Cancer-Specific Mortality in Asian American Women Diagnosed with Gynecologic Cancer: A Nationwide Population-Based Analysis

This article is featured in Highlights of This Issue, p. 493

Although Asian Americans (AA) represent only 7% of the U.S. population, they are the fastest-growing racial group in the U.S. It is estimated that the AA population grew by 72% between 2000 and 2015, from 11.9 million to 20.4 million (1). AAs are projected to become the largest immigrant group in the U.S., surpassing Hispanics by 2055 (2). Despite the rapidly growing population, few studies have comprehensively assessed cancer outcomes in AAs.

The dominant cancer literature tends to aggregate AAs, obscuring their diversity and heterogeneous backgrounds. AAs include individuals who trace their ancestry to East Asia, Southeast Asia, or South Asia, comprising more than 20 countries and over 200 languages. Although, socioeconomic indicators are more favorable in AAs than other U.S. ethnic groups (3, 4), significant differences exist within AA ethnic groups (Table 1).

About 56% of AAs hold a bachelor's degree or more, with higher proportions among Indians (76%), Pakistanis (60%), Koreans (59%), and Chinese (57%), and lower proportions among Filipinos (50%) and Vietnamese (32%). An estimated 66% of AAs are foreign-born, with 30% immigrating to the U.S. in 2010 or later. A higher proportion of Indians (71%), Chinese (69%), and Koreans (69%) are foreign-born, whereas a lower proportion of Japanese immigrants are foreign-born (41%; ref. 5). About 7% of AAs are uninsured, with higher proportions among Koreans (10%), Pakistani (8%), and Vietnamese (8%) and lower proportions among Indians (5%) and Japanese (3%). In addition, 31% of AAs have limited English proficiency varying from 18% in Indians to 48% in Vietnamese (Table 1; ref. 5). All of these factors play important roles in an individual's ability to access and navigate the healthcare system and can significantly impact cancer risk and outcomes.

Given the heterogeneous backgrounds of AAs, a nuanced understanding of differences in cancer-specific mortality by AA ethnicity is critical to identifying opportunities for cancer prevention, screening, and treatment. To fill this knowledge gap, we examined the extent of differences in gynecologic cancer mortality for six of the largest AA ethnicities—Chinese, Indians/Pakistanis, Filipino, Vietnamese, Koreans, and Japanese—compared with non-Hispanic Whites (NHW) from 1991 to 2016. We examined differences in gynecologic cancer mortality because disparities in these cancers are understudied relative to their incidence, particularly among AAs.

In 2021, gynecologic cancers will account for 13% of all new cancers diagnosed (116,760 cases) and 12% of all cancer-related deaths (34,080 deaths) in the U.S. (6). Gynecologic cancers are a leading cause of person–years of life lost and contribute substantially to lost earnings (7). The burden of gynecologic cancer is expected to increase as populations grow and age. Our analysis focused on uterine cancer, the most common gynecologic cancer and the fourth most common cancer diagnosed among U.S. women; ovarian cancer, the most lethal gynecologic cancer and the fifth leading cause of cancer-related deaths among U.S. women; and cervical cancer, the only gynecologic cancer with an effective screening test (6). Collectively, these cancers account for 88% of all new gynecologic cancer diagnoses and 91% of all gynecologic cancer deaths among U.S. women.

We used the Surveillance, Epidemiology, and End Results (SEER) program, which aggregates data from population-based cancer registries throughout the U.S. and includes information on demographics, cancer characteristics, cancer treatment, and mortality. SEER captures 35% of the U.S. population, including 32% of Whites and 58% of AAs (8). The study was deemed exempt by the Institutional Review Board of Columbia University since all data used were publicly available and deidentified.

