Racial Differences in the Association of Endometriosis and Uterine Leiomyomas With the Risk of Ovarian Cancer

The Ovarian Cancer in Women of African Ancestry (OCWAA) consortium was established with the objective to understand racial differences as they relate to risk factors and outcomes in epithelial ovarian cancer.( 25 ) In this analysis, we included individual-level data from 6 OCWAA studies that collected data on endometriosis and/or leiomyomas (4 case-control studies and 2 case-control studies nested within prospective cohorts). Across the six studies, 8,682 total participants (3,124 Black participants and 5,458 White participants) were included with 1,008 Black and 2,237 White participants with ovarian cancer, and 2,116 Black and 3,221 White controls. Five of the 6 studies collected data on endometriosis (African American Cancer Epidemiology Study [AACES], Black Women’s Health Study [BWHS], Cook County Case Study [CCCS], North Carolina Ovarian Cancer Study [NCOCS], Los Angeles County Ovarian Cancer Study [LACOCS]) and 5 of the 6 studies collected data on leiomyomas (AACES, BWHS, NCOCS, LACOCS, Southern Community Cohort Study [SCCS]). Each study collected demographic, reproductive history, lifestyle, and medical history data via self-administered questionnaires and/or interviews. All OCWAA studies classified race based on self-report. Each study obtained informed consent from its participants; the individual studies and the OCWAA Consortium were approved by the relevant Institutional Review Boards (University of Virginia, Duke University, Boston University School of Medicine, University of Illinois at Chicago, University of Southern California, Vanderbilt University Medical Center, and Meharry Medical College). Data on cancer diagnoses were abstracted from medical records and cancer registry reports. Only epithelial ovarian cancer cases defined using International Classification of Diseases for Oncology, Version 3, were eligible for inclusion. Tumor histotype was determined by combining morphology and grade information according to diagnostic guidelines for ovarian carcinomas in 2014 from the World Health Organization (WHO) Classification of Tumors of Female Reproductive Organs.( 26 ) Each study collected data on the diagnosis of endometriosis and/or leiomyomas via standardized questionnaires that were either interviewer-administered or self-administered ( Appendix 1 , available online at http://links.lww.com/xxx). Heterogeneity was quantified by calculating I 2 statistics and corresponding 95% confidence intervals.( 27 – 29 ) As no significant heterogeneity by study was detected, data was pooled for all analyses.( 29 , 30 ) Logistic regression models were used to calculate odds ratios (ORs) and 95% confidence intervals (CIs) for the associations between endometriosis and leiomyomas with ovarian cancer risk, overall, and by race; polytomous models were used for analyses histotype (high-grade serous, low-grade serous, endometrioid, clear cell, mucinous). Wald joint χ 2 tests of histotype-specific coefficients were used to calculate p-values for heterogeneity between histotypes. We also examined the endometrioid, clear cell, and low-grade serous histotypes combined as these histotypes have previously been the most strongly associated with ovarian cancer risk among women with endometriosis.( 19 ) We defined potential confounders as factors potentially associated with both ovarian cancer and endometriosis and/or leiomyomas risk, including education (high school graduate/GED or less, some college, college graduate, graduate/professional school), parity (no full-term pregnancies, 1–2 full-term pregnancies, ≥3 full-term pregnancies), duration of oral contraceptive use (never, <5 years, ≥5 years), age at menarche, body mass index (BMI; <25.0 kg/m 2 , 25.0–29.9 kg/m 2 , 30.0–34.9 kg/m 2 , ≥35.0 kg/m 2 ), smoking status (ever smoker, never smoker), tubal ligation (yes, no), first-degree family history of breast or ovarian cancer (yes, no), menopausal status, postmenopausal hormone use duration (never, <5 years, ≥5 years), and premenopausal hysterectomy (yes/no) with categories created for missing covariate data. Participants were classified as having a premenopausal hysterectomy if their age at hysterectomy was prior to the onset of menopause; participants were classified as not having a premenopausal hysterectomy if their age at hysterectomy was after the onset of menopause, if the reason for the hysterectomy was due to ovarian cancer, if the hysterectomy was less than a year before diagnosis, or if they never received a hysterectomy. As hysterectomy, oral contraceptive use, and postmenopausal hormone use have the potential to be confounders or mediators of the endometriosis and/or leiomyomas and ovarian cancer relations, we examined their influence in two ways. First, we adjusted for each variable to examine as a potential confounder (described above). Second, we applied a counterfactual method of mediation to assess the presence of mediation on the endometriosis or leiomyomas and ovarian cancer risk relation by each of these variables.( 31 ) This method is much more flexible than the product method, allows for non-linearity, and is equivalent to the product method( 32 ) when there are no exposure-mediator interactions. To fit logistic mediation models to this analysis based on both cohort and case-control studies in this consortia, adjustments were needed to satisfy the rare event assumption. Ovarian cancer rates were rare in the cohort studies, but were not in the case-control studies. To account for this, weights for participants in the case-control studies were calculated by a function of race-specific ovarian cancer prevalence in the general and study populations, and these weights were applied within the models of the mediator as the outcome.( 33 ) To implement this, a SAS macro designed by Valeri and VanderWeele was slightly modified to include these weights in the appropriate regression models.( 31 ) Finally, we assessed effect modification by hysterectomy and postmenopausal hormone use as prior studies have observed differential associations between endometriosis and/or uterine leiomyomas and ovarian cancer risk by these factors.( 24 , 34 , 35 ) The results from effect modification analysis include analyses stratified by the effect modifier of interest (i.e., hysterectomy and postmenopausal hormone use) and the estimation of ORs using a single reference category (e.g., women who did not report a history of endometriosis and who did not report a hysterectomy).( 33 , 36 ) Finally, we performed a sensitivity analysis using a variable that compared non-cancer related hysterectomies (including both pre- and postmenopausal hysterectomies) to no hysterectomy or cancer-related hysterectomies. This information was not available for BWHS or SCCS. All statistical analyses were performed using SAS 9.4 (Cary, NC).

