Trends of Ovarian Cancer Incidence by Histotype and Race/Ethnicity in the United States 1992–2019

The effect of risk factors on ovarian cancer differs by histotype, and the prevalence of such risk factors varies by race/ethnicity. It is not clear how ovarian cancer incidence has changed over time by histotype and race/ethnicity. We used the Surveillance, Epidemiology, and End Results Program (SEER-12) 1992–2019 data to examine the trend of ovarian cancer incidence for three histotypes (high-grade serous N = 19,691, endometrioid N = 3,212, and clear cell N = 3,275) and four racial/ethnic groups (Asian/Pacific Islander, Hispanic, non-Hispanic Black, and non-Hispanic White). Joinpoint and age-period-cohort analyses were conducted to analyze ovarian cancer incidence trends. High-grade serous cancer was the most common histotype, but its incidence has significantly decreased over time for all racial/ethnic groups; the decrease was largest for non-Hispanic White women (average annual percent change AAPC during 2010–2019 = −6.1; 95% confidence interval (CI), −8.0 to −4.2). Conversely, clear cell cancer was most common in the Asian/Pacific Islanders, and its incidence has increased over time, particularly among Hispanic and Asian/Pacific Islander women (AAPC during 2010–2019 = 2.8; 95% CI, 0.8 to 4.7, and AAPC = 1.5; 95% CI, 0.7 to 2.2, respectively). Endometrioid cancer incidence has decreased in non-Hispanic White but increased in Hispanic women (AAPC during 2010–2019 = −1.3; 95% CI, −1.9 to −0.8, and AAPC = 3.6; 95% CI, 1.0 to 6.3, respectively). The differential incidence trends by histotype and race/ethnicity underscore the need to monitor incidence and risk factor trends across different groups and develop targeted preventive interventions to reduce the burden of ovarian cancer and disparity by race/ethnicity. Significance: During 1992–2019, high-grade serous ovarian cancer incidence has decreased while clear cell cancer incidence has increased regardless of race/ethnicity. Endometrioid cancer incidence has decreased in non-Hispanic White but increased in Hispanic women. Differential ovarian cancer incidence trends highlight the need for targeted preventive interventions by histotype and race/ethnicity.


Introduction
distinct histotypes: high-grade serous, low-grade serous, endometrioid, clear cell, and mucinous, depending on different cells of origin. Many high-grade serous cancers likely arise in the fallopian tube fimbriae, whereas endometriosis is widely considered the tissue of origin for clear cell and endometrioid cancers (17). Many risk factors for these histotypes are shared, including parity and combined oral contraceptive use, but some are distinct by histotype. As mentioned above, ET use is not associated with the risk of clear cell or mucinous histotypes (14). On the basis of data from CI5 among White women in the United States, there have been decreases in endometrioid and mucinous ovarian cancers during 1988-2007. Still, ovarian cancer rates of serous and clear cell have been relatively constant (4), which was not fully understood.
Ovarian cancer incidence is highest among non-Hispanic White women and lowest among non-Hispanic Black and Asian/Pacific Islander women (2).
Wu and colleagues suggested that much of the difference in incidence by race/ethnicity could be explained by the distribution of underlying risk and protective factors for the disease (18).
It is crucial to analyze the trends of ovarian cancer incidence by histotype and race/ethnicity as the risk factors vary by histotype and the prevalence of risk factors differs by race/ethnicity. Some previous studies have examined the trends of ovarian cancer in the United States. However, limitations of these studies include not evaluating the trends by histotypes (19), and restricting to Black and White women (4,(19)(20)(21). In addition, we extended the previous ovarian trend analysis based on the Surveillance, Epidemiology, and End Results (SEER) Program (22), using the recently updated data until 2019.
To understand both historical and current trends in ovarian cancer rates and to overcome the limitations of the previous studies, we have conducted Joinpoint regression and age-period-cohort analyses using the most recent data from SEER during 1992-2019 (22). We evaluated incidence trend across histotypes and racial/ethnic groups (Asian/Pacific Islander, Hispanic, non-Hispanic Black, and non-Hispanic White) to fully understand the state of ovarian cancer risk in the United States.

Materials and Methods
All analyses are based on publicly available data; therefore, Institutional Review Board approval was not required.

