Cancer Mortality in U.S.-Born versus Foreign-Born Asian American Groups (2008–2017)

This article is featured in Highlights of This Issue, p. 1

Asians account for approximately 60% of the global population (1). In the United States, Asian Americans (AA) represent the fastest growing racial/ethnic group and although AAs are a highly heterogenous group – in social, cultural, and political contexts, individually and nationally – most U.S. research treats them as a single group (2, 3).

Cancer mortality in AAs is a growing topic of interest in healthcare – and disaggregation of AAs by groups has demonstrated important heterogeneity in their health outcomes and health behaviors. Our previous research found that AAs have lower overall cancer-specific mortality rates than non-Hispanic whites but disaggregation revealed important differences in mortality burden of cancer between Asian-American ethnic groups. In particular, infection-related (e.g., stomach and liver) cancer mortality rates are disproportionately high among Koreans, Chinese, and Vietnamese but not among Japanese, Filipino, or Asian Indian decedents (4).

Nativity is an important factor to consider as immigration patterns and acculturation may affect AA cancer burden differently for foreign-born versus U.S.-born AAs. Asian-American immigration to the United States represents more than 30 countries of origin: heterogenous health literacy, lifestyles, diet, body size, neighborhood, socioeconomic status, and infection risk factors which may be directly attributable to nativity and migration trends (5, 6). The prevalence of known risk factors for cancer vary widely across Asian countries of origin. The prevalence of tobacco smoking (2015) was high among Asian males, ranging from 20% in India to 50% in Korea, and was very low among Asian females, ranging from 1.3% in Vietnam to 10.6% in Japan (7). The prevalence of obesity (2016) was higher among females (7.5% in Philippines to 2.6% in Vietnam) and was lower among males (5.9% in China to 1.6% in Vietnam; ref. 8). Importantly, some of these risk factors and health behaviors are a stark contrast to Western risk factors and behaviors, which are adopted, although nonuniformly, by U.S.-born children of immigrants (9, 10). Epidemiologic research of immigration patterns often reveals that recent immigrants have better health outcomes than their native-born counterparts (11, 12). This so called “healthy immigrant effect” (HIE) may have multiple explanations – e.g., healthier, wealthier people tend to migrate, which may be maintained after migration (13). HIE has been seen to dissipate for some measures of health and disease and foreign-born rates have been shown to converge with the rates of the native-born populations within a generation or two (14).

In this study, we add to our previous findings (4) by examining recent cancer mortality burden among AA groups (separately for Asian Indian, Chinese, Filipino, Japanese, Korean, and Vietnamese) with additional years of follow up, and stratification by decedent country of birth (U.S.-born vs. foreign-born).

We analyzed national mortality data from the National Center for Health Statistics (NCHS) reflecting all cancer-related deaths from 2008 to 2017, under a data use agreement with the Centers for Disease Control and Prevention. Death certificates are completed by the healthcare provider, medical examiner or funeral home director, often with input from the decedent's family and medical team. Year of death, 5-year age groups, sex (male or female), race/ethnicity of decedent, underlying cause of death, and country of birth were identified from death certificate data. All cancer-related deaths were determined by the “underlying cause of death” coded by NCHS using International Classification of Diseases (ICD-10).

We selected cancer sites that contribute significantly to male and female AA cancer mortality based on 2016 estimates (15): Lung and bronchus (C34), female breast (C50), liver and intrahepatic bile duct (C22), colon and rectum (C18), pancreas (C25), prostrate (C61), stomach (C16), ovary (C56), leukemia (C91-C95), non–Hodgkin lymphoma (NHL; C82-C85), kidney and renal pelvis (C64, C65), oral cavity and pharynx (C01, C02, C10-C14.0, C03-C09, C14.2-C14.8), and uterine corpus (C54, C55). Following Rehkopf and colleagues (16), we created meaningful categories of cancer sites reflecting established risk factors and potential for secondary prevention allowing for some cancers to be included in more than one category (i.e., breast cancer can be detected by screening and is diet-/obesity-related). “Screen-detected cancers” including colon and rectum (colonoscopy or fecal immunochemical test; ref. 17), female breast (mammography; ref. 18), and prostate (PSA shared decision making; ref. 19) were defined as cancers with screening recommended among the general U.S. population and not only among high-risk groups. “Tobacco-related cancers” including oral cavity and pharynx, leukemia, lung and bronchus, kidney and renal, and pancreas were defined as cancers related to use of tobacco products. “Infection-related cancers” including oral cavity and pharynx (human papillomavirus), liver (hepatitis B virus), NHL (Epstein–Barr virus), and stomach (Heliobacter pylori) were defined as cancer related to viral infections. “Diet-/obesity-related cancers” including colon and rectum, female breast, ovary, and uterine corpus were defined as cancers related to obesity and the associated chronic inflammation.