A total of 2,724 Black participants and 5,281 White participants were included in the endometriosis analysis. Among all participants, 6.4% of Black participants and 7.0% of White participants reported a history of endometriosis, with the corresponding values among cases of 81 (8.5%) for Black participants and 198 (9.0%) for White participants, and 92 (5.2%) and 173 (5.6%) for Black and White controls, respectively. Among Black and White participants with ovarian cancer, those reporting a history of endometriosis were more likely to have endometrioid and clear cell tumors compared to those not reporting a history of endometriosis. Both Black and White participants with endometriosis were more likely to report a premenopausal hysterectomy and to have ever used oral contraceptives than participants without endometriosis ( Table 1 ). The mean age of endometriosis diagnosis was 35 years in both Black and White participants. A history of endometriosis was associated with a higher risk of ovarian cancer in both Black participants (OR=2.12; 95% CI=1.50–3.00) and White participants (OR=1.58; 95% CI=1.26–1.98) (p hetereogeneity =0.22). Statistically significant differences in the effect estimates across histotypes were identified when both Black and White participants were combined (p hetereogeneity <0.001) and when associations were examined by race (p hetereogeneity =0.047 for difference across histotypes for Black participants and p hetereogeneity =0.004 for White participants). When stratified by histotype, the association with endometriosis was strongest for the endometrioid histotype among Black participants (OR=7.06; 95% CI=3.86–12.91) compared to White participants (OR=2.17; 95% CI=1.36–3.45) with a significant difference in the effect estimates by race for only the endometrioid histotype (p hetereogeneity =0.003). Black participants (OR=4.82; 95% CI=1.81–21.85) and White participants (OR=3.48; 95% CI=2.24–5.41) had a higher risk of clear cell ovarian cancer (p hetereogeneity =0.64). A similar increased risk of high-grade serous ovarian cancer was observed in both Black and White participants (OR=1.63; 95% CI=1.06–1.64 and OR=1.31; 95% CI=1.00–1.72, respectively, p hetereogeneity =0.56) ( Table 2 and Figure 1 ). Mediation analysis of endometriosis and ovarian cancer by oral contraceptive use showed evidence for a suppressive indirect effect (IE) overall (Percent mediated −3.9%; IE p=0.02) and among White participants (Percent mediated −7.5%; Indirect Effect p=0.01) and a non-significant effect in Black participants (Percent mediated −0.3%; IE p=0.91)( Appendix 2 , available online at http://links.lww.com/xxx). Mediation analyses suggested that part of the association between endometriosis and ovarian cancer is attributable to hysterectomy overall (Percent mediated 4.6%, IE p=0.08), with a larger percent mediated among White participants (8.1%, IE p=0.36) than Black participants (1.7%, IE p=0.55), although none reached statistical significance ( Appendix 3 , available online at http://links.lww.com/xxx). No mediation was observed by postmenopausal hormone use ( Appendix 3 , http://links.lww.com/xxx). When stratified by premenopausal hysterectomy, the association between endometriosis and ovarian cancer in White participants was stronger in those who did not report a premenopausal hysterectomy (OR=1.85, 95% CI=1.41–2.43) than those who did report a hysterectomy (OR=1.09, 95% CI=0.73–1.64) while in Black participants the higher odds of ovarian cancer among those with endometriosis was observed in both those with and without hysterectomy (OR=2.32, 95% CI=1.51–3.59 for no hysterectomy and OR=1.91, 95% CI=1.05–3.48 for