Prevention of Cancers of the Breast, Endometrium and Ovary

Prevention of cancers of the breast, endometrium and ovary

Malcolm C Pike*,1, Celeste Leigh Pearce1 and Anna H Wu1

1USC/Norris Cancer Center, Los Angeles, CA, USA

A central epidemiological feature of cancers of the breast, in US white females in 1969–1971 before mammo- endometrium and ovary is the sharp slowing down in their graphic screening and widespread use of menopausal rate of increase with age around the time of menopause. estrogen–progestin therapy (EPT) had distorted the The incidence of these tumors by the age of 70 years would picture. In contrast to the curve shown in Figure 1, there be between fourfold and eightfold increased if the rapid is an abrupt slowing down of the rate of increase in increase with age seen in young women continued into old incidence around the age of 50 years. The age-specific age. These phenomena can be explained by the different incidence curves for endometrial and ovarian cancer effects of ovarian hormones on cell division rates in the (Figures 3 and 4) show a similar slowing down of the relevant tissues. Models of these effects provide a rate of increase around the same age. These incidence plausible explanation of most of the known epidemiology curves can be made to fit into the form of Equation (1) of each of the cancers, including the increase in breast by expressing the equation in a more general form: cancer risk from menopausal estrogen–progestin therapy. Some recent epidemiological findings in endometrial and ovarian cancer suggest new avenues for possible chemo- IðtÞ¼a½dðtÞk ð2Þ prevention of these cancers. Oncogene (2004) 23, 6379–6391. doi:10.1038/sj.onc.1207899 where d(t) is the ‘effective tissue age’ of the relevant Keywords: prevention; breast cancer; endometrial can- tissue up to age t. At a given age, the rate of ‘tissue cer; ovarian cancer ageing’ is the rate at which underlying carcinogenic processes such as mutation, proliferation and cell death take place. We assume that it is roughly equal to mitotic rate, or perhaps cell death and resulting replacement Introduction (Cairns, 2002), in some stem cell compartment. For lung cancer, the ‘effective tissue age’ is almost exactly An accelerating increase in incidence with increasing age duration of smoking up to age t, as both early and late is a striking feature of most adult cancers. The relation- stages in lung carcinogenesis are strongly affected by ship between incidence, I(t), and age, t, for these cancers tobacco (Doll, 1971). For hormone-dependent tissues follows quite closely a simple power relationship: ‘effective tissue age’ probably represents at least approximately the cumulative number of stem cell k divisions up to age t. The hypothesis is simply that IðtÞ¼at ð1Þ hormones affect cancer incidence mainly through their effect on mitotic rates in the stem cell compartment, where a is a constant and k is usually between 4 and 5. both by increasing the probability of a DNA-damaging For these cancers, a plot of the logarithm of incidence event becoming fixed as a mutation (Dao, 1981) and against the logarithm of age produces a straight line of through their promoting effect seen in certain animal slope 4 or 5. Figure 1 shows this relationship for experimental systems (Welsch, 1981). This can be colorectal cancer (Cutler and Young, 1975). This almost expressed mathematically through the ‘effective tissue exponential increase in cancer incidence with increasing age’ (d(t) in Equation (2)) to give models that account age can be attributed to a multistep carcinogenic for many of the known risk factors of hormone- progression in which the underlying processes, including dependent cancers (Henderson et al., 1982). mutation, proliferation and regression of intermediate In these terms, Figures 2–4 show that for each of these precancerous lesions, proceed at a constant rate cancer sites the rate of increase of ‘effective tissue age’ throughout adult life (Doll, 1971; Peto, 1977). decreases around the age of 50 years. This change in Whatever the reason for the form of the equation, slope of the age-specific incidence curves at this age deviations from this form are easily recognized and strongly suggests an effect of menopause. By reducing provide fundamental insights into risk factors for the mitosis, menopause slows the rates of both spontaneous deviant cancer. Figure 2 (Cutler and Young, 1975) and environmentally induced mutations in the relevant shows the age-specific incidence rates for breast cancer stem cells. Therefore, early menopause will have a protective effect on each of these cancers. If ceasing *Correspondence: MC Pike; E-mail: [email protected] ovulation reduces risk, early age at menarche should Breast, endometrial and ovarian cancer prevention MC Pike et al 6380 500 100

100 10

1 Inc rate / 100,000 Inc rate / 100,000 10

0.1

0.01 1 30 40 50 60 70 30 40 50 60 70 Age (years) Age (years) Figure 3 Age-speciﬁc incidence rates for endometrial cancer in the Figure 1 Age-speciﬁc incidence rates for colorectal cancer in US Birmingham region of the UK, 1968–1972 (Waterhouse et al., 1976) white women, 1969–1971 (Cutler and Young, 1975)

100 500

100

10 Inc rate / 100,000 Inc rate / 100,000 10

1 1 30 40 50 60 70 30 40 50 60 70 Age (years) Age (years) Figure 2 Age-speciﬁc incidence rates for breast cancer in US white Figure 4 Age-speciﬁc incidence rates for ovarian cancer in the women, 1969–1971 (Cutler and Young, 1975) Birmingham region of the UK, 1968–1972 (Waterhouse et al., 1976)

increase it. The simplest mitotic rate model for these these defects point us to the direction further research cancers must thus have a sharp increase in the rate of should take, keeping in mind that the main purpose of ‘effective tissue ageing’ at menarche and a fairly steep epidemiological research on these tumors is to guide us fall at menopause. in finding ways to prevent them. In this paper, we will discuss cancers of the breast, As we have noted, Figures 2–4 suggest that breast, endometrium and ovary in terms of this model. The endometrial and ovarian cancer are driven by some model was fitted by setting f, the rate of ‘tissue ageing’ at aspect of ovarian function. Oophorectomy will signifi- each period in a woman’s reproductive life, at a different cantly reduce the incidence of all three cancers. This is, constant level. This rate is cumulated to give the however, not an acceptable intervention except for ‘effective tissue age’, d(t), at age t. Thus, for breast women at very high risk of the diseases. The problem cancer, the fitted rate (Figure 5) was f1 ¼ 1 from is to find out precisely what aspect of ovarian function is menarche to first birth, f2 ¼ 0.7 from first birth to responsible for increasing risk, possibly different for the menopause, and f3 ¼ 0.1 after menopause. We will draw three cancers, so that an acceptable and effective attention to defects in the fit of these models and how prevention approach can be developed.

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6381

We are concerned here in particular with ﬁnding f0 +b chemoprevention approaches to prevention. That sig- 1 niﬁcant chemoprevention is possible for endometrial Menarche Start of and ovarian cancer is known since oral contraceptives 0.8 Perimenopausal Period (OCs) provide it, but they do not provide protection against breast cancer. 0.6 FFTP f1 0.4 Cancer of the breast LMP 0.2 f2