All ovarian, uterine, and cervical cancer cases in women ≥18 years between 1991 and 2016 were extracted from SEER-18 (n = 431,813). The year 1991 was used as the first year of study because SEER introduced ethnic subdivisions among AAs in 1991. Our analysis was limited to the largest AA ethnic groups, which include Filipino, Chinese, Japanese, Vietnamese, Korean, and Indian/Pakistani. These ethnic groups account for 85% of the total AA population. Indian and Pakistani ethnicities were combined because SEER did not separate these groups. Black, American Indian/Alaska Native, other AA, Pacific Islander, and Hispanic women of any race were excluded (n = 109,445). Women with more than one cancer (n = 42,415), those diagnosed at autopsy or from death certificates (n = 2,244), and those with no follow-up (n = 9,796) were also excluded. After exclusions, the study population included 267,913 NHW and non-Hispanic AA women with an ovarian, uterine, or cervical cancer diagnosis.

Demographic characteristics, including year of diagnosis (1991–2016), age at diagnosis (continuous and <50 years, ≥50 years), marital status (married, unmarried, unknown), and geographic region (Northeast, Midwest, South, West) were ascertained for all cases. Geographic region was based on U.S. Census delineations and classified on the basis of the location of the SEER registry. County-level measures of median household income, proportion with at least a college education, proportion urban, proportion foreign-born, proportion of families with limited English proficiency, and proportion of families living below 150% of the poverty line were obtained on the basis of linkage to the 2000 U.S. Census provided through SEER. Clinical and treatment characteristics, including cancer stage, grade, histologic subtype, and first course of cancer treatment, were also extracted.

Histologic subtype of cases was based on the International Classification of Diseases for Oncology, third edition (ICD-O3) morphology codes. Epithelial ovarian cancer was classified as high-grade serous, low-grade serous, endometrioid, mucinous, clear cell, carcinosarcoma, malignant Brenner, mixed, and carcinoma, not otherwise specified and nonepithelial ovarian cancer was classified as sex-cord stromal, germ cell, and other (9). Uterine cancer was classified as low-grade and high-grade endometrioid, serous, carcinosarcoma, clear cell, mixed, uterine sarcoma, and other (10), and cervical cancer was classified as squamous cell, adenocarcinoma, adenosquamous, and other.

Cancer treatment was classified as a binary variable (yes, no) according to the type of cancer-directed treatment and stage. For ovarian cancer, stage-appropriate treatment was defined as surgery or surgery and chemotherapy for localized cases and surgery and chemotherapy for regional and distant cases (11). For uterine cancer, stage-appropriate treatment was defined as surgery with or without radiation for localized cases, surgery and chemotherapy or chemoradiation with or without surgery for regional cases, and surgery and chemoradiation or chemotherapy with or without radiation for distant cases (12). For cervical cancer, stage-appropriate treatment was defined as surgery with or without radiation or radiation alone for localized cases, and chemotherapy with or without radiation for regional and distant cases (13).

Demographic and tumor characteristics by race and ethnicity were compared using t test and χ² test. To examine the risk of cancer-specific mortality, we used a competing risk approach using Fine and Gray regression models with noncancer death as a competing event. Subdistribution HRs for cancer mortality were estimated, which can be interpreted as having a direct relationship to the probability of cancer mortality (14, 15). Cancer mortality was defined as the number of women who died with ovarian, uterine, or cervical cancer as the underlying cause of death in a sample of women diagnosed with these cancers (i.e., our study population). Time since cancer diagnosis was used as the time metric and was measured as the number of months between the date of cancer diagnosis and date of death or date of last known contact for those alive.

Cancer mortality was examined by race and ethnicity and age at diagnosis (<50 years and ≥50 years). This age cut point was selected to determine whether associations differed by menopausal status as cervical cancers are most prevalent in premenopausal women and uterine and ovarian cancers present in the perimenopausal and postmenopausal periods. Models were adjusted for age, year of diagnosis, marital status, median household income, proportion with at least a college education, proportion urban, stage, grade, histologic subtype, and cancer treatment. These covariates were chosen a priori on the basis of clinical rationale and prior literature (16–21). Cases with missing stage, grade, and/or treatment were included in the primary analyses. Sensitivity analyses excluding cases with missing data (n = 59,357), stratified analyses by year of diagnosis (1991–2006 and 2007–2016), and analyses adjusting for neighborhood-level socioeconomic status (22) for a subset of cases diagnosed between 2000 and 2016 (n = 216,755) were performed. Cut points for year of diagnosis were based on median values. The proportional hazards assumption was assessed graphically and with Schoenfeld residuals and there was no indication that the assumption was violated. All analyses were performed in Stata version 15.1 (College Station, TX).

