an inverse relationship between selenium and the risk of breast cancer has been reported from correlation studies and from experimental studies in animals, we investigated this association in the Nurses Health Study. We showed that toenail selenium levels reflect prior intake of this element.56 Among the 68,215 women who sent in toenails after the 1982 questionnaire cycle, 434 cases of breast cancer were diagnosed up to June To each case, we matched one control born in the same calendar year and who returned nails in the same month as the case. Among the 434 cases, the mean selenium level was pg/g. This was not different from the mean for control subjects (mean, p/g; P = 0.90). Dividing the cases and control subjects into quintiles of selenium level (based on the distribution among the control group), we observed that the relative risk for those with the highest nail selenium levels was 1.10 (95% confidence interval, ) compared with those in the lowest q~intile.~ The lack of association between toenail selenium and the risk of breast cancer suggests that selenium supplementation later in life will not reduce a woman s risk of breast cancer. We cannot, however, exclude the possibility that intake in early life may influence the subsequent risk. Benign Breast Biopsy Histopathologic Findings Although benign breast disease is associated with an increased risk of breast cancer, different histopathologic changes confer differing levels of risk. A strong relationship was reported between atypical hyperplasia and the risk of breast cancer in women who had undergone biopsy for benign breast di~ease.~ We designed and implemented a study in the Nurses Health Study to replicate these findings. We identified women with breast cancer who had a prior history of breast biopsy for benign breast disease. To each case, we matched control subjects born the same year who underwent a breast biopsy during the same year. Based on 12 1 cases and 488 control subjects compared with women who had no proliferative changes in breast biopsy specimens, we observed a fourfold increase in risk among the women with biopsy changes of atypical hyperplasia (relative risk, = 3.7; range, ) and a twofold increase in risk among women with proliferative changes without atypia (relative risk, 1.6; range, ).59 The relative risk was greater among premenopausal women (risk for atypia, 5.9; range, ) and among women with a family history of breast cancer (relative risk, 7.3; range, ), although the confidence intervals were wide for these subgroup analyses. Atypical hyperplasia may be regarded as a precursor lesion for breast cancer. However, we have little understanding of the factors associated with progression from this lesion to invasive disease. Additional work is needed to advance our understanding and thus help identify factors that may prevent this progression to breast cancer. Adolescent Nutrition Western children are growing taller and attaining maturity at an earlier age. Secular changes are well documented. The stature increase in European children has been documented since the middle of the last century, and it was shown that it occurred in each country.60 The increase in attained height may be a consequence of better adolescent nutrition, better living standards, or better control of serious childhood illnesses. Concomitant with the increase in attained height has been an acceleration of the age at menarche during the past century.61 The age at menarche is estimated to have decreased by 4 months per decade during the period from However, in terms of breast cancer risk, this can explain only a 100% increase in risk if the mean age at menarche remains at older than 17 years as seen in China6 compared with the current mean of younger than 13 years in the United States. Height may be an alternative marker of adolescent nutrition. In a prospective study of women who attained puberty during a period of nutritional variability (i.e., energy restriction during World War 11), height was related positively to the risk of breast cancer, and the association varied by birth However, like age at menarche, this does not offer an approach for the prevention of breast cancer. International variation in the rates of breast cancer cannot be explained fully when a model such as that proposed by Pikez0 is fitted with the data on reproductive events. However, these reproductive events per se

