History of Infertility and Risk of Endometrial Cancer in the Women's Health Initiative

During 25 years of follow-up 1,622 incident cases of endometrial cancer were ascertained. Approximately 17.8% of participants reported a history of infertility. Participants with a history of infertility had a slightly earlier age at menopause (49.9 vs 50.4 years), were less likely to have a BMI ≥30 (25.9 vs 28.3%), were less likely to report regular periods (77.4 vs 85.6%), and were more likely to have a history of smoking (52.1 vs 49.1%) compared to parous participants without a history of infertility ( Table 1 ). A modest, but not statistically significant association was observed between overall infertility and incident endometrial cancer (HR=1.12; 95% CI=0.99–1.26) ( Table 2 ). In sensitivity analysis when nulliparous participants were included in the reference group results were similar (HR=1.10; 95% CI=0.97–1.24). When histologic type of endometrial cancer was investigated, the association with infertility history was slightly stronger among endometrioid endometrial cancer; however, results were not statistically significant (endometrioid: HR=1.13; 95% CI=0.98–1.29 and non-endometrioid: HR=1.04; 95% CI=0.77–1.41) ( Table 2 ). When specific infertility diagnoses were examined, we did not observe any statistically significant associations between type of infertility diagnosis and incident endometrial cancer ( Table 3 ). The association between history of infertility and endometrial cancer was strongest among those with a BMI ≥25 (HR=1.15; 95% CI=0.99–1.33), compared to those with a BMI <25 (HR=0.97; 95% CI=0.77–1.23), although this difference was not statistically significant (p effect modification =0.25). When BMI was further stratified, the associations between those with a BMI 25–29.9 and BMI ≥30 were similar (HR=1.14; 95% CI=0.90–1.43 and HR=1.18; 95% CI=0.97–1.43, respectively). There were 1,809 prevalent endometrial cancers at study baseline. As risk factors may differ between earlier and later onset endometrial cancers we examined the association between history of infertility and endometrial cancer diagnosed before study baseline. The odds of an endometrial cancer diagnosis were 19% higher among those reporting an infertility diagnosis compared to those who were parous with no infertility (odds ratio [OR]=1.19; 95% CI=1.06–1.34). This association was strongest among those reporting an infertility diagnosis due to endometriosis (OR=2.42; 95% CI=1.83–3.19) ( Table 4 ).

The WHI is an ongoing prospective cohort study of 161,808 women who were postmenopausal and between the ages of 50–79 years old at study baseline. Participants originally enrolled between 1993 and 1998 in one of three randomized clinical trials (CT) (hormone therapy, dietary modification, and calcium/vitamin D supplementation; n=68,132) or an observational study (OS) (n=93,676) in 40 United States (U.S.) clinical centers.( 14 ) Women were excluded at baseline if they had medical conditions predictive of survival time of less than 3 years (e.g. class IV congestive heart failure, severe chronic lung disease with jaundice or ascites). Women were excluded from the CT if they had invasive cancer in the last 10 years. At the conclusion of the CTs in 2005, CT and OS participants were invited to consent for further follow-up in the WHI Extension Studies. At baseline and throughout study follow-up, participants have provided extensive information including demographics, medical history, diet, medication use, lifestyle, and behavioral measures. The Institutional Review Board (IRB) at each study site approved the protocols and participants provided written informed consent. Additional consent to review medical records was obtained through signed written consent. This study was reviewed and approved by the IRB of the Fred Hutchinson Cancer Center (approval number: 3467-EXT) and this analysis was approved by the University of Arizona IRB (Protocol number: 2011237760). For the incidence analysis, participants were excluded if they were missing data on infertility (n=1,644), missing data on prevalent cancer at baseline (n=1,386), had prevalent endometrial cancer at baseline (n=2,024), were missing follow-up data (n=657), or had a hysterectomy at baseline (n=64,125) leaving 91,972 participants. For the main incidence analysis, we further limited to parous participants for those without infertility resulting in a final analytic sample of 82,871. For the prevalence analysis, participants were excluded if they were missing data on infertility (n=1,644), missing data on prevalent cancer at baseline (n=1,386), missing parity data (n=519) or were nulliparous among those without infertility (n=13,445), resulting in an analytic sample of 144,814. At study baseline, participants were asked whether they had ever tried to become pregnant for more than one year without becoming pregnant.