Data
Data for ovarian cancer cases and population denominators were obtained from the SEER-12 registry November 2021 submission using SEER*Stat software (version 8.4.0; ref. 23). SEER-12 covers the period from 1992 to 2019. SEER-12 (previously SEER-13) included data from 12 registries (Supplementary Table S1); data from the Detroit registry since 2019 were no longer included in the SEER database. SEER-12 data, rather than SEER-8 (previously SEER-9) or SEER-17 (previously SEER-18), were used because they cover all years during which the Asian/Pacific Islander category was included in the SEER database.
We focused on three main histotypes: high-grade serous, endometrioid, and clear cell, as their etiologies are well understood, they represent the majority of ovarian cancers, and the sample sizes were adequate for our analytic approach.
Although the number of mucinous ovarian cancer cases is comparable with clear cell, we did not include mucinous cancer in the analysis because a majority of mucinous ovarian cancers are considered misclassified gastrointestinal primary tumors (25). Serous grade 2, 3, or 4 and endometrioid grade 3 or 4 were classified as high-grade serous cancer as suggested by the literature (24,26); endometroid grade 1 or 2 were classified as low-grade endometroid cancer. Serous and endometrioid cancers with missing grades (N = 5,313 and N = 582, respectively) were excluded from the main analysis. All clear cell cancers were considered as such regardless of their grade. Our study included 19,691 high-grade serous, 3,212 low-grade endometrioid, and 3,275 clear cell ovarian cancer cases from the SEER-12 registry. The flow chart of inclusion is presented in Supplementary Fig. S1.

Age-standardized Incidence Rates
Age-standardized incidence rates (ASIR) for invasive epithelial ovarian cancer for each histotype (high-grade serous, low-grade endometrioid, and clear cell) were calculated separately for four racial/ethnic groups (Asian/Pacific Islander, Hispanic, non-Hispanic Black, and non-Hispanic White). Other racial/ethnic groups were excluded because of the small sample size ( Supplementary Fig. S1). ASIRs were standardized to the 2000 U.S. population (27).

Joinpoint Trend Analysis
Trend analyses were performed using Joinpoint Regression Program software (version 4.9.1.0; ref. 28). Piecewise-continuous log-linear models by histotype and race/ethnicity were fitted, allowing a maximum of five Joinpoints per model. The final models were selected using permutation tests (29). The method used 4,499 permutation tests to ensure that the probability of overall type 1 error was less than 0.05 (29). Annual percent change (APC) in incidence, average annual percent change (AAPC) for the most recent 5 years (2015-2019) and 10 years (2010-2019), and 95% confidence intervals (CI) were calculated.

Analysis of Incidence by Age-period-cohort
Age-period-cohort analyses were performed by fitting a log-linear model with a Poisson distribution to the observed data to estimate age, period, and cohort effects. To address the well-known nonidentifiability problem of age-periodcohort models, we fitted models with either cohort or period constrained to be zero on average with a slope equal to zero. The final selected model was the one with the lowest Akaike information criterion (AIC). Age-period-cohort analyses were conducted for women ages 30-84 by histotype and race/ethnicity. Age, period, and cohort effects were modeled with natural splines. The birth cohort 1935 and the calendar period 2010 were the reference groups. The analyses were conducted using the Epi package in R (version 4.2.0).

Sensitivity Analyses
In the SEER-12 registry dataset, there were 5,092 patients with serous ovarian cancer with missing grade, but stage data available, among whom 4,725 were at the distant or regional stage. Most distant or regional stage cases are high grade. We, therefore, conducted a sensitivity analysis including these distant or regional stage cases with a missing grade as high-grade serous cases to increase the sample size. In the main analysis above, we included high-grade endometrioid cancer into high-grade serous cancer as the literature suggests that the majority of high-grade endometrioid cancers are high-grade serous cancers. However, we acknowledge that some true high-grade endometrioid cancers exist. Thus, we also conducted a sensitivity analysis in which high-grade endometrioid cancer was not included as high-grade serous cancer.
We also conducted a sensitivity analysis using SEER-17 data (previously SEER-18), which includes more registries but a shorter time frame compared with SEER-12. Sensitivity analyses using the SEER-17 registry included women diagnosed between 2000 and 2019 with the same eligibility criteria described above for SEER-12 analysis. Although the period of SEER-17 is shorter than SEER-12, the number of cases is greater for every histotype and racial/ethnic group because it includes more registries. The Joinpoint analyses using the SEER-17 registry set a maximum of three Joinpoints due to the shorter study period (2000-2019).

