Breast Cancer Research and Treatment 18: Sll-S17, 1991. 9 1991 Kluwer Academic Publishers. Printed in the Netherlands.
The international variation in breast cancer rates: An epidemiological assessment
Brian E. Henderson and Leslie Bernstein
Department of Preventive Medicine, University of Southern California School of Medicine, Norris Cancer Hospital and Research Institute, 1441 Eastlake Avenue, Los Angeles, California 90033, USA
Key words: breast cancer, geographic variation, estrogen production, hormonal exposure, incidence, risk factors Abstract
Part of the international differences in breast cancer incidence rates can be explained by geographic variation in reproductive and other breast cancer risk factors. Age at menarche and age at onset of regular ovulatory menstrual cycles are two such factors; both vary across populations directly according to breast cancer risk, and both are acknowledged as breast cancer risk factors. Consideration of the body of evidence on these factors, as well as that on age at menopause, suggests that the cumulative frequency of ovulatory menstrual cycles is a critical determinant of breast cancer risk. Although age at first term pregnancy explains the majority of the protective effect of parity on breast cancer risk, two recent studies have demonstrated a small residual protective effect of increasing number of births. It appears that pregnancy has paradoxical effects on breast cancer risk in terms of hormone production and metabolism. The initial effect is an increased risk associated with first trimester estrogen exposure. However, the hormonal consequences of completing the pregnancy counteract this negative effect of early pregnancy. The effect of body weight, a breast cancer risk factor for postmenopausal women, can be explained in terms of increased extraglandular conversion of androstenedione to estrone. Further evidence supporting a pathogenic role of estrogens in the development of breast cancer comes from international studies of endogenous hormones in populations with differing risks of breast cancer. These risk factors have been incorporated into a mathematical model which is based on the concept that breast tissue ages according to hormonal (primarily estrogen) exposure; this model closely predicts the incidence rates throughout the world.
There is remarkable variation in the breast cancer incidence rates of different countries [1]. Rates are nearly six times higher in the United States, Canada, or Northern Europe than in Asia or black Africa. These differences in breast cancer incidence rates are not determined by variation in genetic susceptibility [2]. The breast cancer incidence rates of American blacks are very similar to those of
American whites, but differ from those of African blacks. Furthermore, Japanese migrants to Hawaii and California have higher rates of breast cancer than women in Japan [3, 4]. Native Japanese women who migrate to the United States as young adults have a small increase in their breast cancer rates while living in the United States, whereas Japanese women born in the United States have
Address for offprints: B.E. Henderson, Department of Preventive Medicine, University of Southern California School of Medicine, Norris Cancer Hospital and Research Institute, 1441 Eastlake Avenue, Los Angeles, California 90033, USA
S12
BE Henderson and L Bernstein
rates approaching those of their white counterparts. This suggests that the risk of breast cancer is at least partly related to exposures that are acquired early in life. These large international differences in breast cancer incidence can be at least partially explained by geographic variation in breast cancer risk factors.
Age at menarche Early age at menarche has been demonstrated as a risk factor for breast cancer in most case-control studies. In general, breast cancer risk decreases approximately 20 per cent for each year of age that menarche is delayed. In a case-control study of young women, we recorded not only age at onset of menstruation but also age when 'regular' (that is, predictable) menstruation was first established [5]. For a fixed age at menarche, women who established regular menstrual cycles within one year of their first menstrual period had more than double the risk of breast cancer of women who had a five-year or longer delay in establishing regular menses. For women with early menarche (aged 12 or younger) and rapid establishment of regular menstrual cycles (less than one year), breast cancer risk was nearly fourfold greater than that of women with late menarche (aged 13 or older) and long duration of irregular cycles (five or more years). Because the cumulative amount of estrogen produced in the normal luteal phase of an ovulatory cycle is greater than that produced during the comparable period of a nonovulatory cycle, the cumulative frequency of ovulatory cycles serves as an index of cumulative estrogen (as well as progesterone) exposure. Results from an earlier study of circulating homrone levels in daughters of breast cancer cases and in age-matched daughters of controls are consistent with this hypothesis. The daughters of the breast cancer cases, who as a group have at least twice the breast cancer risk of the general population, had higher levels of circulating estrogen and progesterone during the luteal phase of the menstrual cycle (on day 22) than did controls [6]. This result has been confirmed by other investigators [7].
Other supportive evidence for the concept that the cumulative number of ovulatory cycles (that is, cumulative estrogen exposure) is a major determinant of breast cancer risk [8] comes from international studies of the frequency of ovulation in relation to age at menarche and number of years since menarche [9]. In girls aged 15 to 19 years, selected from several populations at varying risk of breast cancer, those with later menarche were more likely to have anovular menstrual cycles than girls with early menarche, given the same number of elapsed years since menarche. Adjusting for years since menarche, the highest frequency of ovulatory menstrual cycles was observed in those populations with the highest breast cancer rates. Apter and Vihko [10], in a longitudinal study of 200 Finnish schoolgirls, also found that those with early menarche establish ovulatory menstrual cycles more quickly than girls with later onset of menstruation. The average intervals from menarche until 50 per cent of cycles were ovulatory were 1, 3, and 4.5 years, respectively, for girls with menarche before age 12, between age 12 and 13, and at age 13 years and older. Forty-four of these girls were later followed into the third decade of life and evaluated endocrinologically [11]. Women with early menarche (before age 12) had higher serum estradiol concentrations during the follicular phase of the menstrual cycle (at average age 27) than did women with later menarche. Furthermore, the serum estradiol concentrations increased more rapidly to the mid-cycle peak in these early-menarche women than in those with later menarche. In addition, follicular phase sex hormone-binding globulin (SHBG) concentrations were about 30 per cent lower in women with early menarche. Over the past 100 years, the age at menarche has progressively decreased both in the United States and in most other areas of the world. A series of extensive cross-sectional studies has demonstrated that age at menarche is directly related to childhood growth patterns [12]. Attainment of a critical body weight-to-height ratio appears necessary for menarche to occur [13]. Chronic malnutrition during childhood delays the age at menarche, whereas the improved nutrition and control of infectious
International variation in breast cancer rates
diseases of childhood of the past decades have combined to lower it. Strenuous physical activity may also delay menarche. Girls who engage in regular ballet dancing, swimming, or running may have a considerable delay in the onset of menses. In one study, ballet dancers had a mean age at menarche of 15.4 years, compared with 12.5 years for controls [14]. Breast development was also delayed in the dancers, and intermittent amenorrhea persisted throughout their teenage years as long they remained active dancers. Even moderate physical activity during adolescence can lead to anovular menstrual cycles. In a study of menstrual cycle patterns during adolescence, we found that girls (aged 14 to 17) who engaged in moderate physical activity (averaging 600 kcal of energy expenditure per week) were 2.9 times more likely to be anovular than were girls engaging in lesser amounts of physical activity [15].