The three major reproductive risk factors for breast Breast Tissue 'Exposure' Rate cancer are late menopause, early menarche and late first 0 full-term pregnancy (FFTP; Kelsey and Bernstein, 1996; 0 10203040506070 Colditz and Rosner, 2000). Women whose natural Age menopause occurs before the age of 45 years have only Figure 5 Model of rate of breast tissue aging model about one-half the breast cancer risk of those whose (LMP ¼ last menstrual period; Pike et al., 1983) menopause occurs after the age of 55 years. The increased breast cancer risk among women with earlier menarche is most noticeable at premenopausal ages. Censuses and Surveys, 1978), which show that breast With regard to late FFTP, women with a FFTP under cancer rates of married women are substantially greater the age of 20 years have about one-half the risk of than those of single women before the age of 25 years nulliparous women, but nulliparous women do not have and exceed the rates of married women only after the as high a risk as women whose FFTP is after about the age of 35 years. When these observations were first made age of 32 years. by Janerich (1979), there was a great deal of argument A definition of the rate of ‘tissue ageing’ with about their validity and interpretation. We now know increasing age that accounts in quantitative terms for that births, and particularly FFTP, are followed by a these three risk factors and the age–incidence curve transient increase in incidence. Pike et al. (1983) showed shown in Figure 2 was given in Pike et al. (1983). This how this can explain the ‘anomaly’ of women with a late definition, shown in Figure 5, assumes that breast ‘tissue age at FFTP having a higher breast cancer risk than ageing’ starts at menarche at a constant rate up to nulliparous women; the benefit of FFTP in reducing the FFTP, then continues at a reduced rate until the start of subsequent breast ‘tissue aging’ rate does not have long the perimenopausal period (around the age of 40 years), enough (before the postmenopausal drop in rate occurs) declining gradually thereafter to a low postmenopausal to compensate for, and then more than compensate for rate. To account for FFTP after around the age of 32 (if the FFTP occurred early enough) the increased rate years being associated with a breast cancer rate that is during the pregnancy. larger than that of nulliparous women, the model includes an abrupt but transient increase in ‘tissue Breast cell mitotic rate during the menstrual cycle ageing’ during the pregnancy. With this simple definition of the rate of ‘tissue ageing’, the effects of menopause, Understanding the relationship of breast cell mitotic menarche, and FFTP are well mimicked with the activity to the phase of the menstrual cycle was essential to following rates. Taking the rate of breast ‘tissue ageing’ relating this model of breast cancer risk to the biology of from menarche to FFTP as 100%, the rate increases to the breast. This information was forthcoming in the early about 400% during FFTP, then falls to 70% until 1980s in a series of studies, which showed that mitotic menopause, and to 10% thereafter. The fitted exponent activity of breast epithelium varies markedly during the of time (k) was 4.5, similar to the best-fitting value for normal menstrual cycle, with peak activity occurring in many epithelial cancers (Pike et al., 1983). The fit of the the luteal phase (Figure 6; Anderson et al., 1982; Pike model is shown by the smooth curve in Figure 2. This et al., 1993). This shows that regular (ovulatory) cycling is model has been extended by Rosner et al. (1994) to positively correlated with mitotic activity and explains include a transient increase during each later pregnancy, why delay in establishment of regular cycles after and other risk factors can be incorporated in the same menarche will decrease breast cancer risk (Henderson way (Colditz and Rosner, 2000). However, the basic et al., 1981; Henderson and Pike, 1981). The pattern is model shown in Figure 5 suffices to explain the under- quite different in the endometrium, in which mitotic lying concept and its implications for chemoprevention. activity is effectively confined to the follicular phase.

Age at FFTP Low breast cancer rates in Japan This model predicts that at young ages nulliparous As of 1970, age-adjusted breast cancer rates were some women as a group will have lower breast cancer rates ﬁve to six times higher in the US than in Japan, although than parous women as a group. This initially surprising the Japanese rate has increased substantially since then. conclusion is supported by breast cancer incidence and This difference was and remains of great interest to mortality data at young ages (Ofﬁce of Population epidemiologists. It cannot be due to differences in

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6382 genetic susceptibility, as the rate in Japanese-Americans Obesity in Hawaii is now similar to the rate in white women The overall effect of obesity is to increase the risk of (Pike et al., 2002). We used the above model to evaluate breast cancer, but this increased risk is restricted to older the role of ages at menarche, first birth and menopause in producing these differences in risk (Hoel et al., 1983). postmenopausal women (Kelsey and Bernstein, 1996). In premenopausal women, obesity is associated with a Older Japanese women had a much later menarche reduction in risk. These effects are predicted by their than American women and this would significantly reduce their risk; but the data on first birth, nulli- estrogen–progesterone milieu. Premenopausal obesity is associated with increased anovulation with decreased parity, and menopause showed that none of the serum hormone levels of both estrogen and, especially, decreased breast cancer rates in Japan could be progesterone. Postmenopausal obesity is associated with attributed to these factors. increased serum estrogen levels, so that the decreased Japanese breast cancer rates remained almost con- risk associated with premenopausal obesity is gradually stant after the age of 50 years (Fujimoto et al., 1979). In eliminated and an increased risk finally attained. There model terms, this implies that ‘tissue ageing’ ceases at is, however, accumulating evidence that breast tissue menopause in Japanese women. The latter effect could levels of estrogen may be uncorrelated with serum be explained by their much lower postmenopausal estrogen levels (Blankenstein et al., 1999). This suggests estrogen levels due to their weight being much lower than that of US women. Although the rates predicted that the above explanation may be fundamentally flawed; more work is needed to clarify this issue. for older Japanese women based on their later age at menarche and with a zero postmenopausal ‘tissue ageing’ rate (f2) are considerably lower than the Mammographic densities observed rates for US white women, they were still more than 2.5-fold greater than the observed Japanese The extent of mammographic densities (i.e., the white rates. The difference was already evident at premeno- areas in a standard mammogram of the breast pausal ages and this suggested that lower levels of representing connective and epithelial tissue) are corre- premenopausal estrogen and progesterone might be seen lated with breast cancer risk, and the variation in risk is in ‘traditional’ Asian women. This was confirmed in greater than that seen with almost all other breast cancer studies conducted during the mid- to late 1980s (Goldin risk factors (Table 1). For technical reasons, in et al., 1986; Bernstein et al., 1990; Key et al., 1990; particular to enable the use of mammograms taken on Shimizu et al., 1990). The reduced levels of serum different machines and by different operators, mammo- estrogens and probably progesterone seen in ’tradi- graphic densities are not usually measured as an tional’ Asian premenopausal women provides a plau- absolute amount but as a percent of the breast area on sible explanation of the remainder of the large difference the mammogram, the mammographic percent density in risk between traditional Japanese and US women. (MPD) (Figure 7). As can be seen from Table 1, MPD varies enormously OCs between different women. This variability has a strong genetic component (Boyd et al., 2002) and is also partly Many epidemiological studies have been conducted to under hormonal control (Spicer et al., 1994; Greendale evaluate the effect of OC use on breast cancer risk. The et al., 2003). Bilateral oophorectomy with add-back main result of these studies as regards our fundamental menopausal estrogen therapy (ET) causes a decrease of understanding of breast cancer etiology is that, whatever approximately 10% in MPD (Gram et al., 2001), and, effect there is, it is small. This suggests that the total extrapolating from the results of a study of the effect of breast cell proliferation of women on OCs should be menopausal hormone therapy on MPD (Greendale very close to that seen during normal menstrual cycles, et al., 2003), this appears to be almost independent of and that is precisely what is seen (Pike et al., 1993). the initial level of MPD. In a randomized study,

Breast Cell Proliferation 400 20 2.5

2 300 15

1.5 200 10 1

Estradiol (pg/ml) 100 5 0.5 Relative Labelling Index

0 0 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 131115212527 5 7 9 13 1719 23 Day of Cycle Day of Cycle Figure 6 Breast mitotic rate by day of cycle

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6383 Greendale et al. (2003) found a 1.2% increase in MPD The effects on breast cancer incidence of oophor- with ET and a 4.6% increase with the EPT of the ectomy (a reduction of B60%) or 5 years of EPT (an Women’s Health Initiative (WHI) trial (Chlebowski increase of about 40%) are thus roughly proportional to et al., 2003). MPD is thus a measure of ovarian hormone the changes observed in MPD with these treatments (a exposure, but only a small amount of variability in decrease of 10% following oophorectomy, and an MPD is accounted for by differences in hormone increase of 4.6% for EPT). exposure. However, change in MPD is closely associated with change in hormone exposure (Gram et al., 2001). Chemoprevention with ‘antihormonal’ drugs

