Offspring sex and risk of epithelial ovarian cancer: a multinational pooled analysis of 12 case-control studies

OCAC was established in 2005 to promote collaborative research on epidemiologic and genetic factors associated with EOC( 32 ). The present analysis included participant-level data for parous women from 12 OCAC case-control studies conducted in Australia, Canada, Germany, the United Kingdom, and the United States with available information on offspring sex( 33 – 45 ). Characteristics of the studies are shown in Table 1 . Because offspring sex was inconsistently reported for non-singleton births across studies and because non-singleton births may differentially impact EOC risk relative to singleton births, we excluded subjects with any non-singleton births (n=528) from current analyses, resulting in 16,343 parous women with all singleton births. We then excluded women missing covariate data (n=35) and women missing offspring sex information (n=335), resulting in a total sample of 15,973 participants for data analysis (6,872 EOC patients and 9,101 controls). All participants provided informed consent and all participating institutions obtained approval from relevant ethics committees. Information on offspring sex for each pregnancy lasting six months or longer (full-term) was self-reported. Based on our previous work, we classified women according to the number of male offspring( 26 ). Ever having given birth to a boy was defined as reporting at least one male offspring among all singleton full-term births. Giving birth to all boys was defined as reporting a male offspring for each full-term, singleton pregnancy. The number of boys was calculated by summing the total number of pregnancies resulting in male offspring. The number of girls was calculated by subtracting the number of boys from the total number of full-term pregnancies. The fraction of births that were boys was defined as the total number of male offspring divided by the number of full-term pregnancies. Information on other relevant variables and potential confounders was obtained from the OCAC core dataset and included age at diagnosis (cases) or interview (controls), race, education, body mass index (BMI) at 18 years of age, recent BMI (defined as previously reported as BMI 1 year prior or 5 year prior to diagnosis/interview or at diagnosis/interview( 46 )), total duration of oral contraceptive (OC) use, number of full-term pregnancies (parity), family history of ovarian or breast cancer, smoking status, and history of endometriosis, adult acne, hirsutism, polycystic ovary syndrome (PCOS), and irregular periods. We used unconditional logistic regression to estimate odds ratios (ORs) and their 95% confidence intervals (95%CIs) for associations between bearing male offspring and EOC risk among parous women. The main multivariate model was adjusted for study site, age at reference (continuous), duration of OC use (never, less than 1 year, 1–4 years, 5–9 years, 10+ years), parity (1, 2, 3, 4, 5+ offspring) and race (white, black, Asian, other). We also considered adjustment for additional ovarian cancer risk factors including education (less than high school, high school, post-high school, college graduate, post graduate), family history of ovarian or breast cancer (yes/no), history of breastfeeding (yes/no), BMI at 18 (=30 kg/m 2 ), recent BMI (=30 kg/m 2 ), history of endometriosis (yes/no), history of irregular periods (yes/no), history of polycystic ovary syndrome (PCOS), adult acne, or hirsutism (yes/no), smoking history (never, ever), and age at menarche (=13 years). These factors did not change the association between bearing a male offspring and EOC risk in general by more than 10% and were therefore not included in final models. Where they did alter associations by more than 10%, we present both the parsimonious model and the more adjusted model. Random effects meta-analyses across study sites of all cancer histotypes showed no evidence of heterogeneity (I 2 =0.0%; p-het=0.57 Figure 1 ). Consequently, all analyses were performed using the pooled dataset adjusted for study site. We performed polytomous logistic regression to evaluate associations between bearing male offspring and EOC risk by the main histotypes (high-grade serous, mucinous, endometrioid, clear cell). We further stratified analyses by number of full-term births to separate associations with offspring sex from those with parity. We also explored models containing terms for total number of male and total number of female offspring and models containing terms for total number of full-term pregnancies and fraction of boys. To identify potential interactions between offspring sex and hormonally-associated exposures for EOC in general and by specific histotypes, we performed stratified analyses by history of endometriosis (associated with excess estrogens( 47 ) or reduced progesterone( 48 )), history of acne or hirsutism or PCOS (associated with excess androgens( 49 – 51 )), age at menarche less than 13 (which is associated with excess estrogens and increased ovulations( 52 – 54 )), recent BMI greater than or equal to 30 kg/m 2 (which is associated with hormonal imbalances( 55 , 56 )), history of irregular periods (associated with hormonal dysregulation( 57 )), history of ever using oral contraceptives (associated with altered hormonal milieu( 58 – 60 )), and history of ever smoking cigarettes (associated with anti-estrogenic effects( 61 )). Interactions and linear trends were assessed with Wald statistics. Stata/SE version 15.1 (StataCorp, College Station, TX) was used to conduct all analyses. All tests were two-sided with significance level of 5%.

