|Home | About | Journals | Submit | Contact Us | Français|
We investigated heterogeneity in ethnic composition and immigrant status among US Asians as an explanation for disparities in breast cancer survival.
We enhanced data from the California Cancer Registry and the Surveillance, Epidemiology, and End Results program through linkage and imputation to examine the effect of immigrant status, neighborhood socioeconomic status, and ethnic enclave on mortality among Chinese, Japanese, Filipino, Korean, South Asian, and Vietnamese women diagnosed with breast cancer from 1988 to 2005 and followed through 2007.
US-born women had similar mortality rates in all Asian ethnic groups except the Vietnamese, who had lower mortality risk (hazard ratio [HR]=0.3; 95% confidence interval [CI]=0.1, 0.9). Except for Japanese women, all foreign-born women had higher mortality than did US-born Japanese, the reference group. HRs ranged from 1.4 (95% CI=1.2, 1.7) among Koreans to 1.8 (95% CI=1.5, 2.2) among South Asians and Vietnamese. Little of this variation was explained by differences in disease characteristics.
Survival after breast cancer is poorer among foreign- than US-born Asians. Research on underlying factors is needed, along with increased awareness and targeted cancer control.
Breast cancer incidence rates have been shown to differ markedly across distinct ethnic subpopulations of US Asian women,1–3 and studies of outcomes after breast cancer similarly have detected wide variation in stage at diagnosis and survival patterns across groups.4–9 For example, in a population-based study, we reported that 5-year survival probabilities and proportions of late-stage disease were as poor in Filipinas as in African American women and that Japanese women had more favorable stage and survival distributions than did non-Hispanic White women.7 Identifying and explaining such disparities are important for public health efforts to diminish disparities in outcomes after breast cancer. However, explanations for the ethnic differences in breast cancer survival in Asians have not been carefully studied aside from the effects of variations in treatment, sociodemographic factors, and tumor characteristics.6
Asian populations in the United States are heterogeneous with respect to immigration and acculturation, 2 factors likely to play a substantial role in explaining different outcomes. However, initial efforts to examine the impact of these factors have been hampered by gaps in population-based data. Birthplace is missing for a substantial proportion (approximately 25%) of cancer patients reported to the National Cancer Institute’s Surveillance, Epidemiology, and End Results Program, and patients with missing information are more likely to be US than foreign born.10 The resulting bias may have skewed the results of previous examinations of the association between immigration status and breast cancer stage and survival11,12 and obscured understanding of how immigration influences ethnic disparities in survival.10,13,14 Degree of Western acculturation is not assessed by cancer registries for immigrant patients, although it may have independent effects on treatment and other factors affecting health outcomes among Asians.15–18
To advance the effort to reduce breast cancer disparities in Asian women, we enhanced population-based cancer registry data to enable the examination of the effects of individual- and neighborhood-level factors on the associations with survival after breast cancer diagnosis among Chinese, Japanese, Filipina, Korean, South Asian, and Vietnamese women in California.19–22 We enhanced the data in 2 ways: (1) we used a novel method for statistically imputing immigrant status by age at assignment of Social Security number (SSN) to designate nativity where that information was missing,2,23–25 and (2) we linked registry data to US Census data to incorporate neighborhood factors characterizing Asian ethnic enclave and socioeconomic status (SES) in our analyses. We defined a neighborhood ethnic enclave as a geographical unit with a greater concentration of Asian population and speakers of Asian languages than other geographical units in California. With this enhanced registry resource, we examined whether individual-level immigrant status, neighborhood-level ethnic enclave, and SES explain differences in survival after breast cancer diagnosis within and across Asian ethnic groups.
Asian women residing in California and diagnosed with a first primary, microscopically confirmed, invasive breast cancer (International Classification of Diseases for Oncology, third edition,26 site codes 500–509, excluding morphology codes 9590–9989) between January 1, 1988, and December 31, 2005, were eligible for the analysis. Cancer cases are reported by statewide mandate to the population-based California Cancer Registry, whose data are also contributed to the Surveillance, Epidemiology, and End Results Program.27,28 The state registry routinely collects patient data from the medical record on age at diagnosis, gender, race, ethnicity, extent of disease, stage, and treatment modalities within the first 4 months after diagnosis; it obtains vital status through hospital follow-up and linkages to vital statistics, death records, and other databases, which, for the deceased, also provide the underlying cause of death.