( 15 ) Participants who responded “yes” were considered to have a history of infertility. Participants were then asked whether they visited a doctor or clinic because they could not get pregnant and whether a reason was found. We investigated heterogeneity in infertility diagnoses by looking separately at self-reported infertility types: ovulatory infertility, tubal or uterine factor infertility, endometriosis, male factor infertility, other, not known. Infertility, by definition, can only be diagnosed among people trying to conceive; therefore, in order to reduce misclassification of our exposure, our primary comparison group was parous women without a reported history of infertility, which we refer to as fertile women. In sensitivity analyses, we expanded our comparison group to include all women (parous and nulliparous) without infertility. All participants completed self-reported health assessment questionnaires annually (OS and Extension) or semi-annually (CT), which collected information on health outcomes including endometrial cancer. Participants who reported endometrial cancer were contacted by WHI field center staff to obtain additional details on health event dates, providers, and a signed release to obtain medical records. Self-reports of incident cancer diagnoses were then verified by a central review of medical records or pathology reports by trained physician adjudicators.( 16 ) At study baseline, participants also reported all prior cancer diagnoses. All invasive incident cancers were documented and coded according to primary site. Endometrial cancer histological subtypes were defined according to the International Classification of Disease for Oncology, 3 rd Edition. Tumor grade and stage were coded using the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) coding system.( 16 ) Vital status was collected periodically through follow-up of participants and surrogates by each clinical center. Systematic searches of the National Death Index were also conducted every two years. Clinical adjudicators determined causes of death through medical record and death certificate review. The National Death Index was also used to confirm cause of death or otherwise unreported deaths for participants who were lost to follow-up. We examined two histologic subtype groups: endometrioid which included endometrioid adenocarcinomas, other adenocarcinomas, and adenosquamous cell carcinomas, and non-endometrioid which included serous adenocarcinoma, papillary serous carcinoma, clear cell carcinoma, carcinosarcoma, mixed cell adenocarcinoma, mullerian mixed tumor, and mesodermal mixed tumor. In the primary analysis, Cox proportional hazard models (described hereafter as the incident analysis) with age (months) as the time scale were used to calculate hazard ratios (HR) and 95% confidence intervals (CI) to estimate the risk of endometrial cancer (overall and by histologic subtypes) with parous participants without a history of infertility as the reference group. In sensitivity analyses, we expanded our comparison group to include nulliparous women without a history of infertility. The proportional hazard assumption was tested and met using a likelihood ratio test. Given that in this postmenopausal study population all infertility diagnoses occurred prior to study enrollment and that risk factors may differ between earlier and later onset endometrial cancers, we conducted secondary analyses using logistic regression examining endometrial cancer cases diagnosed before study baseline (described hereafter as the prevalent analysis). This analysis also allowed more power when examining specific infertility diagnoses. In both the incident and prevalent analyses model 2 was a priori adjusted for variables considered confounding variables or variables associated with WHI study selection including: education (less than high school, high school or GED completed, vocational training, technical school or some college, college degree or more), oral contraceptive use (never or ever), age at menarche (10 or less, 11–12, 13–14, 15+), BMI (18.5–24.9, 25–29.9, ≥30 kilogram [kg]/meter [m] 2 ), smoking status (never, former, current), marital status (married/partnered ever or never), physical activity (continuous metabolic equivalent [MET]-hours/week), alcohol use (non-drinker, past drinker, <1 drinks/week, 1 to <7 drinks/week, 7+ drinks/week), and WHI trial arm (OS or CT). Missing covariates were imputed using multiple imputation by chained equations (MICE)( 17 ). Given that menstrual cycle regularity (i.e. having periods occur about once per month excluding times when participant was pregnant or taking birth control) is a consequence of some infertility types in Model 3 we additional adjusted for menstrual cycle regularity (yes regular, sometimes regular/sometimes irregular, not regular). Since prior endometrial cancer risk factor associations have been shown to differ by BMI,( 18 – 20 ) multiplicative effect modification was examined comparing those with a BMI <25 to BMI ≥25. We also investigated the association with specific infertility diagnoses (ovulatory infertility, tubal or uterine factor infertility, endometriosis, male factor infertility, other, not known). In primary analysis we examined only those who reported a singular cause of infertility (e.g., only tubal/uterine factor infertility) and in secondary analysis allowed participants to contribute to multiple infertility diagnoses (e.g., tubal/uterine factor infertility and male factor infertility). Data is available through the WHI online resource, https://www.whi.org/datasets , while the WHI remains funded and indefinitely through BioLINCC, https://biolincc.nhlbi.nih.gov/studies/whi_ctos/ . Eligible researchers may download the data directly at the WHI online resource. Other researchers may download the publicly available data through BioLINCC, in accordance with NHLBI’s BioLINCC guidelines.