Age at menopause The relationship between age at menopause and breast cancer risk has been known for some time. The rate of increase in the age-specific incidence rate of breast cancer slows sharply at the time of menopause and the rate of increase in the postmenopausal period is only about one-sixth the rate of increase in the premenopausal period [1]. It has been estimated that women whose natural menopause occurs before age 45 have only one-half the breast cancer risk of those who menopause occurs after age 55 [16]. Another way of expressing this result is that women with 40 or more years of active menstruation have twice the breast cancer risk of those with fewer than 30 years of menstrual activity. Artificial menopause, by either bilateral oophorectomy or pelvic irradiation, also markedly reduces breast cancer risk. The effect appears to be just slightly greater than that of natural menopause (defined as cessation of menstrual periods). As shown by Feinleib [17], there is little change in a woman's risk of breast cancer following a unilateral oophorectomy or a simple hysterectomy.
S13
Age at first full-term pregnancy Two of the earliest known and most reproducible features of breast cancer epidemiology are the decreased risk associated with increased parity and the increased risk of single women. MacMahon and colleagues [18] made a major advance in our understanding of the role of pregnancy in altering breast cancer risk through their analysis of an international collaborative case-control study. Single and nulliparous married women were found to have the same increased risk of breast cancer, approximately 1.4 times the risk of parous married women. Among married women in each country, parous breast cancer patients had fewer children than had parous women in the control groups. These authors clearly demonstrated, however, that this protective effect of parity was totally explained by a protective effect of early age at first birth. Those women with a first birth before age 20 had about one-half the risk of nulliparous women. Controlling for age at first birth, subsequent births provided no further protection against breast cancer. Recently, two studies conducted in other populations have observed small residual protective effects of an increasing number of births [19, 20]; this suggests that there may be certain circumstances in which multiparity does offer some further protection. In a recent study in Shanghai, we observed that there was a protective effect of multiple pregnancies, most notably after the fifth pregnancy [20]. The main protective effect is, however, undoubtedly associated with the first full-term pregnancy. Women who have a late first full-term pregnancy are actually at elevated risk of breast cancer compared with nulliparous women [18]. This paradoxical effect of a late first full-term pregnancy has been repeatedly confirmed by other studies. A possible explanation for this effect is suggested by somewhat related observations on breast cancer risk immediately following a full-term pregnancy and following an uncompleted first pregnancy. In a hospital-based case-control study, it was demonstrated that women who gave birth during the three years before interview had higher risk of breast cancer than those whose last birth occurred at least ten years earlier (after adjustment for age, age at
$14
B E Henderson and L Bernstein
first birth and parity, relative risk = 2.7) [21]. Recently, we found that a first-trimester abortion, whether spontaneous or induced, occurring before the first full-term pregnancy was associated with higher risk of breast cancer [22]. This observation has been confirmed in a cohort of Connecticut women who had only one live birth [23]. In that study, abortions after the first full-term pregnancy did not carry any increased risk of breast cancer. Based on these results, we have concluded that there are two contradictory effects of pregnancy on breast cancer risk which are particularly notable in the first pregnancy [24]. This apparent paradox has a physiological explanation based on patterns of estrogen and prolactin secretion and metabolism during pregnancy. During the first trimester of pregnancy, there is a rapid rise in the level of 'free' estradiol (that is, the fraction of estradiol that is biologically available to act upon tissues), an effect that is more apparent in the first rather than subsequent pregnancies [25]. Thus, the net effect of this early part of pregnancy, in terms of estrogen exposure to the breast, is an increase in risk that is equivalent to the exposure from several ovulatory cycles over a relatively short time period. In the long run, however, this negative effect of early pregnancy on breast cancer risk can be overridden by two beneficial, hormonal consequences of completing the pregnancy. Several years ago, we reported that prolactin levels were substantially lower in parous compared with nulliparous women [26], an observation that was replicated recently [27]. In addition, we found that parous women had higher levels of SHBG and lower levels of free estradiol than their nulliparous counterparts [28]. Lactation has not been clearly established to either enhance or protect against development of breast cancer. If the cumulative number of ovulatory cycles is directly related to risk of breast cancer, a beneficial effect of long duration of nursing would be expected because nursing results in a substantial delay in the reestablishment of ovulation following a completed pregnancy. Because of a small proportion of mothers with a large cumulative number of nursing months, most previous epidemiologic studies have not allowed for precise estimates of the effects of lactation on breast cancer
risk. We recently completed a population-based case-control study in China, a population in which long-duration nursing is the norm [20]. In this study, a 30 per cent reduction in breast cancer risk was observed for each five years of nursing experience.
Weight In addition to the menstrual and reproductive risk factors described above, there is a strong relationship between weight and risk of breast cancer. The relationship is critically dependent on age. For women under age 50, there is little or no increased risk associated with increased weight, but by age 60, a 10-kg increase in weight results in approximately an 80 per cent increase in breast cancer risk [29]. Whether this weight effect is one of excess weight (body fat) or absolute weight as such is inclear. Contradictory results have been reported on whether, for example, Quetelet's index (wt/ht2) is correlated with breast cancer risk. Absolute weight appears to be as good an indicator of risk as any function of weight and height. In the postmenopausal period, the major source of estrogen is extraglandular (largely adipose tissue) conversion of the adrenal androgen androstenedione to estrone [30]. The association of breast cancer with cyclic ovarian activity implies that estrogen is important in the pathogenesis of this disease. We have discussed the rationale that hormones, and in particular, estrogen, can directly increase the incidence of breast cancer [31]. The most carefully done international studies comparing estrogen levels in populations at differing risks of breast cancer support the pathogenic role of estrogens. In the early 1970s, MacMahon and his colleagues [32] conducted a series of studies on teenagers and young women to investigate whether some aspect of estrogen metabolism was responsible for the large differences in breast cancer rates between Asia and North America. They knew that the Asian women chosen for study should be a group who would themselves be expected to experience low breast cancer rates, so
International variation in breast cancer rates
they chose study populations as similar as possible to women than in the breast cancer age range. Although they initially reported their results in terms of the relative amounts of various urinary estrogens, their results can also be evaluated in terms of absolute total urinary estrogen levels (which to some degree probably reflect serum estrogen levels) [33]. They found that, in overnight urines collected on the morning of day 21 of the menstrual cycle, total urinary estrogen levels were 36 per cent higher in the North American teenagers. Levels were also elevated in samples collected on the morning of day 10, but only by 9 per cent. These differences might be related to later menarche in the Japanese women, but even larger differences were found in older women, for whom the effects of late menarche should have disappeared. In nulliparous women aged 20 to 24, total urinary estrogen levels in the North American women were 49 per cent higher on day 21 and 38 per cent higher on day 10 (during the follicular phase of the menstrual cycle); similar differences were found among parous women aged 30 to 39. These differences in total urinary estrogen are presumably a reflection on both reduced frequency of ovulatory cycles, and less effective corpus luteum formation in those cycles for which ovulation did
S15
ference. These higher levels of estradiol in white compared with Chinese women could be an important part of the explanation of the twofold to threefold differences in the breast cancer risk of young women in the United States and in Asia. Greater hormonal differences are observed in comparisons of postmenopausal Asian and United States white women. Recently we studied serum estrogen levels in postmenopausal Japanese women in Japan and postmenopausal white women in Los Angeles [41]. The Japanese women studied were deliberately chosen to be from a rural agricultural area in order to get samples which represented as closely as possible the traditional Japanese 'lifestyle' that gave rise to the low rates of breast cancer in Japan. Levels of estrone and of estradiol were both significantly higher (47 per cent and 36 per cent, respectively) in the sera of white women and adjustment for differences in body weight accounted for little of these differences. After the menopause, there is no increase in the age-specific breast cancer incidence rates of Japanese women as there is for white women in the United States. These decreased levels of serum estrogens could well explain this further divergence in the breast cancer incidence rates.