Menopausal hormone therapy Chemoprevention of breast cancer in postmenopausal women by the selective estrogen receptor modulator Epidemiological studies have estimated that menopausal (SERM) tamoxifen has been clearly established (Lipp- estrogen therapy (ET) causes an approximately 10% man and Brown, 1999), and there is good evidence that increase in breast cancer risk after 5 years of use, that is, another SERM, raloxifene, may be equally or even more a 10% increase in risk at the 5-year point after 5 years of effective (Cauley et al., 2001). Furthermore, there is use or a cumulative increase in risk for 5 years of use of good evidence that the modern aromatase inhibitors are approximately half this amount (Collaborative Group more effective in the treatment of breast cancer than on Hormonal Factors in Breast Cancer, 1997), and the tamoxifen and that they will be as good or better than WHI trial of ET in hysterectomized women found no tamoxifen as chemoprevention agents. The main issues increase in risk (Anderson et al., 2004). Epidemiological that need to be resolved with the use of these agents studies of menopausal EPT found a considerably greater relates to important side effects associated with their risk in the range of 25–40% (Magnusson et al., 1999; use. We can confidently predict that steady progress will Colditz and Rosner, 2000; Ross et al., 2000; Schairer be made in improving this ‘antiestrogen’ approach to et al., 2000). This very significant risk associated with postmenopausal chemoprevention. EPT was confirmed by the WHI randomized trial of the The dose of estrogen and progestin in OCs, including effects of the most commonly prescribed EPT regimen, current low dose OCs, must be sufficient to block namely continuous combined EPT with conjugated ovulation and provide a satisfactory control of spotting. equine estrogens (CEE) at 0.625 mg/day and medrox- We have suggested (Pike et al., 1989; Spicer et al., 1991) yprogesterone acetate (MPA) at 2.5 mg/day. The WHI that chemoprevention of breast cancer in young result was a cumulative 26% increase in risk for premenopausal women could be achieved by a contra- approximately 5.2 years of use (Chlebowski et al., ception approach in which ovulation is inhibited by the 2003); this is equivalent to an approximately 50% use of a gonadotropin releasing hormone analog increase at 5 years of use. (GnRHA). Use of a GnRHA to block ovulation permits the add-back dose of estrogen and progestin to be selected purely on the basis of providing protection from Table 1 Breast cancer odds ratios associated with percent mammo- the hypoestrogenism induced by blocking ovulation and a graphic density the need to protect the endometrium; we suggested that % Densities Cases Controls Adjusted odds ratio a dose of estrogen equivalent to that used in ET with very intermittent progestin (10–12 days every 4 months) None 10 25 1.0 would provide such protection. This approach has been o10 29 61 1.2 10À 65 73 2.2 demonstrated to be acceptable to women at high risk of 25À 94 97 2.4 breast cancer (Spicer et al., 1993) with minimal bone loss 50À 90 67 3.4 and significant reductions in mammographic density 75+ 66 31 5.3 (Spicer et al., 1994; Ursin et al., 1998; Gram et al., 2001). aBoyd et al. (1995). Similar reductions in mammographic density were subsequently seen in a small number of women with BRCA1 mutations. As we have noted, total breast cell 30 mitotic activity is very similar in OC cycles and natural 25 ovulatory cycles, whereas the reduction in mammographic density induced by the GnRHA suggests that 20 breast cell proliferation has been significantly reduced. f Women Further clinical development is needed, but the potential 15 reduction in breast cancer risk from such an approach is 10 of the same order of magnitude as the reductions in both endometrial and ovarian cancer in OC users (see below).

Percentage o 5

0 Chemoprevention by mimicking early ﬁrst birth 0 102030405060708090100 Mammographic Percent Density The ‘mechanism’ by which an early ﬁrst birth protects Figure 7 Distribution of mammographic percent densities (Greendale against breast cancer is poorly understood. In the rat, et al., 2003) Russo et al.(1991)showedthatatermpregnancyinduced

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6384 permanent differentiation of terminal end buds and a was given in Pike (1987). This definition considers substantial reduction in breast cancer induced by 7,12- endometrial ‘tissue age’ as moving at rate of f0 ( ¼ 1) dimethylbenzanthracene (DMBA). Russo and Russo during normal menstrual cycles and at rate f1 (o1) (1994) also reported that these same pregnancy-induced during pregnancies and during the postmenopausal breast changes and reduced breast cancer risk could be period. With this definition, the effects of menopause, affected by treating the rats with the placental hormone menarche and parity are well mimicked with the chorionic gonadotropin, and proposed that treating following parameters: f1 ¼ 0.15 and k ¼ 6(Pike, 1987). young women with human chorionic gonadotropin The fit of the model is shown by the smooth curve in (hCG) would similarly significantly reduce their risk. Figure 3. Evidence that such differentiation of the breast occurs The model predicts that each birth will have the same in women with a term pregnancy is weak, but Bernstein quantitative effect on reducing risk; although this was et al. (1995) found in a breast cancer case–control study seen in a number of studies, in general first birth has a of young women in Los Angeles that hCG treatment greater effect than subsequent births. There is also was associated with an approximately 40% reduced risk evidence that a late age at last birth may be associated of breast cancer in nulliparous women. This intriguing with a further reduction in risk independent of parity. observation was associated with a wide confidence OC use for 1 year is considerably less protective than a interval but it needs to be further pursued. The women single birth even a birth after the first. had had hCG injections as part of a now not-practiced weight reduction program (Simeons, 1954) and was not Endometrial cell mitotic rate during the menstrual cycle sufficiently common for it to be effectively pursued by standard epidemiological case–control methods. Look- The relationship of endometrial cell mitotic activity to ing for associated reductions in mammographic densi- the phase of the menstrual cycle is critical in under- ties with use of hCG in nulliparous women as are seen in standing the etiology of endometrial cancer. Endome- parous compared to nulliparous women may be trial cell proliferation is exquisitely sensitive to the rewarding (El-Bastawissi et al., 2001; Jakes et al., effects of estrogens ‘unopposed’ by progestin (see Key 2002; Gapstur et al., 2003). and Pike (1988) for comprehensive list of references). Although there are some gaps in the data, endometrial cell proliferation appears to be close to maximal for some 18 days of the menstrual cycle covering the Cancer of the endometrium follicular and early luteal phase of the cycle, with close to no endometrial cell proliferation for the last 10 days US age-specific incidence data for both endometrial and of the cycle. ovarian cancer are difficult to interpret because the very high rate of hysterectomies and oophorectomies in the Parity US means that the denominator for calculating rates cannot be accurately estimated. Data from the UK As we noted above, there is a greater reduction in risk around 1970 suffer much less from this problem and we with first birth than with subsequent births (Pettersson use such data to illustrate the situation for cancer at et al., 1986; Kvale et al., 1988; Brinton et al., 1992; these two sites. Data from around 1970 also avoid the Albrektsen et al., 1995; Lambe et al., 1999): an distortion of the endometrial cancer rates from the approximately 30% reduction in risk with first birth widespread use of OCs, which decreases risk, and compared to an approximately 15% reduction with each menopausal ET, which increases risk, and the distortion subsequent birth. of the ovarian cancer rates from widespread use of OCs, In their cohort study, Kvale et al. (1988) found that a which decreases risk. Figure 3 shows the age-specific late age at last birth was associated with a significant incidence curve for endometrial cancer in women in the reduction in risk independent of parity. This effect has Birmingham region of the UK, 1968–1972 (Waterhouse now been found in a number of studies (Albrektsen et al., 1976). The suggestion of a strong protective effect et al., 1995; Lambe et al., 1999), although the effect is of menopause is again evident. not always seen (Pettersson et al., 1986; McPherson The major reproductive risk factors increasing en- et al., 1996). In our study (Pearce et al., 2004), we found dometrial cancer risk are late menopause and early that compared to a nulliparous woman, risk of menarche, while increasing parity decreases risk (Key endometrial cancer was reduced by 51% in a woman and Pike, 1988). OC use decreases risk significantly, with a last birth after the age of 35 years and that while risk is greatly increased in obese women. All these additional births (before the last) reduced risk further by results can be explained by the ‘unopposed estrogen’ an average of an additional 15% per birth. A last birth hypothesis for endometrial cancer that risk is increased before the age of 25 years was associated with only a by exposure of the endometrium to estrogen unopposed 12% reduced risk of endometrial cancer. by progesterone or a synthetic progestin. Our results on the effects of age at last birth are A definition of the rate of increase in endometrial in good agreement with the results seen in the three ‘tissue age’ with increasing age that describes in other positive studies although the differential effect quantitative terms the effect of menopause, menarche was not quite so large in the study of Lambe et al. and parity on the age–incidence curve shown in Figure 3 (1999). In the study of McPherson et al. (1996), ‘older