8 Epidemiology of Breast Cancer/Colditz 1487 do not reflect hormone levels, body size, or the number of breast cells. These factors may influence the risk. International comparisons show that rural Chinese women have significantly lower estradiol and testosterone levels than do British women.64 Even after adjusting for body size, hormone levels were significantly lower among premenopausal Chinese women in Shanghai than among white women from Los Angele~.~~ Specifically, women in Los Angeles had 20.6% greater estradiol concentrations than those in Shanghai. Animal studies show that energy restriction early in life profoundly reduces the number of mammary tumors.66 Furthermore, the breast is particularly sensitive to cancer induction during adolescence, as seen among those exposed to ionizing radiati~n.~ Pregnancy induces terminal differentiation of human breast glands, which may have a smaller proliferative component. Animal studies indicate that, after first pregnancy, the secretory cells of the breast gland have a longer cell cycle, are resistant to carcinogens, and have a more efficient DNA repair capacity.68 Whether height or age at menarche is taken as a marker of adolescent diet, from the epidemiologic data, it appears that nutrition at an early age may have a long-term impact on growth and subsequent cancer risk. Although attained height has reached a plateau in the United States, this suggests that the rates of breast cancer may stabilize among women born after 1950 and should not show additional increases. However, the age at menarche has not stabilized. Conclusion Much of the epidemiologic data relating risk factors to the incidence of breast cancer can be synthesized through the estrogen-related factors of length of exposure to cyclic hormones (represented by the woman s age at menopause minus her age at menarche) and, among postmenopausal women, obesity and exogenous hormone use. Survival rates after breast cancer also highlight the role of estrogens in the cause of breast cancer. The number of breast cells may be reflected by the woman s height, which is a marker of adolescent nutrition, and perhaps her age at menarche. The diet in adult life may have little impact on the risk, suggesting that approaches to prevention will not be easy to identify or implement. To address prospectively the relationship between hormone levels and subsequent risk of breast cancer among postmenopausal women, we have collected blood samples from approximately 35,000 women in the Nurses Health Study. Data from this and other similar studies may help identify future avenues for hor- monal blocking or manipulation that could reduce the risk among mature women. References Muir C, Waterhouse J, Mack T, Powell J, Whelan S. Cancer incidence in five continents, vol. 5. Lyon: International Agency for Research on Cancer, National Center for Health Statistics. Vital statistics of the United States vol. 2. Mortality. Part A. Washington, DC: Government Printing Office, DHHS publication no. (PHS) Moolgavkar SH, Day NE, Stevens RG. Two-stage model for carcinogenesis: epidemiology of breast cancer in females. ] Nut1 Cancer Inst 1980; Armstrong B, Doll R. Environmental factors and cancer incidence and mortality in different countries, with special reference to dietary practices. Int ] Cancer 1975; 15: Miller BA, Feuer EJ, Hankey BF. The increasing incidence of breast cancer since 1982: relevance of early detection. Cancer Causes Control 1991; Boyle P, Maissoneuve P, Hsieh C. Epidemiology of breast cancer. In: Zatonski W, Boyle P, Tyczynski J, editors. Cancer prevention: vital statistics to prevention: proceedings of the International Conference October 1988, Warsaw, Poland. Warsaw: PA Interpress, 1990: Prentice RL, Sheppard L. Dietary fat and cancer: consistency of the epidemiologic data and disease prevention that may follow from a practical reduction in fat consumption. Cancer Causes Control 1990; 1: Blot WJ, Devesa SS, Fraumeni JF. Declining breast cancer mortality among young American women. ] Natl Cancer Inst 1987; 78~ Moms DL, Hulka BS. Increases in breast cancer incidence and mortality rates between 1975 and 1986 in white and black Americans: a screening effect or an alarming trend [abstract]? Am ] Epidemiol 1990; Shapiro S, Venet W, Strax P, Venet L, Roeser R. Selection, follow-up, and analysis in the Health Insurance Plan Study: a randomized trial with breast cancer screening. Natl Cancer Znst Monogr 1985; 67: Momson AS, Brisson J, Khalid N. Breast cancer incidence and mortality in the Breast Cancer Detection Demonstration Project. ] Nut1 Cancer Inst 1988; 80: Reis LAG, Hankey BF, Edwards BK, editors. Cancer statistics review Bethesda (MD): National Cancer Institute: Publication no. (NIH) White E, Lee CY, Kristal AR. Evaluation of the increase in breast cancer incidence in relation to mammography use. ] Natl Cancer Inst 1990; Lantz PM, Remmngton PL, Newcomb PA. Mammography screening and the increased incidence of breast cancer in Wisconsin. ] Natl Cancer Inst 1991; 81: Bain C, Speizer FE, Rosner B, Belanger C, Hennekens CH. Family history of breast cancer as a risk indicator for the disease. Am j Epidemiol 1980; 111: Bain C, Willett W, Rosner B, Speizer FE, Belanger C, Hennekens CH. Early age at first birth and decreased risk of breast cancer. Am ] Epidemiol 1981; 114: Willett WC, Stampfer MJ, Colditz GA, Rosner B, Hennekens CH, Speizer FE. Dietary fat and risk of breast cancer. N Engl ] Med 1987; 316:22-8.