Occur.
On the hypothesis that increased ovarian activity increases the risk of breast cancer, one would predict that breast cancer case-control studies evaluating luteal-phase circulating levels of estradiol would find evidence of greater production in the cases. Five such studies have been undertaken, but four of them had small numbers of cases ([34] - 17 cases; [35] - 5 cases; [36] - 10 cases; [37] - 5 cases; compared with [38] - 36 cases) and the results are inconclusive [39]. We recently completed the analysis of two concurrent case-control studies conducted in the United States (Los Angeles) and in China (Shanghai) [40]. Overall cases had 13.5% higher serum estradiol concentrations with a caseto-control excess of 16.6% in Chinese women and 10.8% in United States white women. Los Angeles controls had 20.6% greater estradiol concentrations than Shanghai controls and adjustment for body weight accounted for only 25.7% of this dif-
A model of breast cancer pathogenesis
Pike and coworkers [24] developed a model of breast cancer incidence that incorporates all of the reproductive and endocrine risk factors and provides an excellent fit to the actual age-specific incidence curves for breast cancer in different populations. This model is based on the concept that breast cancer incidence does not increase proportionally with calendar age, but rather with breast tissue age raised to the power 4.5. The concept of breast tissue age is closely associated with the cell kinetics of breast tissue stem cells, which, in turn, are closely associated with exposure of breast tissue to ovarian hormones. The model predicts relative risks that are remarkably consistent with the observed values for each of these risk factors, including the two contradictory effects of pregnancy. This model allows us
S16
BE Henderson and L Bernstein
to explore the degree to which variations in these risk factors among different populations may explain the large international variation in breast cancer rates. As of 1970, age-adjusted incidence rates of breast cancer were some five to six times higher in the United States than in Japan. Data on average ages at menarche, first birth, and menopause among Japanese are available from a 1970 survey [42]. The average age at natural menopause of Japanese women is similar to that of American women for these age groups, but fewer Japanese women have had a surgical menopause. The data on menarche favor a lower breast cancer incidence rate in Japanese women: older Japanese women had a much later menarche than American women, but the data on age at first birth and nulliparity show that none of the decreased breast cancer rates in Japan can be attributed to these factors. Japanese breast cancer rates remain almost constant after age 50 [43]. In model terms, this implies that no further breast tissue aging occurs in Japanese women in the postmenopausal period. This is, in all probability, a reflection of body weight. In 1970 the average weight of postmenopausal women was less than 50 kg and these women were unlikely to have been producing significant amounts of estrogen. The model-predicted breast cancer incidence rates of Japanese women, which allow for their actual distribution of the established breast cancer risk factors, are still considerably lower than the observed rates for United States whites, whereas they are between 2.4- and 3.8-fold greater than the observed Japanese rates. The very late menarche in Japanese women would be expected to result in a substantial delay in the establishment of regular menstrual cycles compared with United States women. Allowing for this delay and incorporating the differences in average cycle length and the lower estrogen levels actually achieved during a typical ovulatory cycle in Japanese women compared with United States white women [32], the predicted Japanese rates are essentially identical to those observed in United States whites.