Among this cohort of postmenopausal women, we observed no statistically significant association between history of infertility and incident endometrial cancer. There was an association between history of infertility and prevalent endometrial cancer cases (i.e., those that occurred before study baseline), with the strongest association for infertility diagnosis due to endometriosis. The associations observed were robust to the choice of reference group (parous participants reporting no infertility and any participant reporting no infertility) and the effect estimates for overall endometrial cancer risk for the incident and prevalent analysis were similar (1.12 vs 1.19, respectively). In one of the most comprehensive examinations of this association to date, the Epidemiology of Endometrial Cancer Consortium (E2C2), conducted a pooled analysis of 14 studies (12 case-control studies and two cohort studies), observing that those who reported infertility had higher risk of endometrial cancer, even after accounting for nulliparity (odds ratio [OR]=1.20; 95% CI=1.13–1.33).( 7 ) Other studies have suggested a stronger relation between infertility and endometrial cancer than observed in our study (HR=1.12) or E2C2; however, these results have come from predominantly premenopausal populations drawn from gynecology and infertility clinics.( 2 , 4 – 6 , 8 , 9 ) These clinic-based results may not be generalizable to the majority of people at risk for endometrial cancer as the mean age of endometrial cancer is 60 years old, well after the age of menopause for most women. Within our study we observed differences in our incidence versus prevalent analyses. In the incidence analysis we examined the association between infertility and endometrial cancer in an entirely postmenopausal population. In contrast, in our prevalent analysis we included pre- and post-menopausal endometrial cancer cases diagnosed before study baseline. However, there are potential limitations of the prevalent analysis of infertility and endometrial cancer. Participants report their endometrial cancer history and their infertility history concurrently, therefore there is the possibility of recall bias, where participants with a history of endometrial cancer overestimate their infertility history or vice versa, however we feel that this type of misclassification is unlikely. Moreover, to be eligible to enroll in the WHI, participants needed to have a predicted survival of at least 3 years and/or no invasive cancer in the last 10 years (for CT). Therefore, the observed association with infertility and prevalent endometrial cancer may not be generalizable to premenopausal endometrial cancer or aggressive subtypes of endometrial cancer with high mortality. However, it is important to acknowledge that while the point estimates were relatively similar between the two analyses (HR for incidence=1.12, OR for prevalent=1.19) the prevalent analysis included a larger sample size overall and for specific infertility diagnoses. Beyond power, the slightly different results between the two analyses could indicate that infertility, an exposure that occurs decades prior to the average at endometrial cancer diagnosis, may be more strongly associated with earlier onset endometrial cancer – an observation that is consistent with prior studies in premenopausal populations.( 2 , 4 – 6 , 8 , 9 ) Future research incorporating information on infertility and premenopausal endometrial cancer with sufficient statistical power to investigate infertility subtypes in warranted. When causes of infertility were examined in the E2C2, the strongest association in the pooled analysis was observed among those with infertility due to endometriosis (OR=1.80; 95% CI=1.34–2.42), with the authors noting that this risk was primarily limited to the younger age groups.( 7 ) This is consistent with our results in that infertility diagnosis due to endometriosis was associated with prevalent endometrial cancer (OR=2.42; 95% CI=1.83–3.19). These cases were on average younger than cases in our incidence analysis where no association was observed for endometriosis (OR=1.17; 95% CI=0.69–1.99). There are a number of complexities of studying infertility and endometrial cancer risk that we sought to address in our study. While infertility, usually defined as trying to become pregnant for 12 months without success, has been shown to be accurately recalled many years after diagnosis,( 21 ) there is the potential for misclassification for the comparison group. Women who are nulliparous may include both women who had infertility (both diagnosed and undiagnosed) and those who were nulliparous by choice or for other reasons. To address this in our primary analysis, we compared women with a history of infertility to parous women without a history of infertility to minimize the possibility of misclassification. Another complexity of studying infertility is that BMI is differentially associated with sub-types of infertility and is a strong risk factor for endometrial cancer.( 22 ) For specific infertility diagnoses, a higher BMI is inversely associated with endometriosis,( 23 ) while in contrast, there is a higher prevalence of overweight and obesity among those with PCOS,( 24 ) with Mendelian randomization analyses indicating that BMI is causally related to PCOS.