The study population included 98,826 AA decedents who were identified on their death certificates as Asian Indian, Chinese, Filipino, Japanese, Korean, or Vietnamese; decedent counts by AA ethnicity are also provided (Table 2). Decedents who were identified as “Other Asian” or as belonging to multiple Asian groups were not included in the present study. The death certificate identified decedents who were born in the U.S. states and territories (U.S.-born) versus outside of the U.S. states and territories (foreign-born). Those reported with “unknown place of birth” (N = 1,279) were excluded from analysis. The present (2003) version of death certificate which includes all 6 major Asian groups was not adopted by all states at the same time in the United States; prior to the 2003 version, death certificates did not have checkboxes for Asian Indian, Korean, or Vietnamese (20). To accommodate this, decedent counts for these three groups were introduced for each state only after they adopted the 2003 form (Supplementary Table S1).

Population denominator counts for 2008 to 2017 were obtained from the American Community Survey (ACS), retrieved from Integrated Public Use Microdata Series (IPUMS USA) and stratified by 5-year age groups, sex, race/ethnicity, and country of birth (21). To account for the 2003 death certificate not being adopted at the same time in all states, the affected population counts were also introduced for each state after their adoption of the 2003 version of the death certificate form which included all 6 Asian groups.

Using decedent counts as the numerator and the census population as the denominator, we calculated age-adjusted 10-year mortality rates (AAMR) and 95% confidence intervals (CIs) spanning 2008 to 2017, stratified by sex, race/ethnicity, nativity, cancer etiology (as well as by specific cancer sites). AAMRs were calculated as deaths per 100,000 persons by applying the stratum-specific mortality rates to the United States 2000 standard population; AAMRs are comparable across strata (groups) as they are reweighted to the same reference population. AAMRs based on small decedent counts tend to have poor reliability (22, 23), so we suppressed them if the stratum-specific decedent counts were less than or equal to 15.

We also calculated 10-year standardized mortality ratios (SMR) and 95% CIs stratified by sex, race/ethnicity, and etiology (as well as by specific cancer sites), using the foreign-born population as the reference group because most Asian ethnic groups had higher foreign-born than U.S.-born death counts. SMRs were indirectly standardized by dividing the stratum-specific observed death counts among the U.S.-born by the expected number of deaths if they had experienced the same age-specific death rates as foreign-born AAs in that stratum. SMRs are not valid for cross-group comparisons, because they are not reweighted by a common standard distribution, but as calculated they do allow for the comparisons within each AA group by estimating the relative excess (or dearth) of deaths observed in U.S.-born decedents compared with what would have been expected if they had had the same age-specific death rates as the foreign-born decedents. In addition to results by cancer etiology, we also provide cancer site-specific results in Supplementary Tables S2 and S3.

To check for potential bias in our AAMRs arising from sampling error in the ACS, we followed Jiang and colleagues' methods (24) and calculated the coefficient of variance (CV) for each sampled denominator data using the ACS-provided replicate weights (25). CVs are defined as the ratio of the sampling error and the denominator count estimate. CVs less than 10% are considered good, producing nearly unbiased AAMRs from sampled denominators, with approximately unbiased and reasonably reliable variance estimator. CVs more than 30% are considered large and can produce slightly biased results (24). CVs were calculated for all categories by age groups, race, gender, and nativity; stratified interquartile ranges (25%–75%) of CV across 18 age groups are presented (Table 1).

For any groups which had CVs more than 30% we calculated bias-corrected AAMRs (95% CI) and SE in those groups following the method described in Jiang and colleagues (24) to compare with the original estimates. Data management and analysis was conducted using SAS version 9.4 (SAS Institute) and R Studio 1.3.1093. PROC STDRATE in SAS was used to calculate direct AAMRs and SMRs. R Studio was used to prepare the denominator data (26). All figures were created using Tableau (27).

There were more foreign-born than U.S.-born individuals in all AA groups overall, and by sex, except for among Japanese Americans. Sixty-five percent to 70% of each AA group were foreign-born. Among Japanese, 35% of males were foreign-born and 49.8% of females were foreign-born (Table 1).

The interquartile ranges presented for the CVs were all below 30%, and most were below 10%. According to Jiang and colleagues (24), large sampling errors (>30% CVs) can produce slightly biased results.