References 1. WHO IARC: Cancer Incidence in Five Continents, vol 5. Muir C, Waterhouse J, Mack T, et al. (eds): Lyon, IARC Sci Pub188, 1987, p 882 2. Haenszel W, Kurihara M: Studies of Japanese migrants. I. Mortality from cancer and other diseases among Japanese in the United States. J Natl Cancer Inst 40: 43--68, 1968 3. Buell P: Changing incidence of breast cancer in JapaneseAmerican women. J Natl Cancer Inst 51: 1479-1483, 1973 4. Dunn J: Breast cancer among American Japanese in the San Francisco Bay area. Natl Cancer Inst Monogr 47: 157, 1977 5. Henderson BE, Pike MC, Casagrande JT: Breast cancer and the estrogen window hypothesis. Lancet 2: 363-364, 1981 6. Henderson BE, Gerkins V, Rosario I, Casagrande JT, Pike MC: Elevated serum levels of estrogen and prolaetin in daughters of patients with breast cancer. N Engl J Med 293: 790-795, 1975 7. Trichopoulos D, Brown JB, Garas J: Elevated urine estrogen and pregnanediol levels in daughters of breast cancer patients. J Natl Cancer Inst 67: 603-606, 1981 8. Henderson BE, Ross RK, Judd HL, Krailo MD, Pike MC: Do regular ovulatory cycles increase breast cancer risk? Cancer 56: 1206-1208, 1985 9. MacMahon B, Trichopoulos D, Brown J, Anderson AP, Aoki K, et al.: Age at menarche, probability of ovulation and breast cancer risk. Int J Cancer 29: 13--16, 1982 10. A p t e r D , V i h k o R: Earlymenarche, arisk factor forbreast cancer, indicates early onset of ovulatory cycles. J Clin Endocrinol Metab 57: 82-86, 1983 11. Apter D, Reinila M, Vihko R: Some endocrine characteristics of early menarche, a risk factor for breast cancer, are preserved into adulthood. Int J Cancer 44: 783-787, 1989 12. Tanner JM: Growth at Adolescence. Biackwell Scientific, Oxford, 1962, pp 1-27 13. Frisch RE, McArthur J. Menstrual cycles: fatness as a determinant of minimum weight for height necessary for their maintenance or onset. Science 185: 949-951, 1974 14. Frisch RE, Gotz-Welbergen AV, McArthur JW, Albright T, Witschi J, et al: Delayed menarche and amenorrhea of college athletes in relation to age at onset of training. JAMA 246: 1559-1563, 1981 15. Bernstein L, Ross R, Lobo R, Haniseh R, Krailo MD, Henderson BE: The effects of moderate physical activity on menstrual cycle patterns in adolescence: Implications for breast cancer prevention. Br J Cancer 55: 681-685, 1987 16. Trichopoulos D, MacMahon B, Cole P: The menopause and breast cancer risk. J Natl Cancer Inst 48: 605-613, 1972 17. Feinleib M: Breast cancer and artificial menopause: A cohort study. J Natl Cancer Inst 41: 315-329, 1968 18. MacMahon B, Cole P, Lin TM, Lowe CR, Mirra AP, et al.: Age at first birth and cancer of the breast. A summary of an international study. Bull WHO 43: 209-221, 1970 19. Layde PM, Webster LA, Baughman AL, Wingo PA, Rubin
I n t e r n a t i o n a l v a r i a t i o n in b r e a s t c a n c e r rates
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
GL, Ory HW: The independent associations of parity, age at first full term pregnancy, and duration of breastfeeding with the risk of breast cancer. Cancer and Steroid Hormone Study Group. J Clin Epidemiol 42: 963-973, 1989 Yuan JM, Yu MC, Ross RK, Gao YT, Henderson BE: Risk factors for breast cancer in Chinese women in Shanghai. Cancer Res 48: 1949-1953, 1988 Bruzzi P , Negri E, La Vecchia C, et al.: Short term increase in risk of breast cancer after full term pregnancy. Br Med J 297: 1096-1098, 1988 Pike MC, Henderson BE, Casagrande JT, Rosario I, Gray GE: Oral contraceptive use and early abortion as risk factors for breast cancer in young women. Br J Cancer 43: 72-76, 1981 Hadjimichael OC, Boyle CA, Meigs JW: Abortion before first live birth and risk of breast cancer. Br J Cancer 53: 281-284, 1986 Pike MC, Krailo MC, Henderson BE, Casagrande JT, Hoel DG: 'Hormonal' risk factors, 'breast tissue age' and the age-incidence of breast cancer. Nature 303: 767-770, 1983 Bernstein L, Depue RH, Ross RK, Judd HL, Pike MC, Henderson BE: Higher maternal levels of free estradiol in first compared to second pregnancy: Early gestational differences. J Natl Cancer Inst 76: 1035-1039, 1986 YuMC, GerkinsVR, Henderson BE, BrownJB,PikeMC: Elevated levels of prolactin in nulliparous women. Br J Cancer 43: 826-831, 1981 Musey VC, Collins DC, Musey PI, Martino-Saltzman D, Preedy JR: Long-term effect of a first pregnancy on the secretion of prolactin. N Engl J Med 316: 229-234, 1987 Bernstein L, Pike MC, Ross RK, Judd HL, Brown JB, Henderson BE: Estrogen and sex hormone-binding globulin levels in nulliparous and parous women. J Natl Cancer Inst 74: 741-745, 1985 de Waard F, Cornelis JP, Aoki K, Yoshida M: Breast cancer incidence according to weight and height in two cities of the Netherlands and in Aichi Prefecture, Japan. Cancer 40: 1269-1275, 1977 Siiteri PK, Hammond GL, Nisker JA: Increased availability of serum estrogens in breast cancer: A new hypothesis. In Pike MC, Siiteri PK, Welsch CW (eds) Hormones and Breast Cancer. Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, 1981, p 87-101
S17
31. Henderson BE, Ross RK, Bernstein L: Estrogens as a cause of human cancer: The Richard and Hinda Rosenthal Foundation Award Lecture. Cancer Res 48: 246-253, 1988 32. MacMahon B, Cole P, Brown JB, Aoki K, Lin TM, et al.: Urine estrogen profiles of Asian and North American women. Int J Cancer 14: 161-167, 1974 33. Parkin DM, Stjernsward J, Muir CS: Estimates of the worldwide frequency of twelve major cancers. Bull WHO 62: 163-182, 1984 34. Bruning PF, Bronfrer JMG, Hart AAM: Non-protein bound oestradiol, sex hormone binding globulin, breast cancer and breast cancer risk. Br J Cancer 51: 479-484, 1985 35. Drafta D, Schindler AF, Milcu M, Keller E, Stroe E, etal.: Plasma hormones in pre- and postmenopausal breast cancer. J Steroid Biochem 13: 793-802, 1980 36. England PC, Skinner LG, Cottrell KM, Sellwood RA: Serum oestradiol-17~ in women with benign and malignant breast disease. Br J Cancer 30: 571-576, 1974 37. Malarkey WB, Schroeder LL, Stevens VC, James AG, Lanese RR: Twenty-four-hour preoperative endocrine profiles in women with benign and malignant breast disease. Cancer Res 37: 4655-4659, 1977 38. Meyer F, Brown JB, Morrison AS, MacMahon B: Endogenous sex hormones, prolactin, and breast cancer in premenopausal women. J Natl Cancer Inst 77: 613-616, 1986 39. Key TJA, Pike MC: The role of oestrogens and progestagens in the epidemiology and prevention of breast cancer. Eur J Cancer Clin Oncol 24: 29-43, 1988 40. Bernstein L, Yuan JM, Ross RK, Pike MC, Lobo R, etal.: Serum hormone levels in premenopausal Chinese women in Shanghai and white women in Los Angeles: results from two breast cancer case-control studies. (submitted) 41. Shimizu H, Ross RK, Bernstein L, Pike MC, Henderson BE: Serum estrogen levels in postmenopausal women: comparison of US whites and Japanese in Japan. Br J Cancer (in press) 42. Hoel DG, Wakabayashi T, Pike MC: Secular trends in the distribution of the breast cancer risk factors: Menarche, first birth, menopause and weight in Hiroshima and Nagasaki, Japan. Am J Epidemiol 118: 78-89, 1983 43. Fujimoto I, Hanai A, Oshima A: Descriptive epidemiology of cancer in Japan: Current cancer incidence and survival data. NCI Monograph 53: 5-15, 1979
The international variation in breast cancer rates: An epidemiological assessment
Brian E. Henderson and Leslie Bernstein
Department of Preventive Medicine, University of Southern California School of Medicine, Norris Cancer Hospital and Research Institute, 1441 Eastlake Avenue, Los Angeles, California 90033, USA
Key words: breast cancer, geographic variation, estrogen production, hormonal exposure, incidence, risk factors Abstract
Part of the international differences in breast cancer incidence rates can be explained by geographic variation in reproductive and other breast cancer risk factors. Age at menarche and age at onset of regular ovulatory menstrual cycles are two such factors; both vary across populations directly according to breast cancer risk, and both are acknowledged as breast cancer risk factors. Consideration of the body of evidence on these factors, as well as that on age at menopause, suggests that the cumulative frequency of ovulatory menstrual cycles is a critical determinant of breast cancer risk. Although age at first term pregnancy explains the majority of the protective effect of parity on breast cancer risk, two recent studies have demonstrated a small residual protective effect of increasing number of births. It appears that pregnancy has paradoxical effects on breast cancer risk in terms of hormone production and metabolism. The initial effect is an increased risk associated with first trimester estrogen exposure. However, the hormonal consequences of completing the pregnancy counteract this negative effect of early pregnancy. The effect of body weight, a breast cancer risk factor for postmenopausal women, can be explained in terms of increased extraglandular conversion of androstenedione to estrone. Further evidence supporting a pathogenic role of estrogens in the development of breast cancer comes from international studies of endogenous hormones in populations with differing risks of breast cancer. These risk factors have been incorporated into a mathematical model which is based on the concept that breast tissue ages according to hormonal (primarily estrogen) exposure; this model closely predicts the incidence rates throughout the world.
