Journal of the National Cancer Institute, Vol. 90: 1774-86, No. 23, December 2, 1998

REVIEW

Hormonal Etiology of Epithelial Ovarian Cancer, With a Hypothesis Concerning the Role of Androgens and Progesterone

Harvey A. Risch

In the United States, ovarian cancer is the fourth most frequent cause of cancer death among women, following lung, breast, and colorectal cancers. Each year, approximately 26 000 women are diagnosed with ovarian cancer and 14 000 die of it. Germline mutations in BRCA1, BRCA2, or other genes have been implicated in a small fraction of cases.
However, it has been suggested that, for the great majority of patients, the risk of epithelial ovarian cancer could be related to ''incessant ovulation'' (i.e., to the chronically repeated formation of stromal epithelial clefts and inclusion cysts following ovulation) or to some type of hormonal stimulation of ovarian epithelial cells, either on the surface of the ovary or within ovarian inclusion cysts, possibly mediated through excessive gonadotropin secretion. From the evidence to date, the relative importance of these two hypotheses -incessant ovulation and gonadotropin stimulation- cannot be distinguished. While either or both may play a role in the development of ovarian cancer, it appears that an additional major factor must also be involved. The purpose of this review is to evaluate evidence for and against the incessant ovulation and gonadotropin hypotheses, as well as to consider the possibility that risk of ovarian cancer may be
increased by factors associated with excess androgenic stimulation of ovarian epithelial cells and may be decreased by factors related to greater progesterone stimulation. Many features of the evidence bearing on the pathophysiology of ovarian cancer appear to support a connection with androgens and progesterone.