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6385 age at last pregnancy appeared protective until together (continuous combined EPT). Continuous controlled for gravidity’, but the authors did not give combined EPT is not associated with regular bleeding, the adjusted (controlled) results. The effect of age at but there is significant spotting. last birth should probably be considered as requiring With the dose of ET most commonly used in the US, further supportive data before it can be assumed as that is, conjugated equine estrogens (CEE) at 0.625 mg/ firmly established. day, we found that endometrial cancer risk is increased approximately 120% at 5 years of use (Pike et al., 1997). OC use For women who received sequential EPT with the progestin medroxyprogesterone acetate (MPA) at 5 or OC use is associated with a significant reduction in 10 mg/day given for around 7 days per month, endometrial cancer risk. At premenopausal ages, endometrial cancer risk was still markedly elevated with the risk is reduced by approximately 10% per year of an estimated increase of 87% at 5 years of use. However, use (Henderson et al., 1983; Cancer and Steroid quite remarkably, when progestin was given for 10 or Hormone Study, 1987a), but this declines with increas- more days the increased risk was estimated to be only ing age. OCs are a mixture of estrogen (almost solely 10% at 5 years of use and this result was not statistically ethenyl-estradiol) and a progestin. Their composition significant from no increase in risk. Continuous is such that they are progestin dominant for the combined EPT was associated with essentially no endometrium, so that endometrial proliferation is increased risk. confined effectively to the days on which the OC is not CEE at 0.625 mg/day, unopposed by a progestin, taken. On the assumption that endometrial proliferation results in endometrial cell proliferation approximating will be proportional to the number of days of that found during the follicular phase of the menstrual unopposed estrogen the reduction in risk with OC use cycle (King and Whitehead, 1984), and this can be is calculated to be about 10%; this is calculated by completely blocked by MPA at 5 or more mg/day (Lane comparing the 18 days of proliferation during a normal et al., 1986). This abolition of cell proliferation and the menstrual cycle to the 7 days on an OC and then observation that ‘Seveny[is the]y number of days that calculating the ratio of the total cell proliferation to the the level of progesterone is above 5 ng/mL in the normal age of 50 years in an ‘average’ woman who does and menstrual cycle’ (Flowers et al., 1983), persuaded many does not use an OC for a year, finally raising the prescribers that 7 days of progestin was sufficient to resulting ratio to the power 6. We note that this form of abolish any risk. However, progestin use for 7 days did calculation suggests that a birth should be associated not completely remove the risk of hyperplasia (Paterson with a 16% reduction in risk; close to that found to be et al., 1980), and 10 or more days of progestin therapy associated with births. became standard practice (King and Whitehead, 1984). It should be pointed out that, as can be seen in Figure 3, Obesity endometrial cancer incidence is increasing rapidly in the premenopausal period, even where obesity is uncom- Obesity is associated with a greatly increased risk of mon, so that the notion that mimicking the progestin endometrial cancer. This is evident both at premeno- phase of the menstrual cycle would provide adequate pausal (Henderson et al., 1983; Cancer and Steroid protection was always suspect. Hormone Study, 1987a) and postmenopausal ages (Pike Key and Pike (1988) had previously argued that, if et al., 1997). At premenopausal ages, it results from the endometrial cell proliferation in the basalis (stem-cell) associated anovular cycles in which the endometrium is layer was the key to increased risk from ET, there would stimulated throughout the cycle, and this is then made still be an increased risk from sequential EPT even with worse during the postmenopausal period when the 10 days of progestin use, since there would still be obesity is associated with both higher levels of estrogen unopposed estrogen for around 15 days per treatment from extraglandular conversion of androgens to estro- cycle (sequential EPT was usually given for 25 days per gens and lower levels of sex-hormone-binding globulin 28-day cycle). This prediction was clearly contradicted so that the estrogen produced is more bioactive (Siiteri by the epidemiological results showing that endometrial and MacDonald, 1973; Siiteri et al., 1981). cancer risk was almost eliminated with 10 days of progestin use. A simple cell proliferation model for Menopausal hormone therapy endometrial cancer is not tenable. In their studies of endometrial tissue after 7 days of Menopausal ET substantially increases a woman’s risk progestin therapy, Flowers et al. (1983) found that such of developing endometrial cancer. To reduce this short progestin use did ‘not cause all the endometrium increased cancer risk, progestins were added to ET, to desquamate to the basalis layer [only] 40 to 50% of menopausal EPT, for between 5 and 15 days (usually 7 the functional layer. was lost’. If these functionalis cells or 10 days) per month in a sequential manner are susceptible to cancer, and if a much greater (sequential EPT). Sequential EPT causes regular bleed- proportion of such cells are lost with longer progestin ing in many women and is associated with other therapy, this could provide an explanation for the sharp negative side effects, so that subsequently, continuous distinction between 7 and 10 days use. It would also be combined therapy regimens were developed in which consistent with the observation of pathologists that estrogen and a lower dose of progestin are always taken early stage tumors often appear to arise in the

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6386 functionalis. This could also provide an explanation for A definition of the rate of increase in ovarian ‘tissue the great protective effect of a late age at last birth. age’ with increasing age that describes in quantitative Such endometrial sloughing may also provide an terms the suggested effect of menopause and the well- explanation for the observations made some 10 years established protective effect of parity on risk, as well as ago that CEE at 0.625 mg/day did not produce the age–incidence curve shown in Figure 4, was given in hyperplasia if 10 mg of MPA was given for 14 days Pike (1987). This definition considers ovarian tissue (i.e., every 3 months (Ettinger et al., 1994; Williams et al., the tissue that gives rise to ovarian cancer) exposure as 1994). Hyperplasia did occur in a further trial in which moving at rate of f0 ( ¼ 1) during normal menstrual estradiol was given at 2 mg/day for 68 days, followed by cycles and at rate f1 (o1) during pregnancies and during norethisterone at 1 mg/day for 10 days and finally the postmenopausal period. With this definition, the estradiol at 1 mg/day for 6days (Cerin et al., 1996). This effect of parity and the age–incidence curve are well may simply be a dose of estrogen effect and would described with f1 ¼ 0.09 and k ¼ 4. The fit of the model is provide a further reason for the steady large increase in shown by the smooth curve in Figure 4. premenopausal endometrial cancer with age. The model predicts that each birth will have the same quantitative effect on reducing risk. This is not seen in Chemoprevention epidemiological studies; first birth provides considerably more protection than subsequent births; that is, nullipar- OC use remains an effective approach to chemopreven- ous women are at especially high risk. This is partly due tion of endometrial cancer: in our data and in recently to the inclusion in the nulliparous group of women who published studies, the protective effect was seen to be were infertile and there is now evidence that they are at very long term. especially high risk. However, this only explains part of If the finding of a major protective effect of a late the considerable excess risk of nulliparous women. It birth can be confirmed, then we may be able to use this now also appears that age at last term pregnancy has a for a short-duration chemoprevention strategy with a major effect on risk and that understanding the etiology long-term preventive effect. To do this, we will need to of ovarian cancer needs to take this into account. The understand the mechanism of the protective effect. model also predicts that approximately 1 year of OC use Further detailed study of the endometrium at or soon would be as protective against ovarian cancer as a full- after delivery, and comparisons with the endometrium term pregnancy. Epidemiological data are in line with in women on OCs and during regular cycling would this if one considers the protection from full-term hopefully lead to a deeper understanding of the origin pregnancies after the first. and possible prevention of endometrial cancer. ‘Ovarian cell’ mitotic rate during the menstrual cycle Cancer of the ovary There have been no studies of the mitotic rate of OSE (either on the surface or in inclusion cysts) during the Figure 4 shows the age-specific incidence curve for menstrual cycle, other than the obvious repair of the invasive epithelial ovarian cancer (ovarian cancer) in surface OSE that occurs after each ovulation. It is also women in the Birmingham region of the UK, 1968–1972 not clear whether knowledge of the mitotic activity of (Waterhouse et al., 1976). The suggestion of a strong the OSE will be informative since it is quite possible, as protective effect of menopause is again evident. A Dubeau (1999) has emphasized, that the OSE is not the number of epidemiological studies have found this cell of origin of ovarian cancers. Dubeau argues that suggested protective effect of early menopause, but this ovarian carcinomas as well as adenomas are similar effect has not been seen as consistently as it has been histologically to the tissue of the fallopian tube, found for breast and endometrial cancer. endometrium and cervix, that is, they appear to have Epidemiological studies have consistently found that been derived from Mullerian tissue, and that the the risk of ovarian cancer is significantly decreased with evidence for metaplasia of the OSE to produce such increasing numbers of births, and Fathalla (1972) tumors is weak. Recent results from the study of ovaries proposed that the mechanism of this effect was the from prophylactic oophorectomies in women with interruption of the tearing of the ovarian surface BRCA1 mutations strongly suggest, however, that the epithelium (OSE) with each ovulation (the ‘incessant ovarian cancers associated with such mutations do arise ovulation hypothesis’). OSE (on the surface or trapped in inclusion cysts (J Boyd, unpublished results presented in inclusion cysts) is regarded by most pathologists, at the American Association for Cancer Research 2004 although not by all (Dubeau, 1999), as the tissue of Annual Meeting). origin of ovarian cancer, and it was the increased cell Although there may be dispute regarding the cell of division of the OSE involved in the repair of the surface origin of ovarian cancer, there is general agreement that after each ovulation that was considered to be the major ovarian cystadenomas and ovarian cancers have the factor increasing ovarian cancer risk. The epidemiolo- same cell of origin. There is some relevant information gical findings of a significant protective effect of OC use concerning mitogenic effects available from in vitro (Newhouse et al., 1977; Casagrande et al., 1979; Cancer studies of cell lines established from the two tumor and Steroid Hormone Study, 1987b) provided further types. Such studies have shown stimulator effects of support for this hypothesis. estrogen and inhibitory effects of progesterone at high