Data from the California Cancer Registry were available for specifically classifying cancer patients as Chinese, Japanese, Filipino, Korean, South Asian (Asian Indian, Pakistani, Sri Lankan, or Bangladeshi), or Vietnamese; together, these groups represent 93% of all Asians in California.21 Registry data on patient race, ethnicity and birthplace are typically abstracted from hospital medical records; these data are obtained primarily through self-report, by assumption of hospital personnel, or by inference from other information, such as race and ethnicity of parents, maiden name, surname, and birthplace, and from death records.29 Approximately 5% of patients are coded in the registry data as “Asian, not otherwise specified”; for 55% of these patients, we were able to determine a more specific Asian ethnic population from birthplace and names (first, maiden, last) by applying the North American Association of Central Cancer Registries Asian/Pacific Islander Identification Algorithm.30 Applying the algorithm also substantially improves the previously documented moderate misclassification of Asians in the cancer registry, because we previously found that most of the misclassification occurs when individuals are classified as “Asian, not otherwise specified.”31
After exclusion of Asian women not in 1 of the 6 ethnic groups of interest, patients with missing or invalid survival time (n=5) or diagnosis at time of death (n=47), and patients whose residence at diagnosis could not be accurately geocoded to a census block group or assigned a neighborhood value (n=232), records for 20747 patients were available for our analysis.
To determine immigrant status,2 we used birthplace information from registry data, which was available for 73% of patients (70% from hospital medical records and 3% from death certificates). For the 26% of patients with unknown birthplace, we imputed immigrant status from the patient’s SSN, which is indicative of the state and year of issuance.23,25 Women who received an SSN when they were aged younger than 25 years were considered US born, and those who received an SSN when they were aged 25 years or older were considered foreign born. We derived this age cutpoint from self-reported nativity in previously interviewed cancer patients (n=1836).10 We maximized sensitivity and specificity through receiver–operating characteristic curves and confirmed the single-age cutpoint by logistic regression. Imputation resulted in immigrant status classifications with 84% sensitivity and 80% specificity of detecting foreign-born status and was similar across the Asian populations. For the 1.7% of records with missing or invalid SSNs, we assigned immigrant status according to the ethnicity–gender–age nativity distribution of the overall sample.
Because individual-level information on SES is not collected by cancer registries,32 we determined neighborhood SES according to patient residence at diagnosis with a composite index derived from US Census 1990 and 2000 block group data.33 We classified neighborhood SES into quintiles by the distribution of the composite SES index across the state of California and then into 2 categories because of small sample sizes in the quintiles: lower SES (quintiles 1, 2, and 3) and higher SES (quintiles 4 and 5).
To characterize residence in an ethnic enclave, we used principal components analysis,34 a standard technique for decomposing the covariance of selected variables that is commonly used for creating multivariable indexes.33 This technique yields a set of linear combinations of variables, which together theoretically capture all of the correlated variability in the data. We applied principal components analysis to selected Census 1990 and 2000 variables at the block group level. We determined which principal components to extract on the basis of Eigen values, screen plots, and interpretability. The census variables included in the Asian-specific ethnic enclave index were the percentages of Asian-language–speaking households that were linguistically isolated, all Asian-language speakers who spoke limited English, recent immigrants, and Asians. This index explained 63% of the variability in the data.
We identified deaths from breast cancer from the underlying cause of death on the death certificate, according to International Classification of Diseases, Ninth Revision35 or 10th Revision36 codes (174–175 and C50, respectively). Survival time was calculated in months from the date of diagnosis to (1) date of death from breast cancer (patients who died from other causes were censored, or considered not to have developed the outcome of interest, at the date of death) or from any cause (for overall survival), (2) date of last known contact, or (3) December 31, 2007 (the end of the study period), whichever occurred earliest. Of the 15793 patients alive at the end of the study period, 90% had had a follow-up date in the last year of the study and 93% in the last 2 years. Foreign-born women were slightly less likely to have had complete follow-up in the last 2 years (92%) than were US-born women (97%).
We conducted survival analyses with multivariable Cox proportional hazards regression models to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for risk of breast cancer–specific mortality and all-cause mortality. We assessed the proportional hazards assumption by visual inspection of the survival curves (log [–log] of the survival distribution function by log [months]) and by tests for time dependency; we observed no significant violations of the assumption.