From 1991 to 2016, we identified 69,113 (25.8%), 157,340 (58.7%), and 41,460 (15.5%) women with ovarian, uterine, and cervical cancer, respectively (Table 2). The average age at diagnosis was lower in AAs (range, 55.2–59.8 years) than NHWs (61.1 years), and a higher proportion of AAs (55.1%–68.3%) were married at the time of diagnosis than NHWs (51.8%). AAs were more likely to reside in counties with higher household income, counties with a higher proportion of individuals with at least a college education, counties with a higher proportion of foreign-born populations, counties with a higher proportion of families with limited English proficiency, and urban counties than NHWs. In addition, a higher proportion of cases in AAs (55.1%–96.7%) were from SEER registries in the West than NHWs (48.0%). About 29.9% of cases in Indian/Pakistani women were from SEER registries in the Northeast compared with 19.1% of cases in NHW women.

The average age at ovarian cancer diagnosis was lower in AAs (53.5–61.2 years) than NHWs (62.5 years; Table 3). Most ovarian cancer cases were diagnosed at the distant stage (57.0%), with a lower proportion of cases in AAs (45.6%–52.0%) diagnosed at the distant stage than NHWs (57.8%). The most common histologic subtype was high-grade serous ovarian cancer (45.1%). AAs were more likely to be diagnosed with clear cell, endometrioid, and mucinous carcinomas than NHWs, with considerable variation by ethnicity. Overall, 72.6% of ovarian cancer cases received stage-appropriate treatment with higher or comparable rates in AAs than NHWs. Like ovarian cancer, the average age at uterine cancer diagnosis was lower in AAs (56.1–61.2 years) than NHWs (63.4 years). Overall, most uterine cancer cases were diagnosed at the localized stage (71.0%), with AAs (62.5%–70.3%) less likely to be diagnosed with uterine cancer at the localized stage than NHWs (71.3%). The most common histologic subtype was low-grade endometrioid (58.2%). AAs were more likely to be diagnosed with high-grade endometrioid and serous carcinomas, and uterine sarcoma than NHWs, with considerable variation by ethnicity. Overall, 84.6% of uterine cancer cases received stage-appropriate treatment with higher or comparable rates in AAs than NHWs. Unlike ovarian and uterine cancer, the average age at cervical cancer diagnosis was higher in AAs (52.7–55.8 years) than NHWs (49.6 years). Overall, most cervical cancer cases were diagnosed at the localized stage (49.9%) and at grade II/III (55.5%). AAs (38.2%–50.3%) were less likely to be diagnosed with cervical cancer at the localized stage than NHWs (50.5%). The most common cervical cancer histologic subtype was squamous cell carcinoma (65.1%). AAs were more likely to be diagnosed with squamous cell carcinoma than NHWs, with considerable variation by ethnicity. Overall, 71.6% of cervical cancer cases received stage-appropriate treatment with higher or comparable rates in AAs than NHWs.

Overall, AAs had a significantly lower risk of ovarian cancer death than NHWs [HR, 0.90; 95% confidence interval (CI), 0.86–0.94]. There was a significant interaction between race and ethnicity and age at diagnosis in AAs compared with NHWs (P = 0.005), with a lower risk of ovarian cancer death in AAs ≥50 years (HR, 0.87; 95% CI, 0.82–0.92). Filipino (HR, 0.87; 95% CI, 0.80–0.94), Chinese (HR, 0.89; 95% CI, 0.81–0.98), and Indian/Pakistani women (HR, 0.84; 95% CI, 0.73–0.97) had a significantly lower risk of ovarian cancer death than NHW women. There was a borderline significant interaction between race and ethnicity and age at diagnosis in Filipino (P = 0.08) and Korean women (P = 0.09) compared with NHW women, with a lower risk of ovarian cancer death in Filipino women ≥50 years (HR, 0.85; 95% CI, 0.77–0.94) and higher risk of ovarian cancer death in Korean women <50 years (HR, 1.23; 95% CI, 0.97–1.56; Fig. 1).