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6387 steroid levels similar to intraovarian levels in ovulatory 20%. Additional births (before the last) reduced risk cycles (Zhou et al., 2002). These and other observations further by some 10% per birth. have given rise to a ‘high-estrogen/low-progestin hy- Our results were in close agreement with the results pothesis’ for the etiology of ovarian cancer. seen in the study of Whiteman et al. (2003). Owing to the strong positive relationship, at least in US studies, Ages at menopause and menarche between late age at first birth and late age at last birth, the result reported by Whittemore et al. (1992) is also Although an increased risk of ovarian cancer in likely to be supportive of a late age at last birth effect. association with late age at natural menopause has been Titus-Ernstoff et al. (2001) also observed a trend of reported in a few studies including our own (Pike et al., increased protective effect with late age at last birth. 2004), no association has been found in many studies This effect was, however, not seen in two other (see Schildkraut et al. (2001) and references therein). population-based case–control studies (Risch et al., Age at menarche has generally not been identified as a 1994; Riman et al., 2002a), and the effect must currently risk factor for ovarian cancer. be considered as not firmly established. The problem Why an effect of age at menopause is not consistently may be the difficulty of distinguishing the effect of ages found is difficult to reconcile with the age–incidence of first birth and last birth and parity when these factors curve for the disease and with both of the proposed are highly correlated. Joint analyses of all these studies etiological hypotheses. It may be that the early stages of should be able to settle this important issue. ovarian cancer affect ovarian function and cause an early menopause. If this is so, then a more clear effect of OC use age at menopause should be seen if one restricted attention to ovarian cancer cases diagnosed after the OC use is consistently found to provide very significant age of 65 years, say. Such analyses have not been protection against ovarian cancer (see Riman et al. carried out. (2002a), Royar et al. (2002) and Pike et al. (2004) and The lack of an effect of early menarche can be references therein). The reduction in risk for 12–15 explained on the high-estrogen/low-progesterone hy- months of OC use is the same as that associated with pothesis since early menarche is associated with a more term pregnancies before the last (assuming the validity rapid onset of regular (ovulating) cycles than a late of the model that proposes a significant effect of age at menarche, so that the amount of exposure to estrogen last birth) or births after the first (assuming no effect of unopposed by progesterone may not be much affected age at last birth). Ovarian function is more completely by age at menarche. suppressed during mid- to late pregnancy than is ever achieved with OC use, and the early part of pregnancy is Parity associated with high progesterone production in the ovary. With OC use, during the 7 days per 28-day cycle There is a greater reduction in risk with first birth than that the OC is not taken, some follicle development with subsequent births (Whittemore et al., 1992; Risch occurs with an associated production of estrogen. et al., 1994; Ness et al., 2000; Titus-Ernstoff et al., 2001; However, these considerations do not help clearly Riman et al., 2002a ; Tung et al., 2003): an approxi- distinguish between the incessant ovulation hypothesis mately 40% reduction in risk with first birth compared and the high-estrogen/low-progestin hypothesis. to approximately 10% with each subsequent birth. This A recent reanalysis of the CASH Study (Schildkraut is partly due to the inclusion in the nulliparous group of et al., 2002) suggested that OC formulations with a women who were infertile, and there is evidence that ‘high-dose/ potency’ progestin may be associated with a they are at especially high risk (Whittemore et al., 1992; greater reduction in risk than those with a ‘low-dose/ Ness et al., 2002). However, based on these results, one potency’ progestin. We observed this effect in our recent can show that excluding such women only reduces the study (Pike et al., 2004), but our results were not first birth reduction to 30%, still much greater than is statistically significant, and Ness et al. (2000) did not seen with subsequent births. observe such an effect. Further data are needed to Further analysis of the parity effect combining data adequately evaluate this issue. from four US population-based case–control studies Rodriguez et al. (1998, 2002b) have proposed, based showed a reduced risk with late age at first birth on long-term studies in macaques of extended exposure (Whittemore et al., 1992). This issue has now been to OCs or the individual components of OCs, that it is a investigated in a number of studies. In our study (Pike direct action of the progestin component on the OSE et al., 2004), we found a major effect of age at last birth; that provides the protection from OCs against ovarian the later the age at last birth the lower the risk of cancer, not the blocking of ovulation and/or follicle ovarian cancer. Compared to a nulliparous woman, the development. In these studies, the progestin component overall risk of ovarian cancer in a woman with a last of OCs, given alone without the estrogen component, birth after the age of 35 years was reduced by 58%, with showed increased apoptosis of the OSE very similar to a reduction of 51% solely due to the last birth. A last that seen with OCs. This most interesting study is, birth before the age of 25 years was associated with only however, difficult to interpret, since the progestin a 16% reduced risk of ovarian cancer; even after component alone would also block ovulation and adjusting for numbers of births this only increased to follicle development, so that one cannot distinguish a