First, we computed models with parameters representing each Asian ethnic and immigrant group to compare mortality with that of US-born Japanese, who were selected as the reference group because they had the highest survival rates of all Asian ethnic groups and the largest sample size. Independent variables were included via stepwise selection to assess the relative impact on the ethnic or immigrant group differences. Second, we computed ethnicity-specific models to assess the relative effects of patient characteristics on breast cancer–specific mortality within ethnic groups.
We did not conduct multilevel analyses because we found minimal spatial clustering of breast cancer cases in census block groups: the majority had only 1 (55%) or 2 (21%) cases. The neighborhood ethnic enclave variable was not included in multivariate models because of its high correlation with immigrant status and neighborhood SES and insufficient sample sizes to permit stratification. Data showing the independent association of ethnic enclave with mortality are available on request. All statistical tests were 2-sided, and we performed all analyses with SAS version 9.1.3 (SAS Institute Inc, Cary, NC). We did not show values with fewer than 5 cases so as to protect patient confidentiality.
Table 1 shows the sociodemographic and clinical characteristics of the cases by ethnic population and immigrant status. The case population was 17% Japanese, 27% Chinese, 36% Filipina, 6% Korean, 6% South Asian, and 8% Vietnamese. With the exception of Japanese women, 59% of whom were US born, the population was predominantly foreign born, ranging from 78% among Chinese, 88% among South Asians, and 92% to 93% among Filipinas, Koreans, and Vietnamese. US-born Japanese were slightly older at diagnosis than were foreign-born Japanese; among the other ethnic groups, foreign-born patients were older than were US-born patients. The distributions of year of diagnosis and marital status also varied according to ethnicity and immigrant status.
For all ethnic populations, US-born women were more likely than were foreign-born women to be diagnosed with localized disease. US-born Chinese and Korean women were more likely than foreign-born women to be diagnosed with estrogen receptor–positive tumors. US-born Chinese, Korean, South Asian, and Vietnamese women were less likely than were their foreign-born counterparts to have advanced-grade disease. With the exception of Japanese women, US-born women were more likely to live in higher-SES neighborhoods and less likely to live in ethnic enclaves (fifth quintile) than were foreign-born women. In general, US-born women had fewer deaths (breast cancer specific and all cause) and higher 5-year survival probabilities than did foreign-born women (Table 2).
Table 3 shows the HRs associated with breast cancer–specific and all-cause mortality for each ethnic population and immigrant group, relative to US-born Japanese. For breast cancer–specific mortality, rates of mortality among US-born Asian groups were statistically similar to those of US-born Japanese, with the stark exception of US-born Vietnamese women, who had a substantially lower risk of mortality (HR=0.3; 95% CI=0.1, 0.9). Except for foreign-born Japanese women, foreign-born ethnic groups had higher mortality than did US-born Japanese women, with HRs ranging from 1.4 (95% CI=1.2, 1.7) among Koreans to 1.8 (95% CI=1.5, 2.2) among South Asians and Vietnamese. A very small proportion of this variation was explained by differences in disease characteristics (stage at diagnosis, grade, histology, and estrogen receptor status); however, mortality did not have significant, independent associations with age at diagnosis, treatment, neighborhood SES, or residence in an ethnic enclave (data not shown). In general, we observed the same associations for all-cause mortality.
Table 4 shows the HRs for breast cancer–specific mortality associated with sociodemographic and clinical factors in each of the 6 ethnic populations. Patterns were similar for all-cause mortality (data not shown). The association between immigrant status and mortality varied from no statistically significant effect among Japanese, Koreans, and South Asians to a risk of mortality nearly 4 times as high for foreign- as for US-born Vietnamese.
The magnitudes of other risk factors’ effects on mortality were largely consistent across ethnic groups, with these exceptions: (1) mortality was 50% to 60% higher among previously married Japanese, Chinese, and South Asians than among women who were currently married and marginally lower among South Asians who were never married; (2) among Koreans, mortality was significantly higher for lobular cancer than for ductal carcinoma (HR=2.7; 95% CI=1.3, 5.7); (3) mortality was significantly higher for estrogen receptor–negative than estrogen receptor–positive tumors among Koreans (HR=2.2; 95% CI=1.5, 3.4) and South Asians (HR=2.6; 95% CI=1.8, 3.8), but this difference was not seen in Filipinas; and (4) mortality was 50% higher among Japanese living in lower-SES than in higher-SES neighborhoods.