9 1488 CANCER Supplement February 25, 2993, Volume 71, No Ottman R, Pike MC, King M-C, Henderson BE. A practical guide for estimating risk for familial breast cancer. Lancet 1983; 2: Pathak D, Speizer FE, Willett WC, Rosner B, Lipnick RJ. Parity and breast cancer risk: possible effect on age at diagnosis. Znt ] Cancer 1986; 37: Pike MC, Krailo MD, Henderson BE, Casagrande JT, Hoe1 DG. "Hormonal" risk factors, "breast tissue age" and the age-incidence of breast cancer. Nature 1983; 303: Bruzzi P, Negri E, LaVecchia C, Decarli A, Palli D, Parazzini F, et al. Short term increase in risk of breast cancer after full term pregnancy. BM] 1988; 297: Janerich DT, Hoff MB. Evidence for a crossover in breast cancer risk factors. Am ] Epidemiol 1982; 116: London SJ, Colditz GA, Stampfer MJ, Willett WC, Rosner B, Speizer FE. A prospective study of relative weight, height and the risk of breast cancer. ]AMA 1989; 262: Tretli S. Height and weight in relation to breast cancer morbidity and mortality: a prospective study of 570,000 women in Norway. lrit ] Cancer 1989; 44: Willett WC, Browne ML, Bain C, Lipnick RJ, Stampfer MJ, Rosner B, et al. Relative weight and risk of breast cancer among premenopausal women. Am ] Epidemiol 1985; 122: Tretli S, Haldorson T, Ottestad L. The effects of premorbid height and weight on the survival of breast cancer patients. Br ] Cancer 1990; 62: Coates RJ, Clark WS, Eley JW, Greenberg RS, Huguley CM, Brown RL. Race, nutritional status, and survival from breast cancer. ] Natl Cancer lnst 1990; 82: Early Breast Cancer Trialists' Collaborative Group. Effects of adjuvant tamoxifen and of cytotoxic therapy on mortality in early breast cancer: an overview of 61 randomized trials among 28,896 women. N Engl ] Med 1988; 319: Key TJA, Pike MC. The role of estrogens and progestagens in the epidemiology of and prevention of breast cancer. Eur ] Cancer 1988; 34~ McPherson K, Vessey MP, Neil A, Doll R, Jones L, Roberts M. Early oral contraceptive use and breast cancer: results of another case-control study. Br ] Cancer 1987; 56: Romieu I, Willett WC, Colditz GA, Stampfer MJ, Rosner 8, Hennekens CH, et al. A prospective study of oral contraceptive use and the risk of breast cancer in women. ]Nut/ Cancer lnst 1989; 81: Romieu I, Berlin JA, Colditz GA. Oral contraceptives and breast cancer: review and meta-analysis. Cancer 1990; 66: Steinberg KK, Thaker SB, Smith SJ, Stroup DF, Zack MM, Flanders WD, et al. A meta-analysis of the effect of estrogen replacement therapy and risk of breast cancer. ]AMA 1991; 265~ Colditz GA, Stampfer MJ, Willett WC, Hunter DJ, Manson JE, Hennekens CH, et al. Postmenopausal hormone use and risk of breast cancer: 12-year follow-up of the Nurses' Health Study. In: Mann RD, editor. Hormone replacement therapy and breast cancer risk. Camforth, UK: Parthenon, 1992: Gambrel1 RD, Maier RC, Sanders BI. Decreased incidence of breast cancer in postmenopausal estrogen-progesterone users. Obstet Gynecol 1983; 62: Ames BN, Kammen HO, Yamasaki E. Hair dyes are mutagenic: identification of a variety of mutagenic ingredients. Proc Natl Acad Sci U S A 1975; 72: Green A, Willett WC, Colditz GA, Stampfer MI, Bain C, Rosner B, et al. Use of permanent hair dyes and risk of breast cancer. ] Natl Cancer Inst 1987; 79: MacMahon 8, Trichopoulos D, Cole P, Brown J. Cigarette smoking and urinary estrogens. N Engl ] Med 1982; 307: London SJ, Colditz GA, Stampfer MJ, Willett WC, Rosner BA, Speizer FE. A prospective study of smoking and the risk of breast cancer. ] Natl Cancer Znst 1989; 81: London SJ, Colditz GA, Stampfer MJ, Willett WC, Rosner BA, Corsano K, et al. Lactation and risk of breast cancer among a cohort of US women. An1 ] Epideniiol 1990; 312: Willett WC, Hunter DJ, Stampfer MJ, Colditz GA, Rosner BA, Hennekens CH, et al. Dietary fat and breast cancer: an 8-year follow-up [abstract]. An7 ] Epidemiol 1991; 134: Willett WC, Stampfer MJ, Colditz GA, Rosner BA, Speizer FE. Relation of meat, fat, and fiber intake to the risk of colon cancer in a prospective study among women. N Engl ] Med 1990; 323: Jones DY, Schatzkin A, Green SB, Block G, Brinton LA, Zeigler RG, et al. Dietary fat and breast cancer in the National Health and Nutrition Examination Survey: 1. Epidemiologic follow-up study. ] Natl