Among parous controls, the study-specific frequency of never bearing a male offspring ranged from 17% to 31%, whereas among parous cases it ranged from 19% to 36% ( Table 1 ). Compared to controls, women with EOC were less likely to have used OCs, had more than one child, attained a college education, reported a history of acne, hirsutism, or PCOS, and reported a history of irregular periods. Case women were more likely to have higher recent BMI, reported histories of endometriosis, and family histories of breast or ovarian cancer ( Table 2 ). Compared to bearing all females, ever having borne a male was not associated with EOC overall (OR=1.05; 95%CI=0.96–1.14; Table 3 ); however, bearing a male offspring was associated with increased risk of mucinous histotype (OR=1.25; 95%CI=1.02–1.54). This association strengthened when we further adjusted for hormonally-associated conditions (endometriosis, irregular periods, acne or PCOS or hirsutism, smoking, history of early menarche and recent BMI; OR=1.45; 95%CI=1.01–2.07). Similarly, giving birth only to boys was not associated with EOC risk overall, whereas compared to giving birth to at least one girl, bearing all male offspring was associated with increased risk of mucinous tumors (OR=1.29; 95%CI=1.07–1.55). The association was slightly strengthened when further adjusted for hormonally-associated conditions (OR=1.35; 95%CI=0.99–1.84). Increasing number of male offspring was associated with increasing risk of mucinous ovarian cancer in both the most parsimonious model (OR=1.16, 1.56, 1.55, for 1, 2 and 3 or more male offspring compared to all female offspring, respectively; trend-p = 0.006) and in a model additionally controlling for hormonally-associated conditions (OR=1.31, 1.84, 2.31, for 1, 2 and 3 or more male offspring, respectively; trend-p = 0.005). There were no associations between increasing number of male offspring and EOC risk overall or for any other histotypes. In models including separate quantitative terms for total number of male offspring and total number of female offspring, each additional offspring was associated with about an 8% decrease in EOC risk overall regardless of whether the offspring was male (OR=0.93; 95%CI=0.90–0.96) or female (OR=0.92; 95%CI=0.89–0.95) ( Table 3 ). While the point estimates for high-grade serous, clear cell, and endometrioid subtypes were similar for both male and female offspring, for the mucinous histotype, each additional female offspring was associated with a 12% decrease in risk (OR=0.88; 95%CI=0.810.96) whereas each male offspring was not associated with risk (OR=1.03; 95%CI=0.95–1.11). The results from models controlling for total number of full-term births also showed that a 25% increase in the fraction of births that were boys was associated with a 9% increase in risk of mucinous EOC (OR=1.09; 95%CI=1.03–1.16). Fraction of male births was not associated with risk of the other subtypes. Stratifying by number of offspring ( Table 3 ) yielded similar patterns of risk associated with increasing male offspring for the mucinous histotype. Among women with exactly one full-term birth, bearing a male offspring was associated with a 22% increased risk of mucinous cancer compared to bearing a female offspring. Among women with exactly two births, compared to bearing all female offspring, bearing exactly one male offspring was associated with a 16% increased risk of mucinous tumors, whereas bearing two male offspring was associated with a 58% increased risk (P-trend=0.01). For mucinous histotype, we further observed interactions with age at menarche ( Table 4 ). Compared to never giving birth to a boy, ever bearing a male offspring was associated with an increased risk of mucinous cancer among women with menarche before age 13 (OR=1.71, 95%CI=1.23–2.38) but no increased risk associated with menarche at a later age (OR=0.99, 95%CI=0.76–1.30; P-interaction=0.02). Results were similar when we examined interactions between menarche and giving birth to all boys (OR=1.55 for early menarche versus OR=1.08 for later menarche; P-interaction=0.08). Among women with menarche prior to age 13, increasing number of male offspring was associated with increasing risk of mucinous tumor (ORs for bearing 1, 2, 3+ male offspring: 1.54, 2.34, 2.24 compared to no male offspring; P-trend =0.002). Among women with later menarche no trend was observed (ORs for bearing 1, 2, 3+ male offspring: 0.94, 1.16, 1.20; P-trend=0.32; P-interaction=0.10). Consistent with this observation, each 25% increase in fraction of male offspring was associated with a significant 18% increase in mucinous cancer among women with earlier menarche but no increase in women with later menarche (P-interaction=0.01). We also observed an interaction between age at menarche and bearing female offspring, with each female offspring associated with a significant 21% reduced risk of mucinous tumors among women with earlier menarche but little or no association among women with later menarche (OR=0.79 versus 0.94 for each female offspring in women with and without early menarche, respectively; P-interaction=0.02). There was no interaction between age at menarche and bearing male offspring (OR=1.04 versus 1.01 for each male offspring in women with and without early menarche, respectively; P-interaction=0.51). No other interactions between hormonal-associated exposures and EOC were observed, except for self-reported history of acne or hirsutism or PCOS and risk of endometrioid cancer ( Table 5 ). Compared to bearing all female offspring, bearing at least one male offspring was associated with reduced risk of endometrioid cancer among women with a history of any of those conditions (OR=0.49, 95%CI=0.28–0.83), but an increased risk among women with no history of any of those conditions (OR=1.64 95%CI=1.14–2.34; P-interaction=0.003). Results were similar when we examined the interaction between reported history of acne/hirsutism/PCOS and number of male offspring (ORs for bearing 1, 2 or 3+ male offspring: 0.47, 0.52, 0.47 versus 1.69, 1.59. 