Health statistics aggregated across broad ethnic categories, such as Asians, may mask disparities between specific ethnic populations. Indeed, studies that present statistics on cancer risk factors, screening, incidence, or outcomes disaggregated by Asian nationality have shown striking variation.1–3,5–7,18,37–47 However, population-based studies have not been able to address the roles in this variation played by differences in immigration and acculturation histories across those groups, in part because of the incompleteness of data on birthplace in registry data.
In our examination of the association between immigrant status and breast cancer survival across 6 Asian ethnic populations, made possible through enhancement of the cancer registry data, we found that, with the exception of Japanese women, foreign-born Asian women had consistently more advanced breast cancer stage at diagnosis and, consequently, lower survival rates than their US-born counterparts. Moreover, these differences in survival by nativity were not explained by the many sociodemographic and clinical factors we evaluated. We further found that immigrant status had a more substantial effect on survival in some ethnic groups (Vietnamese, Chinese, and Filipinas) than in others and that the magnitude of the effects of certain demographic and clinical factors also varied across groups. By contrast, neighborhood SES was associated with survival only in Japanese women.
To our knowledge, only 1 previous population-based study assessed the effect of immigrant status on breast cancer survival. In Surveillance, Epidemiology, and End Results Program data from San Francisco–Oakland, Hawaii, and Seattle–Puget Sound that excluded cases with missing birthplace information, Pineda et al. found no differences in survival after breast cancer by place of birth for Japanese, Chinese, and Filipina women.12 However, because patients in the cancer registry without available birthplace data were more likely than those for whom this information was available to be US born and alive,14 these authors’ exclusion of patients with unknown birthplace from the survival analyses would lead to underestimation of mortality differences, a concern supported by our finding of a significantly detrimental influence of immigrant status.
Although disease stage at diagnosis is the single strongest contributor to differential rates in mortality among cancer patients, we found little contribution of stage or other tumor characteristics to the observed differences in mortality by immigrant status. Within ethnic groups, a significant effect of immigrant status remained for Chinese, Filipina, and Vietnamese women. Clearly, more research is warranted to understand the complex factors that encompass immigrant status.
Studies of healthy populations have found that, compared with their US-born counter-parts, foreign-born Asians present an apparently contradictory profile for breast cancer risk and outcomes. Compared with US-born descendents of immigrants from their countries, foreign-born Asian women generally are thinner and smaller; consume more soy, less fat, and more vegetables; engage in more physical activity; are older at menarche; have children earlier in life and are more likely to breastfeed them; and are less likely to receive postmenopausal hormone replacement therapies,18,48–55 all of which are factors associated with a lower risk of breast cancer. However, they also have lower SES; more linguistic barriers, reported health conditions, and Asian-specific beliefs and health behaviors; and are less likely to have health insurance and to be screened for breast cancer,18,42,56 all of which potentially raise their risk of poor outcomes after diagnosis. Asian cultures tend to value self-sufficiency, fatalism, and familial obligation over individual physical health, and these cultural values have been found to significantly affect breast cancer treatment choices.57–61 Several studies have shown that women with certain Asian ethnicities are less likely than the general population to receive guideline-based or preferred treatment, including adjuvant radiation after breast-conserving surgery6,62 and breast-conserving treatment instead of mastectomy,17,62–65 and that these patterns are in part associated with immigration and acculturation factors.17,65
Factors associated with the process and consequences of immigration may also be relevant. Not only is immigration related directly to stress, particularly when associated with personal or political turmoil, it often requires separation from friends and family, major lifestyle changes, and acceptance of lower-status occupations than immigrants had in their home countries. Future research on the impact of the specific immigrant experience, process, and context may better help to explain the disparities we observed in breast cancer outcomes in foreign-born Asian women.