There is remarkable variation in the breast cancer incidence rates of different countries [1]. Rates are nearly six times higher in the United States, Canada, or Northern Europe than in Asia or black Africa. These differences in breast cancer incidence rates are not determined by variation in genetic susceptibility [2]. The breast cancer incidence rates of American blacks are very similar to those of
American whites, but differ from those of African blacks. Furthermore, Japanese migrants to Hawaii and California have higher rates of breast cancer than women in Japan [3, 4]. Native Japanese women who migrate to the United States as young adults have a small increase in their breast cancer rates while living in the United States, whereas Japanese women born in the United States have
Address for offprints: B.E. Henderson, Department of Preventive Medicine, University of Southern California School of Medicine, Norris Cancer Hospital and Research Institute, 1441 Eastlake Avenue, Los Angeles, California 90033, USA
S12
BE Henderson and L Bernstein
rates approaching those of their white counterparts. This suggests that the risk of breast cancer is at least partly related to exposures that are acquired early in life. These large international differences in breast cancer incidence can be at least partially explained by geographic variation in breast cancer risk factors.
Age at menarche Early age at menarche has been demonstrated as a risk factor for breast cancer in most case-control studies. In general, breast cancer risk decreases approximately 20 per cent for each year of age that menarche is delayed. In a case-control study of young women, we recorded not only age at onset of menstruation but also age when 'regular' (that is, predictable) menstruation was first established [5]. For a fixed age at menarche, women who established regular menstrual cycles within one year of their first menstrual period had more than double the risk of breast cancer of women who had a five-year or longer delay in establishing regular menses. For women with early menarche (aged 12 or younger) and rapid establishment of regular menstrual cycles (less than one year), breast cancer risk was nearly fourfold greater than that of women with late menarche (aged 13 or older) and long duration of irregular cycles (five or more years). Because the cumulative amount of estrogen produced in the normal luteal phase of an ovulatory cycle is greater than that produced during the comparable period of a nonovulatory cycle, the cumulative frequency of ovulatory cycles serves as an index of cumulative estrogen (as well as progesterone) exposure. Results from an earlier study of circulating homrone levels in daughters of breast cancer cases and in age-matched daughters of controls are consistent with this hypothesis. The daughters of the breast cancer cases, who as a group have at least twice the breast cancer risk of the general population, had higher levels of circulating estrogen and progesterone during the luteal phase of the menstrual cycle (on day 22) than did controls [6]. This result has been confirmed by other investigators [7].
Other supportive evidence for the concept that the cumulative number of ovulatory cycles (that is, cumulative estrogen exposure) is a major determinant of breast cancer risk [8] comes from international studies of the frequency of ovulation in relation to age at menarche and number of years since menarche [9]. In girls aged 15 to 19 years, selected from several populations at varying risk of breast cancer, those with later menarche were more likely to have anovular menstrual cycles than girls with early menarche, given the same number of elapsed years since menarche. Adjusting for years since menarche, the highest frequency of ovulatory menstrual cycles was observed in those populations with the highest breast cancer rates. Apter and Vihko [10], in a longitudinal study of 200 Finnish schoolgirls, also found that those with early menarche establish ovulatory menstrual cycles more quickly than girls with later onset of menstruation. The average intervals from menarche until 50 per cent of cycles were ovulatory were 1, 3, and 4.5 years, respectively, for girls with menarche before age 12, between age 12 and 13, and at age 13 years and older. Forty-four of these girls were later followed into the third decade of life and evaluated endocrinologically [11]. Women with early menarche (before age 12) had higher serum estradiol concentrations during the follicular phase of the menstrual cycle (at average age 27) than did women with later menarche. Furthermore, the serum estradiol concentrations increased more rapidly to the mid-cycle peak in these early-menarche women than in those with later menarche. In addition, follicular phase sex hormone-binding globulin (SHBG) concentrations were about 30 per cent lower in women with early menarche. Over the past 100 years, the age at menarche has progressively decreased both in the United States and in most other areas of the world. A series of extensive cross-sectional studies has demonstrated that age at menarche is directly related to childhood growth patterns [12]. Attainment of a critical body weight-to-height ratio appears necessary for menarche to occur [13]. Chronic malnutrition during childhood delays the age at menarche, whereas the improved nutrition and control of infectious
International variation in breast cancer rates
diseases of childhood of the past decades have combined to lower it. Strenuous physical activity may also delay menarche. Girls who engage in regular ballet dancing, swimming, or running may have a considerable delay in the onset of menses. In one study, ballet dancers had a mean age at menarche of 15.4 years, compared with 12.5 years for controls [14]. Breast development was also delayed in the dancers, and intermittent amenorrhea persisted throughout their teenage years as long they remained active dancers. Even moderate physical activity during adolescence can lead to anovular menstrual cycles. In a study of menstrual cycle patterns during adolescence, we found that girls (aged 14 to 17) who engaged in moderate physical activity (averaging 600 kcal of energy expenditure per week) were 2.9 times more likely to be anovular than were girls engaging in lesser amounts of physical activity [15].