There was no difference in the overall risk of uterine cancer death by race and ethnicity (HR, 1.03; 95% CI, 0.97–1.10). However, a significant interaction between race and ethnicity and age at diagnosis was noted in AAs compared with NHWs (P < 0.001), with a higher risk of uterine cancer death in AAs <50 years (HR, 1.26, 95% CI, 1.08–1.46). Chinese (HR, 1.13, 95% CI, 1.00–1.27) women had a higher risk of uterine cancer death than NHW women, whereas Vietnamese women (HR, 0.78; 95% CI, 0.61–0.99) had a lower risk of uterine cancer death than NHW women. There was a significant interaction between race and ethnicity and age at diagnosis in Filipino women (P = 0.008) compared with NHW women, with a higher risk of uterine cancer death in Filipino women <50 years (HR, 1.41; 95% CI, 1.11–1.79; Fig. 2).

Overall, AAs had a significantly lower risk of cervical cancer death than NHWs (HR, 0.80; 95% CI, 0.75–0.87). There was a significant interaction between race and ethnicity and age at cervical cancer diagnosis in AAs compared with NHWs (P < 0.001), with a lower risk of cervical cancer death in AAs ≥50 years (HR, 0.74; 95% CI, 0.61–0.83). Except Japanese women, all AAs (HRs, 0.67–0.84) had a significantly lower risk of cervical cancer death than NHWs. There was a significant interaction between race and ethnicity and age at cervical cancer diagnosis in Filipino (P = 0.001) and Vietnamese (P = 0.03) women compared with NHW women, with a lower risk of cervical cancer death in women ≥50 years (Filipino: HR, 0.71; 95% CI, 0.61–0.83; Vietnamese: HR, 0.63; 95% CI, 0.51–0.78; Fig. 3).

Overall results were unchanged after the exclusion of cases with missing stage, grade, and/or treatment data (Supplementary Table S1), after stratifying the overall analysis by year of cancer diagnosis (Supplementary Table S2), and after adjusting for neighborhood-level socioeconomic status for a subset of cases diagnosed between 2000 and 2016 (Supplementary Table S3).

In this large population-based sample of 267,913 women with gynecologic cancers, we found that AAs as an aggregated group had a lower risk of ovarian and cervical cancer death and a higher risk of uterine cancer death than NHWs after adjusting for demographic factors, tumor characteristics, and treatment. Associations were stronger in women ≥50 years at diagnosis for ovarian and cervical cancer, whereas associations for uterine cancer were stronger in women <50 years at diagnosis. Patterns of cancer mortality were heterogeneous across AA ethnicity, highlighting the importance of disaggregation to better understand risk and protective factors unique to each ethnic subgroup and identify high-risk groups.

Filipino, Chinese, Vietnamese, and Indian/Pakistani women had a lower risk of ovarian cancer death than NHW women. However, Korean women <50 years had a higher risk of ovarian cancer death than NHW women. Among those with uterine cancer, Chinese women had a higher risk of death than NHW women, whereas Vietnamese women had a lower risk of death than NHW women. Filipino women <50 years had a higher risk of uterine cancer death than NHW women. All AAs, except Japanese women, had a lower risk of cervical cancer death than NHWs. In Filipino and Vietnamese women, the lower risk of cervical cancer death was limited to those ≥50 years.