We found few differences between foreign- and US-born Japanese women in prognostic factors or survival patterns after breast cancer. We previously reported similar rates of breast cancer incidence between foreign- and US-born Japanese.2 Women in Japan have become increasingly Westernized and the population prevalence of lifestyle-related breast cancer risk factors has risen significantly. Since World War II, the number of women in the Japanese workforce has risen,66 leading to a declining birth rate.67,68 This societal shift may have resulted in changes in the population distribution of breast cancer risk factors such as higher SES,69 higher body mass index, lower parity, and later age at first birth,68,70 as well as in factors that may affect survival. This may explain why foreign- and US-born Japanese women were more similar than were immigrant and nonimmigrant members of the other Asian populations we studied. Indeed, in our data, neighborhood SES and ethnic enclave did not differ for foreign-and US-born Japanese. In addition, because Japanese women in our study generally were younger when they received an SSN (83% before age 40) than were women in other ethnic groups (65% before age 40), it is likely that foreign-born Japanese immigrated earlier in life and were thus more acculturated than other foreign-born Asians.
Our study had the distinct strength of obtaining data from a database with sufficient statistical power to examine breast cancer outcomes in 6 nationalities of Asian women subclassified by immigrant status and with a significant (19-year) follow-up period. Nonetheless, our data had some limitations. Abstracted from medical records, cancer registry data on race and ethnicity are likely not recorded consistently across reporting facilities and may also be subject to some misclassification.29 We applied an algorithm to minimize this misclassification.
Our measurement of immigrant status also relied on imputation of nativity for 26% of the patients. Although imputation was an improvement over the assumptions about missing nativity data used in previous studies, the resulting classification had 84% sensitivity and 80% specificity. Even if imputed immigrant status was not completely accurate, it was likely to be correlated with acculturation, for which immigrant status served as a proxy.
Our findings of survival differences may also have been biased by the salmon effect,71–76 the phenomenon of migrants with serious illnesses returning home to their birth countries to die. We noted slightly higher loss to follow-up among the foreign- than US-born Asians, suggesting that the HRs comparing foreign-born women to US-born women may have been slightly underestimated. Other limitations of our study included a lack of detailed or complete registry data on clinical factors, treatment, and behavioral, environmental, and genetic factors that may influence survival after breast cancer diagnosis. Some groups stratified by ethnic population and immigrant status yielded small sample sizes.
Asians are the fastest-growing racial group in the United States,77 composing 4.2% of the national population and 12% of the California population.78 The process of immigration and acculturation, as well as intrinsic cultural differences between groups, may have important and lasting effects on the health of these populations, yet very little is known about patterns of health and disease across Asian populations stratified by immigration and acculturation characteristics. US Asian women are the only ethnic group for which cancer far outweighs heart disease as the leading cause of mortality,79–81 with breast cancer as the leading cause of cancer incidence and second leading cause of cancer mortality.1,3
Our study takes the first step toward providing the evidence base for a better understanding of how population-based cancer patterns vary with immigration and acculturation. A paucity of detailed health and other data specific to distinct Asian ethnic groups in the United States has caused health disparities between them to be largely overlooked. By enhancing cancer registry data, we uncovered important breast cancer survival disparities among Asian immigrant groups, which require that we shape public health policies for specific Asian subpopulations in the United States.
This study was supported by the Safeway Foundation, the National Cancer Institute (grant R03 CA117324), and the Surveillance Epidemiology and End Results program. The collection of cancer incidence data used in this study was supported by the California Department of Health Services as part of the statewide cancer reporting program mandated by California Health and Safety Code Section 103885; the National Cancer Institute’s Surveillance, Epidemiology and End Results Program (contract N01-PC-35136 with the Northern California Cancer Center, contract N01-PC-35139 with the University of Southern California, and contract N02-PC-15105 with the Public Health Institute); and the Centers for Disease Control and Prevention’s National Program of Cancer Registries (agreement U55/CCR921930-02 with the Public Health Institute).
The authors thank Preety Kalra, Jane Pham, June Kristine Winters, and Cammie d’Entremont for their help with this article.
Reprints can be ordered at http://www.ajph.org by clicking the “Reprints/Eprints” link.
ContributorsS. L. Gomez supervised the collection of data, originated the study, and supervised all aspects of its implementation. S. J. Shema conducted the data analysis. All authors helped to conceptualize ideas, interpret findings, and review drafts of the article.
Note. The ideas and opinions expressed herein are those of the authors, and endorsement by the State of California’s Department of Health Services, the National Cancer Institute, the Centers for Disease Control and Prevention, or their contractors and subcontractors is not intended and should not be inferred.
Human Participant Protection
This study was approved by the Northern California Cancer Center institutional review board.