Age at menopause The relationship between age at menopause and breast cancer risk has been known for some time. The rate of increase in the age-specific incidence rate of breast cancer slows sharply at the time of menopause and the rate of increase in the postmenopausal period is only about one-sixth the rate of increase in the premenopausal period [1]. It has been estimated that women whose natural menopause occurs before age 45 have only one-half the breast cancer risk of those who menopause occurs after age 55 [16]. Another way of expressing this result is that women with 40 or more years of active menstruation have twice the breast cancer risk of those with fewer than 30 years of menstrual activity. Artificial menopause, by either bilateral oophorectomy or pelvic irradiation, also markedly reduces breast cancer risk. The effect appears to be just slightly greater than that of natural menopause (defined as cessation of menstrual periods). As shown by Feinleib [17], there is little change in a woman's risk of breast cancer following a unilateral oophorectomy or a simple hysterectomy.
S13
Age at first full-term pregnancy Two of the earliest known and most reproducible features of breast cancer epidemiology are the decreased risk associated with increased parity and the increased risk of single women. MacMahon and colleagues [18] made a major advance in our understanding of the role of pregnancy in altering breast cancer risk through their analysis of an international collaborative case-control study. Single and nulliparous married women were found to have the same increased risk of breast cancer, approximately 1.4 times the risk of parous married women. Among married women in each country, parous breast cancer patients had fewer children than had parous women in the control groups. These authors clearly demonstrated, however, that this protective effect of parity was totally explained by a protective effect of early age at first birth. Those women with a first birth before age 20 had about one-half the risk of nulliparous women. Controlling for age at first birth, subsequent births provided no further protection against breast cancer. Recently, two studies conducted in other populations have observed small residual protective effects of an increasing number of births [19, 20]; this suggests that there may be certain circumstances in which multiparity does offer some further protection. In a recent study in Shanghai, we observed that there was a protective effect of multiple pregnancies, most notably after the fifth pregnancy [20]. The main protective effect is, however, undoubtedly associated with the first full-term pregnancy. Women who have a late first full-term pregnancy are actually at elevated risk of breast cancer compared with nulliparous women [18]. This paradoxical effect of a late first full-term pregnancy has been repeatedly confirmed by other studies. A possible explanation for this effect is suggested by somewhat related observations on breast cancer risk immediately following a full-term pregnancy and following an uncompleted first pregnancy. In a hospital-based case-control study, it was demonstrated that women who gave birth during the three years before interview had higher risk of breast cancer than those whose last birth occurred at least ten years earlier (after adjustment for age, age at
$14
B E Henderson and L Bernstein
first birth and parity, relative risk = 2.7) [21]. Recently, we found that a first-trimester abortion, whether spontaneous or induced, occurring before the first full-term pregnancy was associated with higher risk of breast cancer [22]. This observation has been confirmed in a cohort of Connecticut women who had only one live birth [23]. In that study, abortions after the first full-term pregnancy did not carry any increased risk of breast cancer. Based on these results, we have concluded that there are two contradictory effects of pregnancy on breast cancer risk which are particularly notable in the first pregnancy [24]. This apparent paradox has a physiological explanation based on patterns of estrogen and prolactin secretion and metabolism during pregnancy. During the first trimester of pregnancy, there is a rapid rise in the level of 'free' estradiol (that is, the fraction of estradiol that is biologically available to act upon tissues), an effect that is more apparent in the first rather than subsequent pregnancies [25]. Thus, the net effect of this early part of pregnancy, in terms of estrogen exposure to the breast, is an increase in risk that is equivalent to the exposure from several ovulatory cycles over a relatively short time period. In the long run, however, this negative effect of early pregnancy on breast cancer risk can be overridden by two beneficial, hormonal consequences of completing the pregnancy. Several years ago, we reported that prolactin levels were substantially lower in parous compared with nulliparous women [26], an observation that was replicated recently [27]. In addition, we found that parous women had higher levels of SHBG and lower levels of free estradiol than their nulliparous counterparts [28]. Lactation has not been clearly established to either enhance or protect against development of breast cancer. If the cumulative number of ovulatory cycles is directly related to risk of breast cancer, a beneficial effect of long duration of nursing would be expected because nursing results in a substantial delay in the reestablishment of ovulation following a completed pregnancy. Because of a small proportion of mothers with a large cumulative number of nursing months, most previous epidemiologic studies have not allowed for precise estimates of the effects of lactation on breast cancer
risk. We recently completed a population-based case-control study in China, a population in which long-duration nursing is the norm [20]. In this study, a 30 per cent reduction in breast cancer risk was observed for each five years of nursing experience.
Weight In addition to the menstrual and reproductive risk factors described above, there is a strong relationship between weight and risk of breast cancer. The relationship is critically dependent on age. For women under age 50, there is little or no increased risk associated with increased weight, but by age 60, a 10-kg increase in weight results in approximately an 80 per cent increase in breast cancer risk [29]. Whether this weight effect is one of excess weight (body fat) or absolute weight as such is inclear. Contradictory results have been reported on whether, for example, Quetelet's index (wt/ht2) is correlated with breast cancer risk. Absolute weight appears to be as good an indicator of risk as any function of weight and height. In the postmenopausal period, the major source of estrogen is extraglandular (largely adipose tissue) conversion of the adrenal androgen androstenedione to estrone [30]. The association of breast cancer with cyclic ovarian activity implies that estrogen is important in the pathogenesis of this disease. We have discussed the rationale that hormones, and in particular, estrogen, can directly increase the incidence of breast cancer [31]. The most carefully done international studies comparing estrogen levels in populations at differing risks of breast cancer support the pathogenic role of estrogens. In the early 1970s, MacMahon and his colleagues [32] conducted a series of studies on teenagers and young women to investigate whether some aspect of estrogen metabolism was responsible for the large differences in breast cancer rates between Asia and North America. They knew that the Asian women chosen for study should be a group who would themselves be expected to experience low breast cancer rates, so
International variation in breast cancer rates
they chose study populations as similar as possible to women than in the breast cancer age range. Although they initially reported their results in terms of the relative amounts of various urinary estrogens, their results can also be evaluated in terms of absolute total urinary estrogen levels (which to some degree probably reflect serum estrogen levels) [33]. They found that, in overnight urines collected on the morning of day 21 of the menstrual cycle, total urinary estrogen levels were 36 per cent higher in the North American teenagers. Levels were also elevated in samples collected on the morning of day 10, but only by 9 per cent. These differences might be related to later menarche in the Japanese women, but even larger differences were found in older women, for whom the effects of late menarche should have disappeared. In nulliparous women aged 20 to 24, total urinary estrogen levels in the North American women were 49 per cent higher on day 21 and 38 per cent higher on day 10 (during the follicular phase of the menstrual cycle); similar differences were found among parous women aged 30 to 39. These differences in total urinary estrogen are presumably a reflection on both reduced frequency of ovulatory cycles, and less effective corpus luteum formation in those cycles for which ovulation did
S15
ference. These higher levels of estradiol in white compared with Chinese women could be an important part of the explanation of the twofold to threefold differences in the breast cancer risk of young women in the United States and in Asia. Greater hormonal differences are observed in comparisons of postmenopausal Asian and United States white women. Recently we studied serum estrogen levels in postmenopausal Japanese women in Japan and postmenopausal white women in Los Angeles [41]. The Japanese women studied were deliberately chosen to be from a rural agricultural area in order to get samples which represented as closely as possible the traditional Japanese 'lifestyle' that gave rise to the low rates of breast cancer in Japan. Levels of estrone and of estradiol were both significantly higher (47 per cent and 36 per cent, respectively) in the sera of white women and adjustment for differences in body weight accounted for little of these differences. After the menopause, there is no increase in the age-specific breast cancer incidence rates of Japanese women as there is for white women in the United States. These decreased levels of serum estrogens could well explain this further divergence in the breast cancer incidence rates.