The reasons for these mortality differences are likely multifactorial. For ovarian cancer, AAs were younger at diagnosis and more likely to be diagnosed with localized disease and subtypes associated with favorable prognoses, such as mucinous, clear cell, and endometrioid carcinomas than NHWs. Prior studies have shown that among women diagnosed with ovarian cancer, factors such as younger age, nonserous histologic subtype, and early-stage disease are associated with a favorable prognosis (23, 24). In addition, endometriosis is a strong risk factor for clear cell and endometrioid carcinomas (25) and studies have shown that AAs have a higher prevalence of endometriosis than NHWs, particularly Filipino, Japanese, Korean, and Indian women (26, 27). In our study, Filipino, Chinese, and Japanese women were more likely to be diagnosed with clear cell and endometrioid carcinomas. Emerging research has also shown that Chinese and Korean women have a higher genetic predisposition to BRCA1/2 mutations than Filipino and Vietnamese women (28–31). Women with BRCA1/2 mutated ovarian cancer have an improved prognosis, higher response rates to platinum-based chemotherapy, and longer treatment-free intervals than women without BRCA1/2 mutated ovarian cancer (32, 33).

For uterine cancer, although AAs were younger at diagnosis, they were less likely to be diagnosed with localized disease and more likely to be diagnosed with aggressive subtypes associated with poor prognoses, such as high-grade endometrioid and serous carcinomas than NHWs. Among AA subgroups, Filipino, Vietnamese, Korean, and Chinese women were more likely to be diagnosed with high-grade endometrioid carcinoma and Indian/Pakistani women were more likely to be diagnosed with serous carcinoma. Endometrioid carcinomas are associated with unopposed estrogen stimulation and more strongly associated with obesity (34). They are often diagnosed at an early stage and have a favorable prognosis. In contrast, nonendometrioid carcinomas are estrogen-independent and less strongly associated with obesity (35, 36). They are often diagnosed in non-White women and have an unfavorable prognosis (34). In addition, differences in recognition of the importance of abnormal uterine bleeding, which occurs in 90% of women diagnosed with uterine cancer (37), could potentially account for the late-stage diagnosis of uterine cancer in AAs. Studies have also shown that among women diagnosed with uterine cancer, those with preexisting diabetes have a higher risk of uterine cancer mortality than those without diabetes (38). The prevalence of diabetes in all AA ethnic subgroups is higher than NHWs, particularly in Filipino, Japanese, and Indian women (39–41).

For cervical cancer, AAs were older at diagnosis and were less likely to be diagnosed with localized disease than NHWs; however, they were more likely to be diagnosed with squamous cell carcinoma than NHWs, which is associated with a favorable prognosis than other cervical cancer subtypes (42, 43). Among AA subgroups, Chinese, Vietnamese, Korean, and Indian/Pakistani women were more likely to be diagnosed with squamous cell carcinoma. In contrast, Filipino women were more likely to be diagnosed with adenocarcinoma, and Japanese women were more likely to be diagnosed with adenosquamous carcinoma. The reasons for these differences in histologic subtypes by AA ethnicity are unclear. In addition, older age at diagnosis and less localized disease in AAs suggests lower access to screening services. Prior studies have shown that AAs are less likely to have had a Pap test than NHWs, with lower screening rates in Korean and Vietnamese women than Filipino, Chinese, and Japanese women (44, 45).

Furthermore, cancer treatment, lifestyle factors, time since immigration, and mutational landscapes of tumors may also impact differences in mortality and account for the interethnic heterogeneity noted in this study. We found that most AA subgroups had comparable or higher rates of stage-appropriate cancer treatment than NHWs, which may partly explain their comparable or lower cancer mortality. Minimal variation in stage-appropriate treatment within AA subgroups was noted for localized cases, with greater variation for regional and distant cases. These differences may be driven by health insurance status and access to cancer care. In addition, in the Women's Health Initiative studies, postmenopausal women with the healthiest lifestyle index had a 20% lower risk of cancer-specific mortality than those with the least healthy lifestyle index, with the strongest associations among Asian women (46). Prior studies have shown that the prevalence of healthy lifestyle behaviors varies significantly by Asian ethnicity. For example, smoking prevalence varies from 8.2% in Chinese women to 30.5% in Vietnamese women, the prevalence of vigorous physical activity varies from 14.2% in Chinese women to 24.7% in Filipino women, and the prevalence of alcohol use varies from 12.3% in Japanese women to 35.5% in Korean women (47). Moreover, groups with longer immigration histories, such as Japanese and Filipinos, may have acculturated to Western diets and lifestyles and thus have similar gynecologic cancer risk profiles to NHWs than groups with more recent immigration histories, such as Vietnamese and Koreans (48). Studies have also shown that AAs have fewer polymorphisms than NHWs in drug-metabolizing enzymes such as CYP3A4 and CYP3A5, which can lead to improved sensitivity to several anticancer drugs (49).

