Mammographic Breast Density and Risk of Ovarian Cancer in Korean Women

In 2020, there were more than 300,000 incident ovarian cancer cases and more than 200,000 deaths related to ovarian cancer worldwide (1). Ovarian cancer is the eighth most common cause of death in women and the most fatal gynecologic cancer due to late diagnoses (2, 3). Over the last decade, although a decreasing ovarian cancer incidence pattern has been observed in North America, Northern and Southern Europe (4), Asia has witnessed an increasing trend (4). The incidence of ovarian cancer has notably risen in these regions among recent birth cohorts (4). This upward trend in ovarian cancer risk in recent birth cohorts may be attributed not only to changes in socioeconomic and environmental factors but also to the impact of changing lifestyle factors, such as late pregnancy, breastfeeding, smoking, diet, and obesity (4).

The risk factors for ovarian cancer are mainly associated with female hormones, most of which are associated with a risk of breast cancer. Among the risk factors for breast cancer, breast density has been suggested to be a marker of the effect of cumulative endogenous estrogen exposure. It is one of the most significant risk factors, with dense breast tissue associated with a four- to five-fold increase in the risk of developing breast cancer compared with fatty breast tissue (5–9). Considering that circulating and tissue sex hormones may increase the risk of ovarian cancer (10) and their association with dense breast tissue (11), breast density has been suggested to be associated with ovarian cancer (12). However, only one study in the United States (12) investigated the association between high breast density and ovarian cancer risk and identified an association in women aged 50 to 59 years. Considering the progressive increase in the incidence and mortality of ovarian and breast cancers in East Asian countries (4, 13, 14), especially in younger birth cohorts, the association between breast density and ovarian cancer must be explored (4).

To address this, our aim was to investigate the association between mammographic breast density and the risk of developing ovarian cancer in a large-scale cohort of women who have undergone mammography. Given the absence of effective ovarian cancer screening, the potential association between breast density and ovarian cancer risk might provide a new biomarker for ovarian cancer, which would be indispensable in the assessment of ovarian cancer risk.

This cohort study used data from the Korean National Health Insurance Service's (NHIS) National Health Information Database (NHIS-NHID). The NHID was formed by the NHIS in Korea. The NHIS is a mandatory social health insurance which covers all Korean citizens. The NHID is an eligibility database that includes demographic information, income-based insurance contributions, and death registration, healthcare utilization, national health screening, and healthcare provider databases (15). As part of national health screening programs, biennial mammographic breast cancer screening is offered to women aged ≥40 years. This study used data from women who underwent mammography screening between January 1, 2009, and December 31, 2014.

This study was approved by the Institutional Review Board (IRB) of the Hanyang University College of Medicine (approval no. HYUIRB-202106–003–1). On the basis of the approval of the IRB, the NHIS granted permission to use the NHIS-NHID. The requirement for informed consent was waived because all participants had agreed to transfer their screening results to the NHIS-NHID, and the NHIS database was constructed after the anonymization of individual identities.

Of the 8,914,194 women who underwent mammography screening between 2009 and 2014, those who reported a history of cancer before the screening or those with missing information on breast density were excluded. In addition, to avoid inverse causality, we excluded cancer cases or death outcomes recorded within 90 days of the mammography screening. Given that the eligible age for breast cancer screening is 40 years, participants aged <40 years at the time of screening were excluded from the analysis. The final analysis dataset included 8,556,914 women (Fig. 1).

After mammography screening, breast density was reported and classified according to the fourth edition of the Breast Imaging-Reporting and Data System (BI-RADS) density categories (16). BI-RADS category 1 indicates almost entirely fat (parenchyma < 25%); category 2, scattered fibroglandular density (25%–50% parenchyma); category 3, heterogeneously dense (51%–75% parenchyma); and category 4, extremely dense (parenchyma > 75%). In this study, mammography breast density was analyzed on the basis of the four-level ordinal BI-RADS density. If the participants underwent mammography screening more than once during the study period, the results of the first mammograms were used for the analysis.

Ovarian cancer development was ascertained from the healthcare utilization database within the NHID. In Korea, patients with cancer have a special co-payment reduction program, and incident cancer cases are registered in this system, known as the rare incurable disease system. In this study, incident ovarian cancer events were defined as having both the International Classification of Diseases code for ovarian cancer (C56) and a rare incurable disease system code (17). This approach is sufficiently reliable and has been used in previous studies (18). Cancer case studies were conducted until December 2020.