Occur.
On the hypothesis that increased ovarian activity increases the risk of breast cancer, one would predict that breast cancer case-control studies evaluating luteal-phase circulating levels of estradiol would find evidence of greater production in the cases. Five such studies have been undertaken, but four of them had small numbers of cases ([34] - 17 cases; [35] - 5 cases; [36] - 10 cases; [37] - 5 cases; compared with [38] - 36 cases) and the results are inconclusive [39]. We recently completed the analysis of two concurrent case-control studies conducted in the United States (Los Angeles) and in China (Shanghai) [40]. Overall cases had 13.5% higher serum estradiol concentrations with a caseto-control excess of 16.6% in Chinese women and 10.8% in United States white women. Los Angeles controls had 20.6% greater estradiol concentrations than Shanghai controls and adjustment for body weight accounted for only 25.7% of this dif-
A model of breast cancer pathogenesis
Pike and coworkers [24] developed a model of breast cancer incidence that incorporates all of the reproductive and endocrine risk factors and provides an excellent fit to the actual age-specific incidence curves for breast cancer in different populations. This model is based on the concept that breast cancer incidence does not increase proportionally with calendar age, but rather with breast tissue age raised to the power 4.5. The concept of breast tissue age is closely associated with the cell kinetics of breast tissue stem cells, which, in turn, are closely associated with exposure of breast tissue to ovarian hormones. The model predicts relative risks that are remarkably consistent with the observed values for each of these risk factors, including the two contradictory effects of pregnancy. This model allows us
S16
BE Henderson and L Bernstein
to explore the degree to which variations in these risk factors among different populations may explain the large international variation in breast cancer rates. As of 1970, age-adjusted incidence rates of breast cancer were some five to six times higher in the United States than in Japan. Data on average ages at menarche, first birth, and menopause among Japanese are available from a 1970 survey [42]. The average age at natural menopause of Japanese women is similar to that of American women for these age groups, but fewer Japanese women have had a surgical menopause. The data on menarche favor a lower breast cancer incidence rate in Japanese women: older Japanese women had a much later menarche than American women, but the data on age at first birth and nulliparity show that none of the decreased breast cancer rates in Japan can be attributed to these factors. Japanese breast cancer rates remain almost constant after age 50 [43]. In model terms, this implies that no further breast tissue aging occurs in Japanese women in the postmenopausal period. This is, in all probability, a reflection of body weight. In 1970 the average weight of postmenopausal women was less than 50 kg and these women were unlikely to have been producing significant amounts of estrogen. The model-predicted breast cancer incidence rates of Japanese women, which allow for their actual distribution of the established breast cancer risk factors, are still considerably lower than the observed rates for United States whites, whereas they are between 2.4- and 3.8-fold greater than the observed Japanese rates. The very late menarche in Japanese women would be expected to result in a substantial delay in the establishment of regular menstrual cycles compared with United States women. Allowing for this delay and incorporating the differences in average cycle length and the lower estrogen levels actually achieved during a typical ovulatory cycle in Japanese women compared with United States white women [32], the predicted Japanese rates are essentially identical to those observed in United States whites.