Disparities in gynecologic cancer mortality between non-Hispanic Blacks and NHWs have been reported extensively in the literature; (50, 51) however, few studies have focused on AAs. Our results for ovarian cancer agree with the few studies that have examined differences between AAs and other racial and ethnic groups and reported favorable outcomes among AAs. We further noted heterogeneity in ovarian cancer mortality by AA ethnicity and age at diagnosis. Like prior studies (16, 17, 52), we found that AAs were more likely to be diagnosed with localized disease and with mucinous, endometrioid, and clear cell carcinomas than NHWs. To our knowledge, our study is one of the first to evaluate differences in uterine and cervical cancer mortality between AAs and NHWs using detailed demographic, clinical, and treatment information and disaggregated by ethnicity.

Future studies incorporating genetic ancestry and mutational landscapes of ovarian, uterine, and cervical cancers among AAs may provide additional insights into the interethnic differences observed in this study. In addition, studies incorporating multiple environmental, sociodemographic, clinical, and behavioral factors are needed to identify potential targets to lessen the burden of gynecologic cancer in the growing AA population. These data are not available in SEER registries; therefore, studies leveraging data from electronic health records from healthcare systems that have a high representation of AA populations and linkage to cancer registries are needed to address current knowledge gaps.

Our study is subject to some limitations. First, there may be misclassification of race and ethnicity in cancer registries; however, the extent of misclassification of AA race and ethnicity in prior studies suggests that misclassification is minimal (sensitivity 70%–90%; refs. 53, 54). Second, SEER treatment data have been found to be incomplete with varying sensitivity by cancer site, stage, and patient characteristics (55). For ovarian cancer, the sensitivity of SEER data to identify women who received chemotherapy is 84%. Third, we lacked genetic information, information on diagnosed comorbidities and body mass index, and insurance status, which could have influenced cancer treatment decisions and outcomes. Fourth, we were unable to account for individual-level socioeconomic status, which could have influenced access to cancer treatment. Finally, we did not capture nativity status, which may differ by AA ethnicity and influence cancer risk (56). Data on nativity status in SEER have been found to be incomplete and inaccurate (57, 58). Despite these limitations, we were able to leverage a large, population-based dataset to comprehensively examine differences in gynecologic cancer mortality by AA ethnicity. SEER captures approximately 35% of the U.S. population, including 32% of NHWs and 58% of AAs, with >90% case completeness and accuracy. Our analysis also accounted for multiple important confounders, including demographic, clinical, and treatment factors, which are known to impact cancer mortality.

In conclusion, AAs have a lower risk of ovarian and cervical cancer mortality and a higher risk of uterine cancer mortality than NHWs after adjusting for demographic factors, tumor characteristics, and cancer treatment. Disaggregation of AA ethnicity revealed a more nuanced picture, with heterogeneity in associations for ovarian and uterine cancer mortality among Chinese, Korean, Filipino, and Vietnamese women. More research is needed to improve the understanding of factors contributing to these mortality differences.

J.D. Wright reports grants from Merck; and personal fees from UpToDate outside the submitted work. J.M. Genkinger reports grants from NCI T32 CA094061 during the conduct of the study. No disclosures were reported by the other authors.

P.S. Karia: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. P. Tehranifar: Supervision, methodology, writing–review and editing. K. Visvanathan: Writing–review and editing. J.D. Wright: Supervision, methodology, writing–review and editing. J.M. Genkinger: Resources, supervision, validation, methodology, writing–review and editing.

Supported by the NCI (T32 CA094061) to Drs. Genkinger (mentor) and Karia (trainee).The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.