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6388 direct progestin effect from an effect of suppression of 1.42 to 2.66 to 3.43. There was only a very weak follicle development and ovulation. This could be relationship between BMI and risk of Regional/Distant studied using a GnRH analog to suppress ovulation. It disease; the odds ratios over the same increasing BMI will be important in future studies to also study categories were 1.0, 0.89, 1.07 and 1.16. In this same inclusion cysts, as we discussed above. study, we noted from review of pathology reports that many of the Localized disease cases (20% of the total Obesity cases) were diagnosed as a consequence of some other complaint, in many cases vaginal bleeding associated Obesity is associated with anovulation in the premeno- with an endometrial cancer primary or an abdominal pausal period and relatively high estrogen levels in the mass associated with an abdominal nonovarian primary. postmenopausal period. On the incessant ovulation (These cases were omitted in calculating the relationship hypothesis, premenopausal obesity would, therefore, of BMI to risk.) When one considers the reasons why a be expected to be associated with decreased risk, and Localized disease case would be diagnosed, it is clear this effect should carry over at least for a time into the that increasing BMI could simply be associated with the postmenopausal period (Pike, 1987). The incessant reason for a medical work-up rather than with ovarian ovulation hypothesis makes no prediction about the cancer itself. Further investigation of the etiological effect of postmenopausal obesity. significance of BMI is needed. On the high-estrogen/low-progestin hypothesis, one would predict that premenopausal obesity would Menopausal hormone therapy (HT) increase risk, since, during the anovular cycles, follicular levels of estrogen would occur throughout the cycle The current epidemiological evidence obtained from without their effect being interfered with by progester- cohort and population-based case–control studies with one. In the postmenopausal period, obesity would again data on formulation and duration of use of menopausal be predicted to increase risk because of the higher HT suggests that unopposed menopausal ET increases estrogen levels. It is, however, impossible at the current ovarian cancer risk (Weiss et al., 1982; Cramer et al., time to predict what the magnitude of this postmeno- 1983; Lee et al., 1986; Risch, 1996; Purdie et al., 1999; pausal effect would be, since we have no knowledge of Rodriguez et al., 2001; Lacey et al., 2002; Riman et al., the dose–effect relationship between estrogen levels in 2002b; Sit et al., 2002). Our meta-analysis of ET use in the blood and ‘ovarian-cell’ proliferation; as we noted these studies, together with the results of our own study above, the in vitro work on ovarian cell proliferation was that found a nonsignificant increase in risk, found a all done with very high levels of estrogen similar to that statistically highly significant increased risk of approxi- achieved premenopausally within the ovary. mately 24% at 5 years of ET use. The estimated overall What is the epidemiological evidence on the effect of risk associated with EPT use was smaller, 13% at 5 years obesity? A systematic review (Purdie et al., 2001) found of EPT use, and was not statistically significant. a consistent positive association between body size and If BMI is related to ovarian cancer risk only ovarian cancer risk in 11 population-based case–control incidentally through increasing BMI being related to studies and five cohort studies. Since that review, results increased contact with the medical system, then this from two other population-based case–control studies would cast some doubt on the increased incidence seen (Kuper et al., 2002; Lubin et al., 2003) and five cohort with ET use since BMI is associated with increased studies (Fairfield et al., 2002; Lukanova et al., 2002; postmenopausal estrogen levels and decreased sex- Rodriguez et al., 2002a; Engeland et al., 2003; Schouten hormone-binding globulin. One would therefore expect et al., 2003) have been published. Although one of these to see a clear association with BMI if the ET result studies (Lukanova et al., 2002) with a small number of was true. ovarian cancer cases did not find such an effect, all the Some of the observed risk seen with ET use may be a other studies did. The increase in risk between the upper diagnostic bias since we found that the ET effect was and lower quartiles of body mass index (BMI) is around much smaller in our study when we restricted attention 30%. This analysis was essentially restricted to post- to Regional/Distant disease. With Localized disease, we menopausal BMI, but a similar effect has been found estimated the risk at 5 years of ET use as 1.51 and 1.22 with premenopausal obesity (Casagrande et al., 1979; for natural menopause and hysterectomized women Farrow et al., 1989; Tornberg and Carstensen, 1994; respectively, while for Regional/Distant disease these Purdie et al., 2001). Thus, the evidence appears quite risks were 1.11 and 1.10 respectively. A joint careful definite: increasing weight is associated with increasing analysis of the published studies is again needed to help risk. This observed association may, however, be clarify this issue. misleading as regards etiology. In a population-based ovarian cancer case–control Other risk factors study we recently completed (Pike et al., 2004), risk increased over four increasing BMI categories (o25, Hysterectomy has been consistently found to be a 25À,30À and 35 þ kg/m2) from 1.0 to 0.97 to 1.29 to protective factor. Although interpretation of this result 1.46, as was seen in the Purdie et al. (2001) review. in some studies is not always clear as women whose However, the increasing trend was restricted to Loca- hysterectomy included oophorectomy may have been lized disease, for which the risk increased from 1.0 to mistakenly included in the control group, the overall

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6389 results strongly suggest that this effect is real (Pike et al., or ‘terminally differentiate’ a significant proportion of 2004). Tubal ligation has also been consistently found to ‘premalignant’ ovarian cancer precursor cells. Detailed be protective. Use of talc in the genital area has also study of the ovaries, including OSE and inclusion cysts, consistently been found to increase risk. The underlying of women immediately after delivery should be most reasons for these effects are not understood. informative. It would also be important to closely observe the fallopian tubes as microscopic ovarian Histology cancers have been found outside the ovary and closely associated with the fallopian tubes in some studies The above results are generally seen with the different looking for early lesions in women with BRCA1 histological varieties of ovarian cancer, although there mutations undergoing prophylactic oophorectomies have been suggestions of some factors being more (Piek et al., 2001). These observations, of course, are important for endometrioid tumors. This can, however, supportive of the ideas of Dubeau (1999) regarding the only be properly studied if one takes into account the cell of origin of at least some ‘ovarian cancers’. Such strong relationship between histology and stage of studies should also be undertaken in macaques or other disease. Data from the Los Angeles Cancer Surveillance lower primates as was done by Rodriguez et al. (1998, Program, the National Cancer Institute’s SEER registry 2002b) to study the effect of OCs. In vitro experiments for Los Angeles County, showed that the majority on the effects of prolonged pregnancy levels of estrogen (approximately two-thirds) of ovarian cancer cases were and progesterone on ovarian cystadenoma cells may histologically classified as serous, around 20% as also be informative. endometrioid (including clear cell) and 10% as mucinous: for serous tumors, 12% were Localized; for endometrioid, 42%; and for mucinous, 55%. Further detailed analysis of the reasons why the Localized Conclusion tumors came to be diagnosed, may be most informative. Analysis of the age-specific incidence curves for three of Chemoprevention the most important cancers of women, namely, breast, endometrium and ovary, shows that menopause has a What is the current situation regarding chemopreventive large protective effect on each of them. If ovarian strategies against ovarian cancer? OC use remains an function continued till the age of 70 years, extrapo- effective approach: in our data, and in recently lation of the premenopausal incidence curves suggests published studies, the protective effect was seen to be that the incidence rates of these tumors in old age would very long term extending to women over the age of 60 be increased fourfold for breast, fivefold for ovary and years who had used OCs many years in the past. eightfold for endometrium. This protective effect of If the finding of a major protective effect of a late menopause remains the central clue to the etiology of birth can be confirmed, then we may be able to use this these cancers, and hence to their chemoprevention. for a short-term chemoprevention strategy with a long- term preventive effect. To do this, we will need to understand the mechanism of the protective effect. We did not find late incomplete pregnancies or late OC use Declaration of conflict of interest to provide such an effect, so we must look for the distinctive characteristics of a full-term pregnancy. Dr. Pike is associated with Balance Pharmaceuticals, It is possible that the prolonged high levels of Inc.; a company set up to develop the GNRHA progesterone during a term pregnancy will be lethal to contraceptive discussed above.

References

Albrektsen G, Heuch I, Tretli S and Kvale G. (1995). Int. J. Anderson JT, Ferguson DJP and Raab GM. (1982). Br. J. Cancer, 61, 485–490. Cancer, 46, 376–382. Anderson GL, Limacher M, Assaf AR, Bassford T, Beresford Bernstein L, Hanisch R, Sullivan-Halley J and Ross RK. SA, Black H, Bonds D, Brunner R, Brzyski R, Caan B, (1995). Cancer Epidemiol. Biomarkers Prev., 4, 437–440. Chlebowski R, Curb D, Gass M, Hays J, Heiss G, Bernstein L, Yuan JM, Ross RK, Pike MC, Hanisch R, Lobo Hendrix S, Howard BV, Hsia J, Hubbell A, Jackson R, R, Stanczyk F, Gao YT and Henderson BE. (1990). Cancer Johnson KC, Judd H, Kotchen JM, Kuller L, LaCroix AZ, Causes Control, 1, 51–58. Lane D, Langer RD, Lasser N, Lewis CE, Manson J, Blankenstein MA, van de Ven J, Maitimu-Smeele I, Donker Margolis K, Ockene J, O’Sullivan MJ, Phillips L, GH, de Jong PC, Daroszewski J, Szymczak J, Milewicz A Prentice RL, Ritenbaugh C, Robbins J, Rossouw JE, and Thijssen JH. (1999). J. Steroid Biochem. Mol. Biol., 69, Sarto G, Stefanick ML, Van Horn L, Wactawski-Wende J, 293–297. Wallace R, Wassertheil-Smoller S and Women’s Health Boyd NF, Byng JW, Jong RA, Fishell EK, Little LE, Miller Initiative Steering Committee. (2004). JAMA, 291, AB, Lockwood GA, Tritchler DL and Yaffe MJ. (1995). J. 1701–1712. Natl. Cancer Inst., 87, 670–675.