Other risk factors of ovarian cancer were considered as covariates in our analysis. Height and weight were measured at the date of screening and used to calculate the body mass index (BMI, kg/m²). The BMI status was further classified into the following groups according to the Asia-Pacific classification (19): underweight (BMI < 18.5 kg/m²), normal (18.5 kg/m² to <23 kg/m²), overweight (23 kg/m² to <25 kg/m²), and obese (≥25 kg/m²). Furthermore, all participants were required to complete a self-reported questionnaire to assess their health conditions before the screening. We included the following variables as covariates in our analysis: age at screening (continuous variable), BMI (categorical variable; <18.5, 18.5 to < 23, 23 to <25, and ≥25), age at menarche (categorical variable; <15, 15–16, ≥17, and missing), parity (categorical variable; no, one, two or more, and missing), breastfeeding (categorical variable; never, <1 year, ≥1 year, and missing), oral contraceptive use (categorical variable; never, ever, and missing), menopausal status (categorical variable; premenopausal, postmenopausal, and missing), age at menopause (categorical variable; <45, 45–52, ≥53, and missing), hormone replacement therapy for postmenopausal women (categorical variable; never, <5 years, ≥5 years, and missing), smoking status (categorical variable; never, ever, and missing), alcohol consumption (categorical variable; yes, once /week, twice or more/week, and missing), physical activity per week (categorical variable; yes, no, and missing), and family history of cancer in first-degree relatives (categorical variable; yes and no). Physical activity was defined as having moderate (e.g., brisk walking, tennis doubles, cycling at a medium pace) or vigorous (e.g., running, aerobics, fast cycling) physical activity at least 1 day per week. Menopausal status was measured at screening by the following question: “What is your current menopausal status?” with the following response options: still on menstruation, have had a hysterectomy, and have had menopause. Women who reported having menopause were further asked to report their age at menopause. On the basis of this information, menopausal status was classified as premenopausal or postmenopausal status (having had a hysterectomy or having had menopause). For postmenopausal women, the age at menopause and hormone replacement therapy were also adjusted. Missing values of covariates were treated as dummy categories in the analyses.

Descriptive statistics of the baseline characteristics of all participants, the participants who developed ovarian cancer, and those who did not develop ovarian cancer during the follow-up period are presented. Follow-up time was measured in days starting from the screening date until the development of ovarian cancer, the date of death, or December 31, 2020, whichever came first. To assess the trend in the relationship between breast density and the risk of ovarian cancer, the Cochran–Armitage test for trend was used. Cox proportional hazard regression analysis was performed to calculate the HR and 95% confidence intervals (CI) for the association between breast density and ovarian risk, adjusted for age at screening and other covariates. To evaluate whether the association between breast density and ovarian cancer is modified by risk factors, we evaluated the heterogeneity in the associations by the risk factors by the Mantel–Haenszel χ² test and then stratified our analysis by age group, overweight or obesity status, and menopausal status as follows: age at screening (40–49, 50–59, and ≥60 years), BMI status (underweight or normal with BMI < 23 kg/m² and overweight or obese with BMI ≥ 23 kg/m²), and menopausal status (premenopausal and postmenopausal). For the sensitivity analysis, we excluded developed ovarian cancer cases or deaths within 1 year and with 2 years from the screening date and estimated the association between breast density and ovarian cancer. All reported P values were two-sided with a type I error (α < 0.05), and a P value less than 0.05 was considered statistically significant. Statistical analyses were performed using the SAS statistical software (version 9.4, SAS Institute, Cary, NC). The data analyzed in this study were generated and obtained from the Korean NHID. Restrictions apply to the availability of these data, which were used under license for this study. Data are available upon reasonable request to the Korean NHID.

Of the 8,556,914 participants in this study, 9,246 incident ovarian events were recorded after a median follow-up period of 10.0 years (interquartile range, 8.1–11.0 years; Fig. 1). The mean (SD) age of the total population was 53.8 (11.0) years (Table 1). BI-RADS categories 1, 2, 3, and 4 contained 25.1%, 26.3%, 31.6%, and 17.0% of the study population, respectively, and the proportion of BI-RADS 3 and 4 was 50% in women with incident ovarian cancer. Premenopausal women made up 48% of the study population and 50.8% were postmenopausal. The proportion of women with a family history of cancer was 2.3% in those with ovarian cancer and 1.9% in those without ovarian cancer.