0.78, for women with and without this history, respectively, P-interaction=0.007). An interaction was also observed between reported history of those androgenic conditions and bearing female offspring, with each female offspring associated with reduced endometrioid cancer risk in women with no reported history compared to those with such a history (OR=0.80 vs 1.02 for each female offspring in women without and with a history, respectively; P-interaction 0.03). There appeared to be no interaction between a history of those androgenic conditions and bearing male offspring (OR=0.82 vs 0.87 for each male offspring in women without and with a history, respectively; P-interaction=0.44).

In this pooled analysis of data from 6,872 parous women with EOC and 9,101 parous controls, sex of offspring was not associated with maternal EOC risk overall. However, bearing male offspring was associated with less protection against mucinous cancers. When examining the per-pregnancy association, offspring sex was not associated with EOC risk overall or for high-grade serous, clear cell, and endometrioid histotypes, but was associated with risk of mucinous tumors. In particular, bearing female offspring was associated with decreased risk of mucinous tumors among parous women, whereas bearing male offspring appeared to have no relation to that histotype. We observed no interactions between offspring sex and hormonally-associated exposures, except among women with mucinous tumors and menarche prior to age 13 and among women with endometrioid tumors and a history of acne, hirsutism, or PCOS. Among women with menarche before age 13, bearing male children was associated with higher risk of mucinous cancer than in women with later menarche. Among women with a history of acne, hirsutism, or PCOS, bearing male children was associated with lower risk of endometrioid cancer than in women without those conditions. Five studies have reported the association between offspring sex and ovarian cancer risk( 26 – 30 ), including two studies included in this pooled analysis (HOPE and AUS). In the HOPE Study, conducted in western Pennsylvania, USA from 2003–2008, compared to bearing all female offspring, bearing any male offspring was associated with lower risk of EOC (OR=0.92) and bearing all male offspring was associated with even lower risk (OR=0.86)( 30 ). A earlier population-based study of 511 cases and 1136 controls conducted in eastern Pennsylvania, USA from 1994–1998 by the same group reported similar findings – relative to all female offspring, bearing all male offspring was associated with decreased EOC risk (OR=0.80)( 26 ). These findings were supported by a nested case-control study within the population-based Swedish Fertility Register that included 7,407 women diagnosed with EOC between 1961 and 2001 and 37,658 controls( 27 ): compared to bearing all female offspring, bearing a male child was associated with reduced EOC risk in a dose-response fashion (ORs: 0.92, 0.87, 0.82, for 1, 2 or 3+ boys, compared to all girls)( 27 ). In contrast, the Australia-wide population-based study (AUS) conducted between 2002 and 2005 and included in this pooled analysis reported no association between offspring sex and EOC for parous women in general but a 2-fold increased risk of the mucinous histotype associated with bearing only male offspring( 29 ). Notably, excluding AUS data from the current analysis did not appreciably affect the observed association with mucinous tumors. A population-based cohort study of 5,092 EOC cases in the Norwegian national registry also reported no EOC-offspring sex association in general( 28 ). However, that study reported an increased risk of endometrioid tumors among women who gave birth only to girls compared to those who gave birth only to boys (incidence ratio 1.35 based on 475 cases). Although there are histotype differences in the magnitude of the protective effect, greater parity has consistently been associated with reduced EOC risk( 7 , 8 ), especially among non-mucinous disease; however, the mechanism underlying this association remains unknown. Two theories have dominated the literature: suppressed ovulation( 9 ) and lowered gonadotropin levels( 62 ). Pregnancy, regardless of fetal sex, should equally affect ovulation and gonadotropin secretion; thus, our results suggest the possibility of additional mechanisms. Reducing inflammation( 12 ) and altering circulating steroid hormones( 11 ) have been postulated. During pregnancy, both maternal hormonal and immune milieus differ by fetal