Within each cancer etiology grouping, there were more foreign-born AA decedents than U.S.-born in every AA group except for Japanese. Proportions by etiology grouping are not provided as the categories are not mutually exclusive. By etiology, tobacco-related cancer mortality was the most common (Table 2). By site, tumors of the lung and bronchus were most common (28.1%) followed by colon and rectum (13%) and liver and intrahepatic bile ducts (11.3%; Supplementary Table S2).

Foreign-born males had some of the highest AAMRs for tobacco-related cancers compared with the lowest AAMRs which were among U.S.-born females. AAMRs were generally higher for foreign-born AAs than for their U.S.-born counterparts except for Japanese males [Foreign-born AAMR 43.02 (38.72–47.31); U.S.-born AAMR 55.38 (53.05–57.72); Fig. 1] and Filipino women [Foreign-born AAMR 26.99 (26.15–27.82); U.S.-born AAMR 39.28 (35.35–43.21); Fig. 2], where foreign-borns had lower AAMRs in comparison with their US-born counterparts. Asian Indian females had similar AAMRs by nativity [Foreign-born AAMR 15.9 (14.97–16.84); U.S.-born AAMR 15.04 (9.32–20.75); Fig. 2].

The SMRs for tobacco-related cancers' reveal that U.S.-born AAs experience significantly lower death rates compared with their foreign-born counterparts except for a few groups. U.S.-born Japanese males were 30% [SMR 1.30, (1.25–1.36)] more likely to succumb to tobacco-related deaths than their foreign-born counterparts and U.S.-born Filipino females were 42% [SMR 1.42 (1.28–1.55)] more likely to die of tobacco-related cancer than their foreign-born counterparts (Fig. 3).

Foreign-born AAs had some of the highest AAMRs for screen-detected cancers with the lowest AAMRs being among U.S.-born AAs (Figs. 1 and 2). AAMRs were consistently higher for foreign-born AAs except for Japanese males, for whom the AAMR was higher among the U.S.-born [foreign-born AAMR 22.86 (19.66–26.06); U.S.-born AAMR 27.77 (25.15–29.39); Fig. 1].

The U.S.-born AAs experienced significantly lower rates of screen-detected cancer-related deaths compared with their foreign-born counterparts, except for Japanese males. Japanese U.S.-born males were 28% [SMR 1.28 (1.21–1.35)] more likely to die of screen-detected cancers than their foreign-born counterparts (Fig. 3).

Diet-/obesity-related cancer AAMRs are highest among females: Filipino [foreign-born AAMR 31.34 (30.46–32.21); U.S.-born AAMR 27.54 (24.43–30.65)], Japanese [foreign-born AAMR 31.02 (29.25–32.78); U.S.-born AAMR 31.47 (29.69–33.25)], and Chinese [Foreign-born AAMR 25.87 (25.14–26.60); U.S.-born AAMR 27.10 (24.73, 29.47)]. Among females, the AAMRs were similar by nativity except for among Asian Indian, Korean, and Vietnamese where the foreign-born AAMRs were higher than US-born AAMRs (Fig. 2). Among males, the AAMRs are similar by nativity except for among Chinese and Koreans for whom the foreign-born AAMRs were higher than their U.S.-born counterparts (Fig. 1).

Korean [female SMR 0.57 (0.42, 0.72); male SMR 0.51 (0.29, 0.73)] and Vietnamese [female SMR 0.56 (0.33, 0.80); male SMR 0.41 (0.13, 0.70)] had significantly lower rates of cancer-related deaths related to diet among US-born compared with their foreign-born counterparts. In most other groups, the SMRs were null (Fig. 3).

All foreign-born groups had higher infection-related cancer AAMRs than U.S.-born groups, among males and females, except for Filipino males who had similar AAMRS by nativity [foreign-born AAMR 22.66 (21.66–23.67); U.S.-born AAMR 21.20 (17.96–24.44); Figs. 1 and 2]. The highest AAMRs were among foreign-born males, specifically Korean [AAMR 41.54 (39.54–43.53)] and Vietnamese [AAMR 41.39 (39.68–43.09); Fig. 1]. The lowest AAMR as among U.S.-born Korean females [AAMR 6.80 (4.13, 9.46); Fig. 2].

Most U.S.-born AA groups experienced significantly lower rates of infection-related cancer deaths compared with their foreign-born counterparts. The lowest SMR was among Vietnamese males [SMR 0.38 (0.23–0.52)] and Asian Indian females [SMR 0.38 (0.15–0.62); Fig. 3].