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6390 Boyd NF, Dite GS, Stone J, Gunasekara A, English DR, Henderson BE and Pike MC. (1981). Hormones and Breast McCredie MR, Giles GG, Tritchler D, Chiarelli A, Cancer . Pike MC, Sitteri PK and Welsch CW (eds). Cold Yaffe MJ and Hopper JL. (2002). N. Engl. J. Med., 347, Spring Harbor Laboratory: Cold Spring Harbor, pp. 115–130. 886–894. Henderson BE, Casagrande JT, Pike MC, Mack T, Rosario I Brinton LA, Berman ML, Mortel R, Twiggs LB, Barrett RJ, and Duke A. (1983). Br. J. Cancer, 47, 749–756. Wilbanks GD, Lannom L and Hoover RN. (1992). Am. J. Henderson BE, Pike MC and Casagrande JT. (1981). Lancet, Obstet. Gynecol., 167, 1317–1325. 2, 363. Cairns J. (2002). Proc. Natl. Acad. Sci. USA, 99, 10567–10570. Henderson BE, Ross RK, Pike MC and Casagrande JT. Cancer and Steroid Hormone Study (1987a). JAMA, 257, (1982). Cancer Res., 42, 3232–3239. 796–800. Hoel DG, Wakabayashi T and Pike MC. (1983). Am. J. Cancer and Steroid Hormone Study (1987b). N. Engl. J. Med., Epidemiol., 118, 78–89. 316, 650–655. Jakes RW, Duffy SW, Ng FC, Gao F, Ng EH, Seow A, Lee Casagrande JT, Pike MC, Ross RK, Louie EW, Roy S and HP and Yu MC. (2002). Cancer Epidemiol. Biomarkers Henderson BE. (1979). Lancet, 2, 170–173. Prev., 11, 608–613. Cauley JA, Norton L, Lippman ME, Eckert S, Krueger KA, Janerich DT. (1979). Lancet, 1, 1240–1241. Purdie DW, Farrerons J, Karasik A, Mellstrom D, Ng KW, Kelsey JL and Bernstein L. (1996). Annu. Rev. Public Health, Stepan JJ, Powles TJ, Morrow M, Costa A, Silfen SL, Walls 17, 47–67. EL, Schmitt H, Muchmore DB, Jordan VC and Ste-Marie Key TJ and Pike MC. (1988). Br. J. Cancer, 57, 205–212. LG. (2001). Breast Cancer Res. Treat., 65, 125–134. Key TJ, Chen J, Wang DY, Pike MC and Boreham J. (1990). Cerin A, Heldaas K and Moeller B. (1996). N. Engl. J. Med., Br. J. Cancer, 62, 631–636. 334, 668–669. King RJ and Whitehead MI. (1984). Medical Management of Chlebowski RT, Hendrix SL, Langer RD, Stefanick ML, Gass Endometriosis . Raynaud JP, Ojasoo T and Martini L (eds). M, Lane D, Rodabough RJ, Gilligan MA, Cyr MG, Raven Press: New York, pp. 67–77. Thomson CA, Khandekar J, Petrovitch H, McTiernan A Kuper H, Cramer DW and Titus-Ernstoff L. (2002). Cancer and WHI Investigators. (2003). JAMA, 289, 3243–3253. Causes Control, 13, 455–463. Colditz GA and Rosner B. (2000). Am. J. Epidemiol., 152, Kvale G, Heuch I and Ursin G. (1988). Cancer Res., 48, 950–964. 6217–6221. Collaborative Group on Hormonal Factors in Breast Cancer Lacey Jr JV, Mink PJ, Lubin JH, Sherman ME, Troisi R, (1997). Lancet, 350, 1047–1059. Hartge P, Schatzkin A and Schairer C. (2002). JAMA, 288, Cramer DW, Hutchison GB, Welch WR, Scully RE and Ryan 334–341. KJ. (1983). J. Natl. Cancer Inst., 71, 711–716. Lambe M, Wuu J, Weiderpass E and Hsieh CC. (1999). Cancer Cutler SJ and Young JL. (1975). Natl. Cancer Inst. Causes Control, 10, 43–49. Monogr., 41. Lane G, Siddle NC, Ryder TA, Pryse-Davies J, King RJB and Dao TL. (1981). Hormones and Breast Cancer. Pike MC, Siiteri Whitehead MI. (1986). Fertil. Steril., 45, 345–352. PK and Welsch CW (eds). Cold Spring Harbor Laboratory: Lee NC, Wingo PA and Peterson HB. (1986). Aging, Cold Spring Harbor, pp. 281–298. reproduction, and the climacteric Mastroiani Jr L and Doll R. (1971). J. R. Stat. Soc. Ser. A., 134, 133–165. Paulsen CA (eds). Plenum Press: New York. Dubeau L. (1999). Gynecol. Oncol., 72, 437–442. Lippman SM and Brown PH. (1999). J. Natl. Cancer Inst., 91, El-Bastawissi AY, White E, Mandelson MT and Taplin S. 1809–1819. (2001). Ann. Epidemiol., 11, 257–263. Lubin F, Chetrit A, Freedman LS, Alfandry E, Fishler Y, Engeland A, Tretli S and Bjorge T. (2003). J. Natl. Cancer Nitzan H, Zultan A and Modan B. (2003). Am. J. Inst., 95, 1244–1248. Epidemiol., 157, 113–120. Ettinger B, Selby J, Citron JT, Vangessel A, Ettinger VM and Lukanova A, Toniolo P, Lundin E, Micheli A, Akhmedkhanov Hendrickson MR. (1994). Obstet. Gynecol., 83, 693–700. A, Biessy C, Lenner P, Krogh V, Berrino F, Hallmans G, Fairﬁeld KM, Willett WC, Rosner BA, Manson JE, Riboli E and Kaaks R. (2002). Int. J. Cancer, 99, 603–608. Speizer FE and Hankinson SE. (2002). Obstet. Gynecol., Magnusson C, Baron JA, Correia N, Bergstrom R, Adami H- 100, 288–296. O and Persson I. (1999). Int. J. Cancer, 81, 339–344. Farrow DC, Weiss NS, Lyon JL and Daling JR. (1989). Am. J. McPherson CP, Sellers TA, Potter JD, Bostick RM and Epidemiol., 129, 1300–1304. Folsom AR. (1996). Am J Epidemiol., 143, 1195–1202. Fathalla MF. (1972). Obstet. Gynecol. Surv., 27, 751–768. Ness RB, Cramer DW, Goodman MT, Kjaer SK, Mallin K, Flowers CE, Wilborn WH and Hyde BM. (1983). Obstet. Mosgaard BJ, Purdie DM, Risch HA, Vergona R and Wu Gynecol., 61, 135–143. AH. (2002). Am. J. Epidemiol., 155, 217–224. Fujimoto I, Hanai A and Oshima A. (1979). Natl. Cancer Inst. Ness RB, Grisso JA, Klapper J, Schlesselman JJ, Silberzweig Monogr., 53, 5–15. S, Vergona R, Morgan M, Wheeler JE and SHARE Study Gapstur SM, Lopez P, Colangelo LA, Wolfman J, Van Horn Group. (2000). Am. J. Epidemiol., 152, 233–241. L and Hendrick RE. (2003). Cancer Epidemiol. Biomarkers Newhouse ML, Pearson RM, Fullerton JM, Boesen EAM Prev., 12, 1074–1080. and Shannon HS. (1977). Br. J. Prev. Soc. Med., 31, Goldin BR, Adlercreutz H, Gorbach SL, Woods MN, Dwyer 148–153. JT, Conlon T, Bohn E and Gershoff SN. (1986). Am. J. Clin. Ofﬁce of Population Censuses and Surveys (1978). Decennial Nutr., 44, 945–953. Supplement, Occupation and Mortality, 1970–72 (DS No. Gram IT, Ursin G, Spicer DV and Pike MC. (2001). Cancer 1.). HMSO: London. Epidemiol. Biomarkers Prev., 10, 1117–1120. Paterson MEL, Wade-Evans T, Sturdee DW, Thom MH and Greendale GA, Reboussin BA, Slone S, Wasilauskas C, Studd JW. (1980). BMJ, 22, 822–824. Pike MC and Ursin G. (2003). J. Natl. Cancer Inst., 95, Pearce CL, Wu AH, Wan P and Pike MC. (2004) In 30–37. preparation.