sex. Carriage of a male fetus is associated with lower maternal levels of estradiol and hCG( 14 , 15 , 18 ) and higher maternal levels of progesterone( 16 ) and testosterone( 19 ). While the role of hCG in EOC etiology is unclear, progesterone is believed to protect against EOC while estrogens and androgens may increase risk( 11 ) in a histotype-specific way( 20 , 21 ). Whether the observed maternal hormonal differences by fetal sex are large enough to matter in the context of the high hormonal levels of pregnancy is unknown. Women carrying male fetuses also exhibit more proinflammatory/proangiogenic immune milieus than women carrying female fetuses( 17 ). Pregnancy outcomes also vary by offspring sex, with preterm birth, higher birth weight, and gestational diabetes associated with males( 63 – 65 ), and increased risk of maternal hypertensive disorders and asthma flares associated with females( 66 , 67 ). Genetic and metabolic profiles of the placenta also vary by fetal sex( 68 ), and both hormones and cells derived from the fetoplacental unit persist in maternal circulation for years after pregnancy ends( 69 ). Moreover, male-origin microchimerism, which arises predominantly but not exclusively from fetal cells acquired during pregnancy( 70 ) and persists for decades after pregnancy( 71 ), has recently been associated with reduced rates of ovarian cancer( 72 ). Fetal sex also influences maternal physiology( 22 , 23 ), and pregnancy conditions that differ by fetal sex, such as preeclampsia and gestational diabetes, may impact future maternal health outcomes( 24 , 25 ). Together, these observations suggest that fetal sex-based differences can have long-term health consequences and support a potential link between offspring sex and EOC risk. Despite this apparent biologic plausibility, the results of this study did not show any overall relationship between offspring sex and EOC risk. However, we did observe relationships with offspring sex for the mucinous histotype in general and specifically for women with menarche prior to age 13. We further observed an association for endometrioid tumors in relation to maternal androgenic conditions. It is now accepted that while pregnancy protects against EOC in general, the protection varies by histotype. In the Ovarian Cancer Cohort Consortium (OCCC), ever bearing offspring provided a 31% decrease in risk in general, with a greatest protection seen for the clear cell histotype (RR=0.35, 95%CI:0.27–0.47) and the least protection observed in the serous histotype (RR=0.81, 95%CI=0.73–0.90) ( 8 ). The Million Women Study also reported a differing protective effect against EOC associated with every bearing offspring based on histotype, with the greatest effect seen among clear cell cases and the least seen among serous cases ( 7 ). Both studies also report histotype differences based on the number of offspring. Given these differences in protective effect of pregnancy by hisotype, it is possible that the relationship between offspring sex and EOC could also vary by histotype. Thus, while our histotype-specific observations are plausible, the underlying biologic reasons for these observations are unclear. Mucinous EOC is a relatively infrequent histotype, representing some 5–20% of cases( 73 ); however, epidemiologic evidence supports a substantially different risk-factor profile than that of the other histotypes( 74 ). Notably, apart from pregnancy, the relationships between hormonal exposures and mucinous tumors are less pronounced or perhaps nonexistent compared to other histotypes( 74 ), suggesting that alteration in the hormonal milieu may not account for our mucinous-disease findings in general and among women with menarche prior to age 13. In addition to higher endogenous estrogen exposure, earlier age at menarche is associated with earlier and more prolonged ovulation( 52 , 53 ). That observation, however, cannot explain the mucinous-specific association because increasing lifetime ovulations are associated with increased ovarian cancer risk overall( 75 – 77 ). Moreover, histotype-specific results show no relationship between lifetime ovulations and the mucinous subtype( 77 ). Similarly, it is unclear why the relationship between offspring sex and endometrioid tumors should vary based on history of androgenic conditions, as endometrioid tumors are more closely associated with estrogenic exposures( 78 – 80 ) and possibly higher circulating androgen levels in the post-menopause( 20 ). Regardless of the underlying biology, our findings underscore the need to further understand the mechanisms whereby pregnancy impacts EOC risk. Moreover, they reflect the heterogeneous etiologic nature of ovarian cancer( 81 ), which is no longer believed to be a single disease but a group of diseases with separate etiologic origins. EOC histotypes exhibit differing clinical behavior and are believed to have different or differentially evolved cells of origin leading to distinct carcinogenic pathways( 82 ). Epidemiologic studies further support the multifactorial origin of EOC, with most well-established risk factors exhibiting substantial heterogeneity by histotype( 8 , 74 ). Our results lend further population-based support to the distinct etiology of EOC histotypes, and in particular for that of mucinous tumors