The cumulative ovarian cancer incidence was higher in those with dense breasts (BI-RADS 3 or 4) versus those with fatty breasts (BI-RADS 1 or 2; with P value = 0.006; Fig. 2). The risk of ovarian cancer was higher in women with a higher breast density, and the results remained significant even after adjusting for age and other covariates (Table 2; Fig. 3). Figure 3 presents the association between the BI-RADS breast density categories and the development of ovarian cancer in the total population and for each age group after adjusting for other covariates. Compared with women with BI-RADS breast density category 1, women with BI-RADS 2 had a 1.08-fold (95% CI, 1.02–1.15) increased risk of ovarian cancer, and those with BI-RADS 3 and 4 had a 1.16-fold (95% CI, 1.09–1.24) and 1.24-fold (95% CI, 1.15–1.34) increased risk, respectively. A dose–response relationship between the level of breast density and the risk of ovarian cancer was observed, with a P_trend of 0.002 (Table 2). After excluding developed ovarian cancer and deaths within 1 or 2 years of the screening date, the results were similar to the current results (Supplementary Table S1).

The P values were significant for the heterogeneity among age group and menopausal status, as well as for overweight/obesity status, with P < 0.001 and P = 0.005, respectively. Subgroup analysis by age group was consistent regarding the increased ovarian cancer risk in women with a higher level of breast density (P_trend 0.021, 0.003, and <0.001 for age groups 40–49, 50–59, and ≥60 years, respectively; Table 2). An increase in ovarian cancer risk in women with BI-RADS density categories 3 and 4 was found in the 40 to 49 age group, but these results were not statistically significant: adjusted HR 1.04 (95% CI, 0.91–1.19) and 1.12 (95% CI, 0.98–1.29) for BI-RADS 3 and 4, respectively. Statistically significant results were found in women aged 50 to 59 years with BI-RADS density category 4 and in women aged >60 years with BI-RADS density 3 and 4, compared with women in the same age groups with BI-RADS density 1 [adjusted HRs of 1.16 (95% CI, 1.02–1.33), 1.25 (95% CI, 1.12–1.40), and 1.32 (95% CI, 1.06–1.65) in women with BI-RADS 2, 3, and 4, respectively].

Findings from the stratified analysis by BMI and menopausal status (Table 3) yielded results consistent with the main analysis. Overall, there was no difference in ovarian cancer risk across the BMI strata and menopausal status strata. In both BMI groups and pre- or postmenopausal women, those with BI-RADS 2, 3, and 4 had an increased risk of ovarian cancer with adjusted HRs.

Our findings showed that a higher mammographic breast density was associated with an increased risk of ovarian cancer. Compared with women with almost entirely fatty breasts, women with scattered fibroglandular densities, heterogeneous, or extremely dense breast tissues were at an increased risk of ovarian cancer after a 10-year median follow-up period with a dose–response relationship. The increased ovarian cancer risk among women with a higher level of dense breasts was consistent regardless of age, BMI, or menopausal status. To the best of our knowledge, cumulative evidence on shared risk factors in ovarian and breast cancers has not been widely investigated as a risk factor for ovarian cancer, except for the findings from the Breast Cancer Surveillance Consortium (BCSC) and current study (12). The BCSC cohort, which reported an overall increased 5-year ovarian cancer risk by up to 1.20-fold in women with dense breast tissues compared with those with fatty breasts along with a significant dose–response relationship, was consistent with our results (12). While mammographic breast density has been established as a strong risk factor for developing breast cancer (7–9, 20), our findings suggested an added potential relationship between breast density and ovarian cancer.

The subgroup analysis based on age group demonstrated a significant increase in ovarian cancer risk in women with dense breasts in all age groups (40–49, 50–59, and ≥ 60), whereas the results of the study by the BCSC study revealed an elevated risk associated with increased breast density only in women aged 50 to 59 years. In addition, our study demonstrated an increased risk of ovarian cancer in women with denser breasts, irrespective of their menopausal status. Despite the suppression of ovarian function during menopausal transition, primarily in women in their 50s, high breast density was suggested to be a risk factor for ovarian cancer (12). During ages 50 to 59 years, hormone changes associated with menopausal transition are associated with decreased breast density (21). Postmenopausal hormone therapy, which is commonly prescribed to alleviate menopausal symptoms and maintain dense breast tissue despite changes, was suggested as the possible reason for the significant association observed in women aged 50 to 59 years in the BCSC study (12). The difference in the results across different age groups in our study and the BCSC study (12) might be attributed to the varying prevalence of external hormone use during menopausal transition and the differing susceptibility of the Korean and Western women to ovarian cancer risk, as observed in the risk of breast cancer. In the BCSC cohort, 18.6% of women were current users of hormone replacement therapy at the time of mammographic screening (12), whereas 7.4% of this study population had ever used hormone replacement therapy. Another study in Korea identified that 6% to 8% of women aged ≥40 years had ever used hormone replacement therapy, which was much lower than those (30%–65%) reported in the Western population (22). In Korea, breast cancer risk is highest in premenopausal women (aged 40–49 years) and decreases with age (23), which follows a different pattern from that observed in Western women, whose breast cancer incidence increases in postmenopausal women (24). Thus, it might be suggested that the susceptibility of Asian and Western women to changes in circulating sex hormones according to age would be different. Previous results indicating different associations between sex hormone-related factors, including menstrual cycle, hysterectomy, parity-related factors, and ovarian cancer risk according to race would support this interpretation (25, 26).