Progesterone
Evidence for a possible protective role of progesterone in the etiology of ovarian cancer starts with consideration of the increased sex hormone activity during pregnancy. Over the first month of pregnancy, maternal LH and FSH decline strongly with the increase in trophoblast hCG (162). The hCG also stimulates the corpus luteum to continue producing progesterone and not regress (102). After the seventh week, the luteal-placental shift occurs in which the functional capacity of the corpus luteum
of pregnancy drops, while the massive placental production of progesterone during pregnancy begins (102). In addition, the placenta extracts maternal (and later, fetal) adrenal androgens, which remain at stable maternal serum concentrations while both production and utilization rates increase; maternal serum estrone and estradiol are made from the adrenal androgens (102). During pregnancy, the placental synthesis thus causes 10-fold increases in maternal circulating progesterone levels (102). Maternal testosterone and androstenedione levels increase some twofold to threefold, although most of the testosterone is bound to the pregnancy-induced higher levels of sex hormonebinding globulin, preventing virilization of female fetuses (102).
These maternal ovarian androgens are in any case dwarfed by the huge estrogen and progesterone concentrations. In terms of the pathogenesis of ovarian cancer, we suggest that the additional protective aspect of pregnancy not mediated through suppression of ovulation may be due to the 8-9 months of elevated progesterone. As we have noted, it seems unlikely to be due to the pregnancy estrogens, since most of the evidence relating estrogens to risk of ovarian cancer (as well as to endometrial cancer and perhaps breast cancer) points either to no effect or to increase in risk.
With respect to oral contraceptive usage, it is uncertain whether the synthetic progestational agents in these preparations directly convey the decreased risk consistently seen according to duration of use; the magnitude of risk decrease is consistent with protection due to ovulation suppression or to androgen reduction (see above). The contraceptive progestins vary somewhat in their androgenic and estrogenic properties (147,148). Those progestins considered to be relatively androgenic in terms of their clinical side effects (e.g., norgestrel and levonorgestrel) (163) also appear to lower total and free serum testosterone the most (148). Epidemiologic studies (164,165) have shown no difference in the reduction of ovarian cancer risk between norgestreltype contraceptives and other combined agents. More than 75% of oral contraceptive usage during the 1960s-1970s was of progestins with ''low androgenicity'' (163,164). During oral contraceptive use, endogenous progesterone synthesis seems to stay as low as that during the early follicular phase of the menstrual cycle, without follicle maturation or corpus luteum function (166). However, given that the progestational potency of the synthetic 19-nortestosterone progestins is more than 100 times that of progesterone (167,168) and that serum levels of progestins absorbed from oral contraceptives are comparable to lutealphase progesterone levels [e.g., 5 ng/mL (78)], the net progestational environment within the ovary is likely to be quite high (169). Thus, the decreased risk of ovarian cancer with oral contraceptive use could also be due to the cyclic progestational climate.
Another piece of evidence suggests that combined oral contraceptives do offer ovarian cancer protection beyond that potentially from suppression of ovulation. A case-control study was large enough to have identified sufficient numbers of subjects who had used progestin-only types of oral contraceptives (165). These progestin-only formulations do not totally suppress ovulation and some ovulatory cycles typically occur (169); up to 40% of women using this method can have regular ovarian function, with normal estrogen and luteal-phase progesterone synthesis (166). In the case-control study, relative to never use of progestin-only contraceptives, the risks were 0.39 for use less than 3 years' duration and 0.21 for use 3 years or longer, with trend P 4 .009. These reduced risks appear comparable to those of the combined oral contraceptives or perhaps a little more protective (165). Thus, these progestin-only contraceptives create a progestational hormonal environment with a
reduced risk that cannot in total be attributed to ovulation suppression.
Given that combined oral contraceptives convey a similar degree of protection but with less ovulation, we infer that risk reduction associated with ovulation suppression cannot comprise the total protection given by the combined preparations and that the net benefit is probably due to the progestational component.
Nevertheless, a similar degree of protection is not yet clearly seen for usage of depot medroxyprogesterone acetate (DMPA). DMPA is a long-acting 17-acetoxy progesterone compound that suppresses endogenous progesterone synthesis and ovulation;
estradiol levels remain in the early- to mid-follicular-phase range (<100 pg/mL) (170). Serum levels of DMPA stay about 1 ng/mL for 3 months after injection; these levels inhibit the midcycle peak in gonadotropins but do not seem to change basal LH and FSH levels (170). Only three studies have examined usage of DMPA and risk of ovarian cancer. A small case-control study in Shanghai (65) found an elevated OR of 2.8 (95% CI - 0.9-8.5), although few subjects had ever been exposed, and even use of combined oral contraceptives was not found to be protective in that study. In a follow-up study of 5000 black American women, an OR of 0.8 (95% CI40.1-4.6) was seen for ever use of DMPA (171). Last, the large and more definitive World Health Organization international collaborative case-control study observed for nonmucinous ovarian cancer a significantly decreased risk of 0.42 (95% CI 4 0.15-0.96) with ever usage (172). DMPA use may thus protect against the development of
ovarian cancer, although further studies are needed to confirm this fact.
The effects of physical exercise may also bear on the hypothesis of progesterone activity and ovarian cancer. We return to the prospective study of 31 000 Iowa women followed for 7 years (150). In addition to the association observed with waist-to-hip ratio, a significant increasing trend in risk of ovarian cancer was seen according to increasing value of an index of usual physical activity. Other studies have also suggested increased risk with employment in jobs categorized as having moderate (compared with low) physical activity levels: manual workers (173), physical education teachers (174), and jobs with little sitting time (175). Physical activity may not be related to serum androgens (or progesterone) postmenopausally (176,177) but premenopausally is associated with a shortened luteal phase (178-182), resulting in lower luteal progesterone levels (183,184). This finding applies both to female nonathletes as well as athletes.
Even moderate recreational physical activity without amenorrhea or other menstrual disturbances is associated with decreased progesterone levels (185). Women with menstrual cycles shortened by decrease in length of the luteal phase would also spend relatively greater proportions of time in the follicular phase and therefore may have somewhat more ovarian exposure to androgen production (186). Intense physical activity also produces transient elevations of serum testosterone and other androgens
(187). However, if ovulation is indeed involved in the etiology of ovarian cancer, women whose regular physical activity is intense or frequent enough to cause amenorrhea may be at decreased risk due to the suppression of ovulation.
As we have noted above, a study using monoclonal antibodies methods showed that virtually all specimens of normal ovarian epithelium contained progesterone receptors (127). Defects in the progesterone receptor could lead to reduced effectiveness of available progesterone and thus to increased risk of ovarian cancer according to our hypothesis. This finding has apparently been seen: A relatively common germline polymorphism variant in the hormone-binding domain of the progesterone receptor was associated with twofold increased risk (P<.025) (188). While this finding relating to the Alu insertion was not subsequently confirmed by a second group of investigators (189), preliminary analysis of sister-matched ovarian cancer case-control study data of the author also shows a 60% increased risk for women with this variant (Risch HA: unpublished data).
Finally, it is interesting to consider the effects of multiple gestation. Women who have delivered (naturally occurring) dizygotic twins appear to have higher gonadotropin levels during their reproductive years (190-194) and in general may be more likely to double ovulate (195) compared with women who have had singleton pregnancies only. Thus, they should be at increased risk of ovarian cancer according to either the incessant ovulation or gonadotropin hypotheses. Such does not appear to be the case. In a record-linkage study of mothers of dizygotic twins, no excess of ovarian cancer cases appeared (196). In fact, a case-control study that examined history of twin pregnancy found somewhat decreased risk with this factor (total parityadjusted OR40.68; 95% CI40.33-1.38) (62), and another by the author (56) also suggests decreased risk for nonmucinous ovarian cancer (parity-adjusted OR 4 0.42; 95% CI 4 0.14- 1.26) (Risch HA: unpublished data). Some evidence exists that premenopausal women with a history of twinning may have greater follicular-phase serum progesterone levels (195), and serum progesterone appears to be higher after double compared with single ovulations (197). Twin pregnancies also involve greater daily production and serum levels of progesterone (198- 199). Thus, reduced risk for women who have had twins could be conveyed through greater ovarian progesterone exposures.