References 1. WHO IARC: Cancer Incidence in Five Continents, vol 5. Muir C, Waterhouse J, Mack T, et al. (eds): Lyon, IARC Sci Pub188, 1987, p 882 2. Haenszel W, Kurihara M: Studies of Japanese migrants. I. Mortality from cancer and other diseases among Japanese in the United States. J Natl Cancer Inst 40: 43--68, 1968 3. Buell P: Changing incidence of breast cancer in JapaneseAmerican women. J Natl Cancer Inst 51: 1479-1483, 1973 4. Dunn J: Breast cancer among American Japanese in the San Francisco Bay area. Natl Cancer Inst Monogr 47: 157, 1977 5. Henderson BE, Pike MC, Casagrande JT: Breast cancer and the estrogen window hypothesis. Lancet 2: 363-364, 1981 6. Henderson BE, Gerkins V, Rosario I, Casagrande JT, Pike MC: Elevated serum levels of estrogen and prolaetin in daughters of patients with breast cancer. N Engl J Med 293: 790-795, 1975 7. Trichopoulos D, Brown JB, Garas J: Elevated urine estrogen and pregnanediol levels in daughters of breast cancer patients. J Natl Cancer Inst 67: 603-606, 1981 8. Henderson BE, Ross RK, Judd HL, Krailo MD, Pike MC: Do regular ovulatory cycles increase breast cancer risk? Cancer 56: 1206-1208, 1985 9. MacMahon B, Trichopoulos D, Brown J, Anderson AP, Aoki K, et al.: Age at menarche, probability of ovulation and breast cancer risk. Int J Cancer 29: 13--16, 1982 10. A p t e r D , V i h k o R: Earlymenarche, arisk factor forbreast cancer, indicates early onset of ovulatory cycles. J Clin Endocrinol Metab 57: 82-86, 1983 11. Apter D, Reinila M, Vihko R: Some endocrine characteristics of early menarche, a risk factor for breast cancer, are preserved into adulthood. Int J Cancer 44: 783-787, 1989 12. Tanner JM: Growth at Adolescence. Biackwell Scientific, Oxford, 1962, pp 1-27 13. Frisch RE, McArthur J. Menstrual cycles: fatness as a determinant of minimum weight for height necessary for their maintenance or onset. Science 185: 949-951, 1974 14. Frisch RE, Gotz-Welbergen AV, McArthur JW, Albright T, Witschi J, et al: Delayed menarche and amenorrhea of college athletes in relation to age at onset of training. JAMA 246: 1559-1563, 1981 15. Bernstein L, Ross R, Lobo R, Haniseh R, Krailo MD, Henderson BE: The effects of moderate physical activity on menstrual cycle patterns in adolescence: Implications for breast cancer prevention. Br J Cancer 55: 681-685, 1987 16. Trichopoulos D, MacMahon B, Cole P: The menopause and breast cancer risk. J Natl Cancer Inst 48: 605-613, 1972 17. Feinleib M: Breast cancer and artificial menopause: A cohort study. J Natl Cancer Inst 41: 315-329, 1968 18. MacMahon B, Cole P, Lin TM, Lowe CR, Mirra AP, et al.: Age at first birth and cancer of the breast. A summary of an international study. Bull WHO 43: 209-221, 1970 19. Layde PM, Webster LA, Baughman AL, Wingo PA, Rubin
I n t e r n a t i o n a l v a r i a t i o n in b r e a s t c a n c e r rates
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
GL, Ory HW: The independent associations of parity, age at first full term pregnancy, and duration of breastfeeding with the risk of breast cancer. Cancer and Steroid Hormone Study Group. J Clin Epidemiol 42: 963-973, 1989 Yuan JM, Yu MC, Ross RK, Gao YT, Henderson BE: Risk factors for breast cancer in Chinese women in Shanghai. Cancer Res 48: 1949-1953, 1988 Bruzzi P , Negri E, La Vecchia C, et al.: Short term increase in risk of breast cancer after full term pregnancy. Br Med J 297: 1096-1098, 1988 Pike MC, Henderson BE, Casagrande JT, Rosario I, Gray GE: Oral contraceptive use and early abortion as risk factors for breast cancer in young women. Br J Cancer 43: 72-76, 1981 Hadjimichael OC, Boyle CA, Meigs JW: Abortion before first live birth and risk of breast cancer. Br J Cancer 53: 281-284, 1986 Pike MC, Krailo MC, Henderson BE, Casagrande JT, Hoel DG: 'Hormonal' risk factors, 'breast tissue age' and the age-incidence of breast cancer. Nature 303: 767-770, 1983 Bernstein L, Depue RH, Ross RK, Judd HL, Pike MC, Henderson BE: Higher maternal levels of free estradiol in first compared to second pregnancy: Early gestational differences. J Natl Cancer Inst 76: 1035-1039, 1986 YuMC, GerkinsVR, Henderson BE, BrownJB,PikeMC: Elevated levels of prolactin in nulliparous women. Br J Cancer 43: 826-831, 1981 Musey VC, Collins DC, Musey PI, Martino-Saltzman D, Preedy JR: Long-term effect of a first pregnancy on the secretion of prolactin. N Engl J Med 316: 229-234, 1987 Bernstein L, Pike MC, Ross RK, Judd HL, Brown JB, Henderson BE: Estrogen and sex hormone-binding globulin levels in nulliparous and parous women. J Natl Cancer Inst 74: 741-745, 1985 de Waard F, Cornelis JP, Aoki K, Yoshida M: Breast cancer incidence according to weight and height in two cities of the Netherlands and in Aichi Prefecture, Japan. Cancer 40: 1269-1275, 1977 Siiteri PK, Hammond GL, Nisker JA: Increased availability of serum estrogens in breast cancer: A new hypothesis. In Pike MC, Siiteri PK, Welsch CW (eds) Hormones and Breast Cancer. Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, 1981, p 87-101
S17
31. Henderson BE, Ross RK, Bernstein L: Estrogens as a cause of human cancer: The Richard and Hinda Rosenthal Foundation Award Lecture. Cancer Res 48: 246-253, 1988 32. MacMahon B, Cole P, Brown JB, Aoki K, Lin TM, et al.: Urine estrogen profiles of Asian and North American women. Int J Cancer 14: 161-167, 1974 33. Parkin DM, Stjernsward J, Muir CS: Estimates of the worldwide frequency of twelve major cancers. Bull WHO 62: 163-182, 1984 34. Bruning PF, Bronfrer JMG, Hart AAM: Non-protein bound oestradiol, sex hormone binding globulin, breast cancer and breast cancer risk. Br J Cancer 51: 479-484, 1985 35. Drafta D, Schindler AF, Milcu M, Keller E, Stroe E, etal.: Plasma hormones in pre- and postmenopausal breast cancer. J Steroid Biochem 13: 793-802, 1980 36. England PC, Skinner LG, Cottrell KM, Sellwood RA: Serum oestradiol-17~ in women with benign and malignant breast disease. Br J Cancer 30: 571-576, 1974 37. Malarkey WB, Schroeder LL, Stevens VC, James AG, Lanese RR: Twenty-four-hour preoperative endocrine profiles in women with benign and malignant breast disease. Cancer Res 37: 4655-4659, 1977 38. Meyer F, Brown JB, Morrison AS, MacMahon B: Endogenous sex hormones, prolactin, and breast cancer in premenopausal women. J Natl Cancer Inst 77: 613-616, 1986 39. Key TJA, Pike MC: The role of oestrogens and progestagens in the epidemiology and prevention of breast cancer. Eur J Cancer Clin Oncol 24: 29-43, 1988 40. Bernstein L, Yuan JM, Ross RK, Pike MC, Lobo R, etal.: Serum hormone levels in premenopausal Chinese women in Shanghai and white women in Los Angeles: results from two breast cancer case-control studies. (submitted) 41. Shimizu H, Ross RK, Bernstein L, Pike MC, Henderson BE: Serum estrogen levels in postmenopausal women: comparison of US whites and Japanese in Japan. Br J Cancer (in press) 42. Hoel DG, Wakabayashi T, Pike MC: Secular trends in the distribution of the breast cancer risk factors: Menarche, first birth, menopause and weight in Hiroshima and Nagasaki, Japan. Am J Epidemiol 118: 78-89, 1983 43. Fujimoto I, Hanai A, Oshima A: Descriptive epidemiology of cancer in Japan: Current cancer incidence and survival data. NCI Monograph 53: 5-15, 1979