hysterectomy). However, these differences were not statistically significant (all p>0.05) ( Table 3 ). When pre- and postmenopausal hysterectomies were examined, the associations were similar (data not shown). A total of 2,995 Black participants and 4,876 White participants were included in the leiomoyoma analysis. Among all participants, 43.2% of Black participants and 21.5% of White participants reported a history of leiomoyomas, with the corresponding values among cases of 444 (46.2%) for Black participants and 488 (24.5%) for White participants, and 851 (41.9%) and 559 (19.4%) for Black and White controls, respectively. Among Black and White participants with ovarian cancer, the distribution of histotypes was similar regardless of race or history of leiomyomas. Both Black and White participants with a history of leiomyomas were more likely to report a hysterectomy than those without a history of leiomyomas. While Black participants were more likely to report tubal ligation than White participants, the percentage reporting tubal ligation within each racial group did not vary materially by history of leiomyomas ( Table 4 ). The mean age of leiomyoma diagnosis was 37 years in Black participants and 40 years in White participants. A history of leiomyomas was associated with a slightly elevated risk of ovarian cancer overall (OR=1.26; 95% CI=1.12–1.41), and in Black (OR=1.34; 95% CI=1.11–1.62) and White participants (OR=1.22; 95% CI=1.05–1.41). Among Black participants, a history of leiomyomas was associated with a higher risk of all histotypes, with statistically significant associations observed for the endometrioid (OR=2.05; 95% CI=1.24–3.39) and high-grade serous (OR=1.32; 95% CI=1.06–1.64) histotypes (p heterogeneity for differences across histotypes=0.002). No statistically significant associations were observed between leiomyomas and specific histotypes in White participants, however, suggestions of higher risk were present in all histotypes except low grade serous and no significant differences in the associations across histotypes were observed (p heterogeneity =0.33) ( Table 5 and Figure 2 ). No significant mediation of the association between leiomyomas and ovarian cancer was observed for hysterectomy, postmenopausal hormone use, or oral contraceptive use (all p>0.05). When stratified by premenopausal hysterectomy, the association between leiomyomas and ovarian cancer was present only among those who did not report a premenopausal hysterectomy for both racial groups. Black participants who had a history of leiomyomas and no premenopausal hysterectomy had 53% greater odds of ovarian cancer (95% CI=1.23–1.89) compared to those without leiomyomas, while there was no association with leiomyomas in the premenopausal hysterectomy group (OR=0.92, 95% CI=0.62–1.36). This interaction was of borderline significance on both the multiplicative (p=0.05) and additive (p=0.08) scales. A similar pattern was observed among White participants (OR=1.31, 95% CI=1.10–1.55 among those with no premenopausal hysterectomy and OR=1.00, 95% CI=0.74–1.35 among those with a premenopausal hysterectomy), but the interaction was not significant on either the multiplicative (p=0.21) or the additive (p=0.28) scales ( Table 6 ). When a common reference group was applied (those who did not have leiomyomas and did not have a hysterectomy), Black and White participants in all other exposure categories had a higher risk of ovarian cancer ( Table 6 ). There were no significant interactions between hormone therapy and leiomyomas in any subset ( Table 6 ). Associations were similar when pre- and postmenopausal hysterectomies were examined (data not shown).