Cancer Znst 1987; 79: Howe GR, Friedenreich CM, Jain M, Miller AB. A cohort study of fat intake and risk of breast cancer. ] Natl Cancer Znst 1991; 83~ Williams RR, Horm JW. Association of cancer rates with tobacco and alcohol consumption and socioeconomic status of patients: interview study from the Third National Cancer Survey. Natl Cancer lnst 1977; 58: Willett WC, Stampfer MJ, Colditz GA. Does alcohol consumption influence the risk of developing breast cancer? In: DeVita VT, Hellman S, Rosenberg SA, editors. Important advances in oncology. Philadelphia: JB Lippincott, 1989: Willett W, Stampfer MJ, Colditz GA, Rosner 8, Hennekens CH, Speizer FE. Moderate alcohol consumption and risk of breast cancer. N Engl] Med 1987; 315: Giovannucci E, Colditz GA, Stampfer MJ, Rimm EB, Litin L, Sampson L, et al. The assessment of alcohol consumption by a simple self-administered questionnaire. Am ] Epideniiol 1991; Schrauzer GN, McGinness JE, Ishmael D, Bell LJ. Alcoholism and cancer: 1. Effects of long-term exposure to alcohol on spontaneous mammary adeno-carcinoma and prolactin levels in C3H/St mice. ] Stud Alcohol 1979; 40: Longnecker MP, Berlin JA, Orza MJ, Chalmers TC. A meta-analysis of alcohol consumption in relation to risk of breast cancer. ]AMA 1988; 260: Howe G, Rohan T, Decarli A, lscovich J, Kaldor J, Katsouyanni K, et al. The association between alcohol and breast cancer risk: evidence from the combined analysis of six dietary case-control studies. Int ] Cancer 1991; 47: Reichamn ME, Judd JT, Longcope C, Schatzkin A, Nair PI', Taylor PR, et al. Alcohol consumption and hormone levels in a controlled diet study of premenopausal women [abstract]. Am ] Epidemiol 1991; 134: Stampfer MJ, Colditz GA, Willett WC, Speizer FE, Hennekens CH. A prospective study of moderate alcohol consumption and risk of coronary heart disease and stroke in women. N Engl ]Med 1988; 319~ Stampfer MJ, Colditz GA, Willett WC, Manson JE, Arky RA, Hennekens CH, et al. A prospective study of moderate alcohol drinking and risk of diabetes in women. Am ] Epidemiol 1988; 128: Hunter DJ, Stampfer MJ, Colditz GA, Manson J, Rosner B, Hennekens CH, et al. A prospective study of consumption of vitamins A, C, and E and breast cancer risk [abstract]. An1 ] Epideniiol 1991; 134: Hunter DJ, Moms JS, Chute CG, Kushner E, Colditz GA, Stampfer MI, et al. Predictors of selenium concentration in human toenails. Am ] Epidemiol 1990; 132:

10 Epidemiology of Breast Cancer/CoZditz Hunter DJ, Moms JS, Stampfer MJ, Colditz GA, Speizer FE, Willett WC. A prospective study of selenium status and breast cancer risk. IAMA 1990; 264: Dupont WD, Page DL. Risk factors for breast cancer in women with proliferative breast disease. N Engl Med 1985; 312: London SJ, Connolly JL, Schnitt SJ, Colditz GA. A prospective study of benign breast disease and risk of breast cancer. IAMA 1992; 267: Boas F. Observations on the growth of children. Science 1930; 72~ Tanner JM. The development of the female reproductive system during adolescence. Clin Obstet Gynecol 1960; 3: Chen J, Campbell TC, Junyao L, Pet0 R. Diet, lifestyle and mortality in China. Oxford, UK: Oxford University Press, 1990: Chen J, Campbell TC, Junyao L, Pet0 R. Diet, lifestyle and mortality in China. Oxford University Press, Vatten LJ, Kvinnsland SK. Body height and risk of breast cancer: a prospective study of 23,831 Norwegian women. BY I Cancer 1990; 61~ Wang DY, Key TJA, Pike MC, Boreham J, Chen J. Serum hormone levels in British and rural Chinese females. Breast Cancer Research and Treatment 1991; Bemstein L, Yuan J-M, Ross RK, Pike MC, Hanisch R, Lob0 R, et al. Serum hormone levels in pre-menopausal Chinese women of Shanghai and white women in Los Angeles: results from two breast cancer case-control studies. Cancer Causes and Control 1990; 1~ Ip C. Quantitative assessment of fat and calories as factors in mammary carcinogenesis in an experimental model. Prog Clin Biol Res 1990; 346: McGregor OH, Land CE, Choi K, Tokuoka S, Lui PI, Wakabayashi T, et al. Breast cancer incidence among atomic bomb survivors, Hiroshima and Nagasaki, ]Nut1 Cancer Irist 1977; 59: , 69. Russo J, Kay LK, Russo IH. Differentiation of the mammary gland and susceptibility to carcinogenesis. Breast Cancer Research and Treatment 1982; 2:5-73.