Oncogene Breast, endometrial and ovarian cancer prevention MC Pike et al 6391 Peto R. (1977). Origins of Human Cancer . Hiatt HH, Watson Russo J and Russo IH. (1994). Cancer Epidemiol. Biomarkers JD and Winsten JA (eds). Cold Spring Harbor Laboratory: Prev., 3, 353–364. Cold Spring Harbor, 1403–1428. Schairer C, Lubin J, Troisi R, Sturgeon S, Brinton L and Pettersson B, Adami HO, Bergstrom R and Johansson EDB. Hoover R. (2000). JAMA, 284, 691–694. (1986). Acta Obstet. Gynecol. Scand., 65, 247–255. Schildkraut JM, Calingaert B, Marchbanks PA, Moorman Piek JM, van Diest PJ, Zweemer RP, Jansen JW, Poort- PG and Rodriguez GC. (2002). J. Natl. Cancer Inst., 94, Keesom RJ, Menko FH, Gille JJ, Jongsma AP, Pals G, 32–38. Kenemans P and Verheijen RH. (2001). J. Pathol., 195, Schildkraut JM, Cooper GS, Halabi S, Calingaert B, Hartge P 451–456. and Whittemore AS. (2001). Obstet. Gynecol., 98, 85–90. Pike MC. (1987). J. Chron. Dis., 40, 59S–69S. Schouten LJ, Goldbohm RA and van den Brandt PA. (2003). Pike MC, Kolonel LN, Henderson BE, Wilkens LR, Hankin Am. J. Epidemiol., 157, 424–433. JH, Feigelson HS, Wan P, Stram DO and Nomura AMY. Shimizu H, Ross RK, Bernstein L, Pike MC and Henderson (2002). Cancer Epidemiol. Biomarkers Prev., 11, 795–800. BE. (1990). Br. J. Cancer, 62, 451–453. Pike MC, Krailo MD, Henderson BE, Casagrande JT and Siiteri PK and MacDonald PC. (1973). Handbook of Physiol- Hoel DG. (1983). Nature, 303, 767–770. ogy Section 7, Part I, Vol. 2 American Physiological Society: Pike MC, Pearce CL, Peters R, Cozen W, Wan P and Wu AH. Washington, DC, pp. 615–629. (2004). Fertil. Steril., 82, 186–195. Siiteri PK, Hammond GL and Nisker JA. (1981). Hormones Pike MC, Peters RK, Cozen W, Probst-Hensch NM, Felix JC, and Breast Cancer . Pike MC, Siiteri PK and Welsch CW Wan PC and Mack TM. (1997). J. Natl. Cancer Inst., 89, (eds). Cold Spring Harbor Laboratory: Cold Spring Harbor, 1110–1116. pp. 87–106. Pike MC, Ross RK, Lobo RA, Key TJA, Potts M and Simeons ATW. (1954). Lancet, 2, 946–947. Henderson BE. (1989). Br. J. Cancer, 60, 142–148. Sit ASY, Modugno F, Weissfeld JL, Berga SL and Ness RB. Pike MC, Spicer DV, Dahmoush L and Press MF. (1993). (2002). Gynecol. Oncol., 86, 118–123. Epidemiol. Rev., 15, 17–35. Spicer D, Shoupe D and Pike MC. (1991). Contraception, 44, Purdie DM, Bain CJ, Siskind V, Russell P, Hacker NF, Ward 289–310. BG, Quinn MA and Green AC. (1999). Br. J. Cancer, 81, Spicer DV, Pike MC, Pike A, Rude R, Shoupe D and 559–563. Richardson J. (1993). Contraception, 47, 427–444. Purdie DM, Bain CJ, Webb PM, Whiteman DC, Pirozzo S Spicer DV, Ursin G, Parisky YR, Pearce JG, Shoupe D, and Green AC. (2001). Cancer Causes Control, 12, 855–863. Pike A and Pike MC. (1994). J. Natl. Cancer Inst., 86, Riman T, Dickman PW, Nilsson S, Correia N, Nordlinder H, 431–436. Magnusson CM and Persson IR. (2002a). Am. J. Epidemiol., Titus-Ernstoff L, Perez K, Cramer DW, Harlow BL, Baron JA 156, 363–373. and Greenberg ER. (2001). Br. J. Cancer, 84, 714–721. Riman T, Dickman PW, Nilsson S, Correia N, Nordliner H, Tornberg SA and Carstensen JM. (1994). Br. J. Cancer., 69, Magnusson CM, Weiderpass E and Persson IR. (2002b). J. 358–361. Natl. Cancer. Inst., 94, 497–504. Tung KH, Goodman MT, Wu AH, McDufﬁe K, Wilkens LR, Risch HA, Marrett LD and Howe GR. (1994). Am. J. Kolonel LN, Nomura AM, Terada KY, Carney ME and Epidemiol., 140, 585–597. Sobin LH. (2003). Am. J. Epidemiol., 158, 629–638. Risch HA. (1996). Gynecol. Oncol., 63, 254–257. Ursin G, Astrahan MA, Salane M, Parisky YR, Pearce JG, Rodriguez C, Calle EE, Fakhrabadi-Shokoohi D, Jacobs EJ Daniels JR, Pike MC and Spicer DV. (1998). Cancer and Thun MJ. (2002a). Cancer Epidemiol. Biomarkers Prev., Epidemiol. Biomarkers Prev., 7, 43–47. 11, 822–928. Waterhouse J, Muir C, Correa P and Powell J (eds). (1976). Rodriguez C, Patel AV, Calle EE, Jacob EJ and Thun MJ. Cancer Incidence in Five Continents . IARC Scientiﬁc (2001). JAMA, 285, 1460–1465. Publication No. 1 IARC: Lyon, France. Rodriguez GC, Nagarsheth NP, Lee KL, Bentley RC, Walmer Weiss NS, Lyon JL, Krishnamurthy S, Dietert SE, Liff JM DK, Cline M, Whitaker RS, Isner P, Berchuck A, Dodge and Daling JR. (1982). J. Natl. Cancer Inst., 68, 95–98. RK and Hughes CL. (2002b). J. Natl. Cancer Inst., 94, Welsch CW. (1981). Hormones and Breast Cancer . Pike MC, 50–60. Siiteri PK and Welsch CW (eds). Cold Spring Harbor Rodriguez GC, Walmer DK, Cline M, Krigman H, Lessey Laboratory: Cold Spring Harbor, pp. 299–315. BA, Whitaker RS, Dodge R and Hughes CL. (1998). J. Soc. Whiteman DC, Siskind V, Purdie DM and Green AC. (2003). Gynecol. Invest., 5, 271–276. Cancer Epidemiol. Biomarkers Prev., 12, 42–46. Rosner B, Colditz GA and Willett WC. (1994). Am. J. Whittemore AS, Harris R, Itnyre J and the Collaborative Epidemiol., 139, 819–835. Ovarian Cancer Group. (1992). Am. J. Epidemiol., 136, Ross RK, Paganini-Hill A, Wan PC and Pike MC. (2000). J. 1184–1203. Natl. Cancer Inst., 92, 328–332. Williams DB, Voigt BJ, Yao SF, Schoenfeld MJ and Judd HL. Royar J, Becher H and Chang-Claude J. (2002). Int. J. Cancer, (1994). Obstet. Gynecol., 84, 787–793. 95, 370–374. Zhou H, Luo M, Schonthal AH, Pike MC, Stallcup Russo IH, Koszalka M and Russo J. (1991). Br. J. Cancer, 54, MR, Zheng W and Dubeau L. (2002). Cancer Biol., 1, 481–484. 300–306.

Oncogene