In this consortium of 6 studies, both Black and White participants with a history of endometriosis had a higher risk of ovarian cancer, with the strongest associations among Black participants for the endometrioid and clear cell histotypes. Hysterectomy modified the association between endometriosis and ovarian cancer, but only among White participants. Leiomyomas were associated with a modestly increased risk of ovarian cancer among both Black and White participants. When stratified by hysterectomy, this association persisted only among those without a premenopausal hysterectomy, irrespective of race. Observational studies have consistently found an association between endometriosis and epithelial ovarian cancer, specifically endometrioid and clear cell ovarian cancer histotypes.( 19 – 21 ) However, prior studies were conducted in predominantly White populations and did not compare effect estimates by race. The largest prior study of Black participants to examine this association was a pooled analysis including data from AACES, NCOCS, LACOCS (which are also included in our analyses) and 9 other case-control studies in the Ovarian Cancer Association Consortium, which also found a similar stronger association between endometriosis and ovarian cancer risk among Black participants compared to non-Hispanic White participants, but did not report associations for endometriosis-associated histotypes.( 17 ) Few studies have examined the association between leiomyomas and ovarian cancer risk and to our knowledge none have examined the association among Black women. A Danish registry study found an association between leiomyomas and ovarian cancer risk (OR=1.36; 95% CI=1.16–1.60). However, the association was present only among those who had been diagnosed with leiomyomas within the year preceding the ovarian cancer diagnosis, suggesting that detection bias might explain the observed association.( 37 ) Dixon-Suen et al., also examined leiomyomas and ovarian cancer risk, and reported positive associations that persisted across all histotypes.( 24 ) This is consistent with our study in which elevated effect estimates were present for most histotypes, although not all reached statistical significance. Consistent with results from recent studies, hysterectomy modified the association between endometriosis( 24 , 34 , 38 ) and leiomyomas( 24 ) and ovarian cancer risk. In our analyses, the positive association between both endometriosis and leiomyomas and ovarian cancer risk was strongest among participants who had not had a premenopausal hysterectomy. The exception to this was among Black participants with endometriosis, who had a higher risk of ovarian cancer regardless of hysterectomy status. Hysterectomy eliminates the risk of leiomyoma recurrence, but it may not eliminate risk of recurrent endometriosis. Among those with endometriosis and leiomyomas, hysterectomy may indicate the duration or severity of these conditions, since alternative treatments for both conditions are available. It is possible that removal of the fallopian tubes along with the uterus may explain part of the lower risk of ovarian cancer observed among those with a hysterectomy, but this would be expected to impact primarily high-grade serous ovarian cancers that arise from the fallopian tube. Limitations of this study include potential misclassification of endometriosis and leiomyomas through self-report. For both conditions, the prevalence among controls was lower than current population estimates, which may reflect under-reporting by participants, cohort effects in incidence, and/or period effects in diagnosis. While recall bias is a possibility in the retrospective case-control studies, in the prospective BWHS and SCCS (included in these analyses), we observed similar reporting patterns to that of the case-control studies indicating that recall bias does not likely explain our results. Thus, across our included studies, it is most likely that misclassification of endometriosis and leiomyomas was non-differential with respect to the outcome of ovarian cancer. In addition, the percentage of participants reporting endometriosis and leiomyomas differed across the included studies, with generally higher prevalence reported in more recent studies ( Appendix 1 , http://links.lww.com/xxx), suggesting period effects in diagnosis. In individual studies that included both Black and White participants, Black participants were more likely than White participants to report leiomyomas and White women were more likely than Black participants to report endometriosis, which is consistent with current diagnostic patterns.( 13 , 15 ) Racial differences in endometriosis diagnosis may reflect disparities in access to appropriate diagnostic methods, as opposed to biological differences in incidence.( 14 ) In conclusion, both Black and White participants with endometriosis had a higher risk of ovarian cancer, and hysterectomy modified this association among White participants. Leiomyomas were associated with a modestly increased risk of ovarian cancer among Black and White participants, with hysterectomy modifying the risk in both groups. Further research is needed to understand how racial differences in access to care and treatment options (e.g., hysterectomy), impact or modify ovarian cancer risk.

Endometriosis and uterine leiomyomas are common gynecologic conditions that have significant impact on reproductive age women.( 1 – 12 ) At present, endometriosis is more likely to be diagnosed among White women than Black women,( 13 ) however, racial differences in the diagnosis of endometriosis may reflect disparities in access to care as opposed to biological differences in incidence.( 14 ) Conversely, leiomyomas are two to three times more common in Black women than White women, but reasons for this difference are not well understood.( 15 ) The incidence of ovarian cancer is approximately 30% higher in White women than Black women.( 16 ) Reproductive conditions such as endometriosis and leiomyomas and/or treatments related to these conditions (e.g., hysterectomy, hormone use) could contribute to this difference, but research has been limited in this area.( 17 ) The association between hysterectomy and ovarian cancer risk is conflicting, with studies examining the association prior to 2000 indicating a lower risk of ovarian cancer among those with a hysterectomy, while post 2000 studies suggest a higher risk.( 18 ) For endometriosis, the association with ovarian cancer is well-established with epidemiologic data indicating that women with endometriosis have two- to three-fold greater risk of endometrioid and clear cell ovarian cancer,( 19 – 21 ) but these estimates are from studies comprised predominantly of White women. A consistent association of leiomyomas with ovarian cancer risk has not been demonstrated. However, leiomyomas may indirectly impact ovarian cancer risk, as leiomyomas are the most common indication for hysterectomy in the U.S.,( 22 , 23 ) and hysterectomy has been associated with lower ovarian cancer risk in women with leiomyomas.( 24 ) The objective of the present study was to evaluate associations of endometriosis and uterine leiomyomas with epithelial ovarian cancer risk among Black and White participants, taking into consideration the impact of hysterectomy, oral contraceptive use, and post-menopausal hormone therapy use, on these associations.