Age-period-cohort Analyses
The high-grade serous cancer incidence increased until ages 60-65 years and then decreased with slightly different peaks depending on racial/ethnic groups; for example, the peak for non-Hispanic Black women came about 10 years later compared with non-Hispanic White women ( Supplementary Fig. S2). The incidence of low-grade endometroid and clear cell cancers peaked at ages 45-50 and 55-60, respectively ( Supplementary Fig. S2). This indicates that high-grade serous cancers occur later in life compared with low-grade endometrioid and clear cell cancers.
AICs for the age-period-cohort models were not different from the age-cohort models for most histotype and racial/ethnic categories (except the high-grade serous cancer in non-Hispanic White women; Supplementary  Fig. S3).
Among non-Hispanic White and Asian/Pacific Islander women, there was a decreasing trend by birth cohort for high-grade serous cancer (Fig. 2). The incidence of low-grade endometrioid cancer was decreasing for non-Hispanic White women but increasing in recent birth cohorts among Hispanic women though the CIs were wide. Among non-Hispanic Black women, the trend was relatively constant for high-grade serous and low-grade endometrioid cancers, but the CIs were wide. The wide CIs were possibly due to small sample sizes. In general, the results suggested that clear cell cancer incidence has been increasing by birth cohort. There was a bump in incidence for the 1950-1955 birth cohort for clear cell cancer among all racial/ethnic groups and, more generally, in recent birth cohorts for non-Hispanic White and Black women.
Sensitivity analyses exploring the inclusion of advanced-stage serous cancers missing grade did not materially affect the results presented above. Sensitivity analyses excluding high-grade endometrioid cancer from high-grade serous cancer yielded similar results ( Supplementary Fig. S4). Furthermore, using SEER-17 data, which includes more registries, but a shorter time frame compared with SEER-12, provided similar findings from those using SEER-12.

Aged 30-84 years
High grade serous A.