The risk of ovarian cancer is associated with the number of ovulatory cycles (27). The ovulatory cycles are associated with factors, including the duration of the time not being pregnant and not breastfeeding between menarche and menopause. Thus, parity-related factors, such as the number of children or duration of breastfeeding, could share a similar direction of association with the risk of breast and ovarian cancers (28, 29). Therefore, considering that parity-related factors are associated with breast density (30, 31) and are a marker of the cumulative estrogen effect, these factors might also serve as a marker of ovarian cancer risk (12). Moreover, genetic risk factors, including a family history of cancer (32, 33) and mutations in BRCA1/2 (34) and other low- to intermediate-risk genes, (35) are shared between breast and ovarian cancers, suggesting a shared mechanism in the development of breast and ovarian cancer.

This study had some limitations. First, it used BI-RADS breast density measurements, which were reported by different radiologists at multiple screening centers in Korea, often leading to a moderate inter-observer agreement (36). However, in Korea, a mammography education program to standardize the performance of radiologists is available, which may increase the reproducibility of interpretation (37). Inter-radiologist variability was assessed in randomly selected films from the Korean National Breast Cancer Screening Program, which showed an inter-radiologist variability of 0.83, indicating a very high level of agreement (38). Second, our database did not contain information regarding the clinical characteristics of ovarian cancer, including tumor characteristics and other risk factors, such as age at first birth. Therefore, we were unable to assess the differences in the association between breast density and ovarian cancer according to the clinical characteristics. A study from the Korean Central Cancer Registry reported that approximately half of the ovarian cases are classified as serous adenocarcinoma, followed by mucinous adenocarcinoma with 16% and endometrioid adenocarcinoma and clear cell adenocarcinoma with 9% each (39). Among other countries, Korea has one of the highest proportion of mucinous carcinoma in (40). According to the histology of ovarian cancer, parity-related factors or hormone-related factors demonstrate different associations (41). Because our study did not include information on the histology of ovarian cancer, we could not assess the heterogeneity in the association between breast density and various histologic subtypes. Future studies considering these histology types are warranted to gain a comprehensive understanding. In addition, we did not include information on the family history of ovarian cancer; instead, only information on the family history of cancer in general was used. Furthermore, in the current study, we did not consider the time-varying nature of adjusted covariates due to the limitations of the database, which is another limitation worth mentioning. Future studies with further consideration of changes in the covariates as and breast density in the association with ovarian cancer risk are needed. Another point worth mentioning is that our database only included women aged 40 years or older. Currently, biennial mammographic breast cancer screening is offered to all Korean women aged 40 years or older as part of national health screening programs. The overall screening rate across all age groups is approximately 60%, with slight variation in prevalence across different age group and screening years (42). For instance, in the year 2010, the breast cancer screening rate was 59.4, 66.7, 59.1, and 52.1 in age groups 40–49, 50–59, 60–69, and 70–74, respectively (42). Therefore, the bias resulting from higher adherence to screening in a specific age group, driven by age-based screening recommendations, was minimized in our current database.

The study included more than 8 million women from population-based breast cancer screening, which broadly represents the demographic composition of women in Korea and other East Asian populations. Furthermore, the study had a retrospective cohort design, with many cases of incident ovarian cancer and complete ovarian cancer ascertainment from the health insurance database with cancer codes for reimbursement, covering the whole population. Our results also accounted for confounding factors, including adiposity, age at menarche, age at menopause, menopausal status, parity, postmenopausal hormone therapy, and other health behaviors. Given the progressive increase in the incidence and mortality of ovarian cancer among Asian women (4, 13, 14) due to the changes in lifestyle factors (43, 44), our findings suggested that breast density might be considered for inclusion in the future risk assessment of ovarian cancer.

In conclusion, this study found a modest increase in ovarian cancer risk in women with dense breasts. Findings from this study provide evidence for a potential relationship between breast density and ovarian cancer and may help reinforce existing interventions to prevent or detect ovarian cancer in women with dense breasts who have a higher risk.

No disclosures were reported.

T.X.M. Tran: Conceptualization, data curation, formal analysis, validation, visualization, methodology, writing–original draft, writing–review and editing. S. Kim: Data curation, project administration. B. Park: Conceptualization, supervision, validation, methodology, writing–review and editing.

Boyoung Park received grant from the National Research Foundation of Korea funded by the Korean government (MSIT; grant no. 2021R1A2C1011958).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.