compared to the others( 8 , 74 ). A strength of the present study is the use of participant-level data from 12 population-based case-control studies spanning three continents. The large sample size resulted in increased statistical power to examine histotype-specific associations, which individual studies could not adequately do. In addition, pooling data from population-based case-control studies with detailed lifestyle, reproductive, and medical history data enabled us to control for potential confounders and to stratify by hormonally-associated exposures, which the population-based registry studies were unable to do. The included studies were all population-based, and the majority of studies used in-person interviews to obtain data on offspring sex and other exposures, increasing the generalizability of findings. Study-specific data were carefully cleaned, harmonized, and entered into a single dataset, further increasing confidence in the quality of the data and allowing us to adjust for a single set of standard confounders. Finally, all available OCAC studies with information on offspring sex were included, thus mitigating the possibility of publication bias. Despite these strengths, some limitations should be considered. First, data were self-reported; thus, potential confounding variables could be influenced by case/control status, which could distort our findings. Moreover, due to missing data, we were not able to assess relationships between offspring sex and some factors that may influence ovarian cancer risk, such as age at first pregnancy. We also can not eliminate the possibility of unknown confounders influencing results. Selection bias is also a concern as controls participating in these studies may differ from cases by factors related to offspring sex or EOC risk, including unknown factors that could not be accounted for in the analyses. Validation in prospective cohorts is needed to address these concerns. Because our study population was predominately white, we could not evaluate the impact of offspring sex in non-white women and how it may differ across race. Finally, we cannot eliminate the possibility that our findings are due to chance. In conclusion, offspring sex appears to affect differentially EOC risk based on histotype and, possibly, in combination with other host factors. Our findings support the distinct etiologic pathways among EOC histotypes and suggest that current etiologic models of EOC may be incomplete. Our findings also suggest the need to better understand how pregnancy affects EOC risk. Confirmation of these findings in prospective cohorts is needed to improve our understanding of EOC etiology, thereby paving the way for new avenues of prevention research for this highly fatal disease.

Ovarian cancer is the fifth most common cancer among women in developed countries and the most fatal gynecological malignancy( 1 ). In 2018, more than 295,000 women were newly diagnosed with the disease and over 185,000 women died from it worldwide( 1 ). More than 70% of cases are diagnosed at late stages when 5-year survival is less than 30%( 2 ). This high fatality coupled with the lack of a screening test for early detection( 3 ) makes it critical to understand risk factors in order to help inform prevention strategies( 4 ). Ever bearing children is associated with about a 30% decrease in risk of epithelial ovarian cancer (EOC) in general ( 5 ) and increasing parity increases protection ( 6 ), although the magnitudes of the relationship vary by histotype ( 7 , 8 ). The exact mechanism underlying the protective effect of pregnancy remains unknown, although it is frequently attributed to ovulation suppression that accompanies pregnancy( 9 ). However, an ovulation alone cannot explain the magnitude of the protective effect( 10 ), suggesting that other pregnancy-associated factors may impact EOC risk. Alterations in the maternal hormonal and immune milieus may be such factors( 11 – 13 ). Fetal sex potentially affects these environments during pregnancy( 14 – 21 ), can impact maternal physiology( 22 , 23 ), and is associated with conditions that have long-term maternal health consequences( 24 , 25 ). Together these data support the possibility that offspring sex may impact maternal EOC risk. Few epidemiologic studies have explored the relationship between offspring sex and EOC, and results have been inconsistent( 26 – 30 ). Methodological limitations including small sample sizes overall and for specific histotypes may account for these disparate findings. EOC is a heterogeneous disease consisting of distinct histotypes exhibiting varied risk factor profiles( 8 ) and likely having distinct etiologic pathways( 31 ). The main aim of this study was to evaluate the associations between offspring sex and EOC in an international collaborative investigation using pooled data from 12 case-control studies participating in the Ovarian Cancer Association Consortium (OCAC). Secondarily, we wished to evaluate associations by histotype. The large sample size of the pooled analysis enabled more robust estimates of the associations between offspring sex and EOC overall and by histotype than previously reported. In addition, the pooled analysis enabled exploration of potential interactions with hormonally-associated exposures.