Discussion
We analyzed trends in ovarian cancer incidence in the United States by histotype and race/ethnicity using SEER-12 registry 1992-2019. The study showed that while high-grade serous cancer is the most common histotype among all racial/ethnic groups, its incidence has been decreasing over time. Conversely, clear cell cancers have been increasing over time and are more common in Asian/Pacific Islander women than in other racial/ethnic groups. Low-grade endometrioid cancer incidence has decreased in non-Hispanic White women but increased in Hispanic women. Our results were similar to a previous analysis of ovarian cancer trends by histotype and race/ethnicity using SEER-18 data during 2000-2013 (30). However, we examined the trends over an extended period and used more updated data (1992-2019) compared with the previous analysis. We also applied Joinpoint and age-period-cohort analyses to examine the trends as well as the period and birth cohort effects on the trends.
The changes in incidence may be due to temporal changes in risk and protective factors for these cancers but may also reflect improvement in histotype classification.
In our analyses, we included high-grade endometrioid cancer into high-grade serous cancer, as suggested by the literature. It is evident that high-grade endometrioid ovarian cancers show different gene profiles from low-grade endometrioid tumors but similar gene profiles as high-grade serous tumors (31).
In addition, high-grade serous carcinoma may be misclassified as high-grade endometrioid cancer. Gilks and colleagues reviewed 176 tumors originally classified as endometrioid invasive epithelial ovarian cancer and reclassified 50 cases as serous tumors, of which 49 were high-grade serous tumors (26). The authors also looked at the expression of WT-1, which is supposed to be predominantly positive among serous tumors but mostly negative among endometrioid tumors. WT-1 immunostaining supported the reclassification because 68.2% of those 50 tumors were WT-1 positive, which is similar to the proportion of WT-1 positivity in the group of serous carcinomas (77.8%) but is remarkably higher than the proportion of WT-1 positivity in the endometrioid tumors classified by this review (3.9%; refs. 26). These references suggest that there are etiologic similarities between high-grade endometrioid and high-grade serous tumors, and thus it is appropriate to group them. We conducted sensitivity analyses excluding high-grade endometrioid tumors from high-grade serous tumors and found similar results.
The underlying reasons for the decrease in the incidence of advanced-stage, high-grade serous cancer in all racial/ethnic groups over the past 20 years are likely multifaceted. First, estrogen-alone menopausal hormonal therapy (ET) use is associated with an increased risk of high-grade serous cancer (14,15). In 2002, the prevalence of ET use decreased dramatically following the WHI results linking menopausal estrogen plus progestin (EPT) use with increased risk of cardiovascular heart disease, stroke, and invasive breast cancer (16). Although the WHI result was specific to EPT, both ET and EPT use declined sharply in the United States (32) and has continued to decrease at least through 2013 (33). Use of ET is not associated with the risk of clear cell ovarian cancer (14), consistent with our observation that the incidence of this histotype is not decreasing but rather increasing.
Another driver for the decreasing trend in high-grade serous cancer incidence could be the increase in opportunistic salpingectomy, which is the removal of the fallopian tubes during abdominal surgery or in lieu of tubal ligation. Highgrade serous cancer is thought to arise from the fallopian tube fimbriae (17); thus, salpingectomy is recommended as a preventive strategy for high-grade serous cancer (34)(35)(36)(37). The number of salpingectomies in the United States increased in 2008 and during the 2011-2013 period went up by 71% (38). This coincided with the sharp decline in the number of high-grade serous cases from 2011 to 2018, especially from 2014 to 2018 (24%). However, neither the decrease in ET use nor the increased uptake of opportunistic salpingectomy would fully explain the declining high-grade serous incidence we observed in the late 1990s.
We also found an increasing trend in clear cell cancer incidence for all racial/ethnic groups, especially among Asian/Pacific Islanders. This may also be explained by improvement in the diagnosis of this histotype. There was a pattern of shifting away from classifying tumors as "carcinoma, not otherwise specified" (NOS) to clear cell carcinoma in all racial/ethnic groups, but most striking in Asian/Pacific Islanders. Using SEER-12 registry data, we found that during 1992-2019, the proportion of NOS among Asian/Pacific Islanders decreased from 20% to 11%, while the proportion of clear cell cancer increased by the same amount, from 9% to 18% (Supplementary Fig. S5 . Therefore, the impact of the decline in parity and combined oral contraceptive use on high-grade serous incidence would likely be less than one on clear cell cancer. The observation that the endometrioid cancer incidence decreased in non-Hispanic White but increased in Hispanic women in recent years may be partially explained by the racial/ethnic differences in the decline in parity and combined oral contraceptive use, given the strength of these associations (39, 40). Also, the increase in clear cell rates may be due to the changes in parity and combined oral contraceptive use.
Two additional interesting observations for birth cohort effects are the small but noticeable peak around birth cohorts 1950-1955 for clear cell cancer in all racial/ethnic groups and that the peak for high-grade serous cancer incidence is about 10 years later in non-Hispanic Black women compared with non-Hispanic White women. The peak in clear cell cancer incidence might partially be explained by the exposure to diethylstilbestrol (DES), a potent synthetic estrogen, which was given to pregnant women in the United States from the early 1940s through the 1960s for the prevention of spontaneous abortion and premature delivery (41). Use of DES in pregnancy was discontinued in 1971 when there was evidence of a strong association between prenatal exposure to DES and an elevated risk of clear cell adenocarcinoma of the vagina and cervix (42). There is some evidence that an increased risk of clear cell ovarian cancer may be associated with prenatal DES exposure (43), although population studies did not find an association, possibly due to small sample sizes (44)(45)(46). The 10-year later peak in high-grade serous incidence for non-Hispanic Black women is likely due to a large number of hysterectomies (likely with an oophorectomy) that occurred in this population (47). Generally, it is likely that the population at risk denominator for calculations for ovarian cancer incidence for non-Hispanic Black women is inflated because of the high prevalence of hysterectomy with oophorectomy.
There are some limitations of the study. First, the ecologic nature of the study does not allow for causal inferences. The SEER data do not contain information on risk factors for ovarian cancer. Second, U.S.-born and foreign-born women may have different cohort effects, but we could not examine the difference because the SEER registry removed the country of birth variable in recent submissions due to bias (personal communication with the SEER support team). Third, there might be a risk of misclassifying ovarian cancer histotypes across the registries. To minimize this risk, we conducted a sensitivity analysis including patients with advanced-stage serous cancer (distant or regional stage) with a missing grade as high-grade serous cases. Despite these limitations, the SEER data used in our study are the gold standard for cancer registration (48).
In conclusion, this study provides the most current overview of ovarian cancer trends by histotype and race/ethnicity in the United States and examines the influence of period and birth cohort on the trends. We found that high-grade serous cancer incidence has decreased overall, while clear cell cancer incidence has increased. The differences in the trends by histotype and race/ethnicity can partially be explained by the variations in the prevalence of risk factors by year and cohort across race/ethnicity and their different effect magnitudes by histotype, as well as the changes in diagnostic classification. It is crucial to develop targeted preventive interventions to effectively reduce the burden of ovarian cancer and disparity by race/ethnicity. There is a need to continue monitoring the incidence trends of ovarian cancer, particularly clear cell cancer to understand whether the increase in incidence was due to changes in risk factor prevalence or the shift in histotype classification. In addition, it is important to keep monitoring the incidence trends of ovarian cancer by race/ethnicity to develop targeted prevention strategies.