( 25 ) These associations can make interpreting the association between overall infertility and endometrial cancer risk more complex, necessitating examining the association by infertility subtypes and adequately accounting for body size. Although we may have been underpowered to detect a statistically significant effect modification, our results indicate that the association between history of infertility and endometrial cancer risk may vary by BMI, with no association observed among those with a BMI<25 and the suggestion of an elevated risk among those with a BMI ≥25. Consistent with our findings, in the E2C2, when the association was examined by body size, the highest odds of endometrial cancer was among those with infertility and a BMI ≥30; however, as in our analysis, this difference was not statistically significant.( 7 ) These results are consistent with other studies that indicate that the risk factors for endometrial cancer may have different impacts depending on body size.( 18 – 20 ) Strengths of our study include the large cohort size, long follow-up period, and detailed information on covariates allowing for a robust examination of the association of infertility with endometrial cancer. The prospective study design avoids the potential for recall and selection biases that are inherent in case-control studies that collect data after cancer diagnosis and clinic-based studies that enroll selective populations. Limitations of this study include that participants were enrolled at ages 50 to 79 years and WHI excluded those with predictive survival of less than three years, thus creating a cohort at baseline that was generally healthier than the US population.( 26 ) We also did not have information on primary versus secondary infertility, age at infertility, or infertility related treatments which could provide insight into particular period or mechanisms of risk. In this study we only had information on infertility attributed to endometriosis and therefore did not have information on endometriosis diagnosis in the absence of infertility which may constitute a majority of endometriosis cases ( 27 ). Further studies should examine endometriosis with and without infertility in relation to endometrial cancer to more fully understand the underlying mechanisms. Overall, in a population of postmenopausal women, we did not observe any statistically significant association of infertility with endometrial cancer. Most specific causes of infertility were not associated with endometrial cancer risk; however, infertility diagnosis due to endometriosis was associated with endometrial cancer among prevalent cases, indicating that infertility-associated endometriosis may impact endometrial cancer risk among younger postmenopausal and premenopausal women. Future longitudinal studies with well detailed endometriosis data in premenopausal people that follows participants well beyond the menopausal transition and past the peak age of endometrial cancer diagnosis may help clarify this association.

Infertility, defined as trying to conceive for ≥12 months of unprotected intercourse without conception, burdens approximately 10–15% of couples in the United States.( 1 ) Several studies have suggested an association between history of infertility and increased risk of endometrial cancer.( 2 – 7 ) However, there are important limitations in the existing literature, and more recent studies have observed mixed results when taking fertility treatment into account.( 8 , 9 ) One of the biggest issues in the existing literature is the lack of statistical consideration of covariates associated with both infertility and endometrial cancer. For example, infertility is strongly associated with reproductive factors, including nulliparity and number of births, and these factors are also risk factors for endometrial cancer, complicating the interpretation of prior analyses. In addition, larger body size, another established risk factor for endometrial cancer, has been associated with certain subtypes of infertility (i.e., polycystic ovary syndrome [PCOS]) which is purported to increase endometrial cancer risk.( 10 , 11 ) Finally, many prior studies have evaluated the association between infertility history, infertility treatment, and endometrial cancer in populations from infertility clinics, which tend to be younger, and have limited follow-up time (<10 years). The median age of endometrial cancer is 64 years old, with over 50% of endometrial cancer cases diagnosed in women over age 60.( 12 ) However, despite the burden of endometrial cancer occurring mostly among postmenopausal women, this population have not been widely studied in relation to infertility.( 2 , 4 – 6 , 8 , 9 ) Therefore, our objective was to investigate the association between history of infertility and risk of postmenopausal endometrial cancer within the Women’s Health Initiative (WHI), a long-standing prospective cohort with over 25 years of follow-up. The rich data available in WHI allowed us to mitigate weaknesses in prior studies including incorporating heterogeneity across infertility diagnoses and accounting for potential confounders (e.g., parity, body mass index [BMI]). In addition, given the age of participants at WHI enrollment (ages 50–79 years) and the enrollment period (1993–1998), we examined the association in an in-vitro fertilization (IVF) naïve population,( 13 ) allowing a clear examination of infertility independent of more invasive infertility treatments. Finally, given that some risk factors may differ between earlier and later onset endometrial cancers, and all infertility diagnoses occurred prior to study enrollment, we examined the association between infertility and endometrial cancer cases diagnosed before study baseline.