Some of the 18 age-group categories among U.S.-born Vietnamese males and females had CVs more than 30%. The sampling error-bias–corrected estimate AAMRs for the U.S.-born Vietnamese males and females didn't differ from our original AAMRS. The SEs in the original estimates increased slightly after correction but remained small in general (Supplementary Table S4).

We analyzed nationally representative death certificate data to estimate cancer mortality rates among AAs by nativity status (US-born vs. foreign-born) between 2008 and 2017. We observed that, Japanese and Filipino were the only AA group to have significantly higher mortality rates among U.S.-born than among foreign-born. For all other groups, compared with foreign born AAs, U.S.-born AAs generally had significantly lower mortality burden of infection-related cancers (SMRs: 0.38–0.94) and screen-detected cancers (SMRs: 0.39–0.91). In addition, for tobacco-related and infection-related cancers, males always had higher death rates than females. Among obesity-related cancers, mortality burden was higher among females, most likely driven by breast, ovarian, and uterine cancers (Supplementary Table S3). Among screen-detected cancers, mortality rates were similar for males and females.

Few studies have examined cancer mortality among AAs by nativity, which may influence cultural practices, lifestyle, health beliefs, and acculturation to U.S. normative behaviors. Two studies looked at Asian cancer mortality by nativity but did not disaggregate by AA groups; one reported lower all-cancer–related death rates among foreign-born and the latter reported higher gastric cancer mortality rates among foreign-born (28, 29). Recently, more attention has been given to Asian-American group disparities, as disaggregated Asian ethnicity data has become available on mortality databases and national surveys (National Health Interview Survey, National Health and Nutrition Examination Survey; refs. 4, 30–33).

In our study, among infection-related cancer types, foreign-born AAs had higher mortality rates than U.S.-born for all AA groups. Many of the infections driving these cancers are highly prevalent in Asian countries. For instance, hepatitis B virus (HBV) prevalence is 4.3% in the United States and much higher in some Asian countries: Philippines (10.4%), Vietnam (8.1%), China (6.3%), Republic of Korea (4.4%), India (3.0%), and Japan (1.0%; refs. 34–36). Unsurprisingly, Vietnam, China, and South Korea rank in the top 10 countries in the world with the highest incidence of liver cancer (37). In the United States, AAs have the highest rates of HBV infection and some of the highest rates of liver cancer (38, 39). This pattern is also seen for gastric cancer in which H. pylori prevalence is 34% in the United States compared with Vietnam (70%), India (64%), China (56%), Korea (54%), and Japan (54%; ref. 40). Aggregated Asian-Americans, Native Hawaiians, and Pacific Islanders have about twice the incidence and death rates of stomach cancer than non-Hispanic Whites (41). Liver cancer and stomach cancer incidence is increasing in many AA groups, being among the five most diagnosed cancers (42). NHL accounts for 3% of cancer deaths in the United States and AAs have the lowest incidence and mortality rate by ethnicity (43).

While foreign-born and U.S.-born Japanese and Filipino Americans have some of the highest death rates from screen-detected cancers, Koreans and Vietnamese had the highest relative differences in death rates by nativity status. Mortality from screen-detected cancers is affected by access to care, appropriate treatment, and follow up. When aggregated, AAs have been shown to have similar rates of cancer screening adherence as non-Hispanic Whites, but when disaggregated, AAs' cancer screening was heterogenous (44). Incomes vary widely among AAs, and income inequality among Asian groups is the greatest amidst any racial group in the United States and rising (45). Among AAs, Koreans (15.0%) and Vietnamese (14.7%) have the highest proportion of individuals living under the federal poverty level (46). Asian cancer screening attendance is influenced by access to care (health insurance, poverty), language discordance, cultural discordance, propensity to seek care for acute symptoms, health literacy, and health awareness—all of which are modified by or related to nativity, specifically recency of immigration, and acculturation to U.S. culture and practices (44, 47–50).

Overall, we found that Asians born outside the United States had higher death rates than those born in the United States; three notable exceptions were screen-detected cancers and tobacco-related cancers among Japanese males and tobacco-related cancers among Filipino females. The HIE postulates that people immigrating to a higher-income country tend to be healthier on average than those born in the high-income country; this phenomenon appears to fade the longer the individual resides in the new country perhaps due to acculturation or just regression to the average (11, 13). While this may be true in some cases (51, 52), our findings refute this theory for cancer mortality in AAs, as we observed higher death rates among foreign-born than U.S.-born AAs. Additionally, another study of the burden of cancer in AAs between 2003 to 2011 found that non-Hispanic Whites had higher death rates than all Asian groups for cancer sites (stomach, colorectal, pancreatic, lung, breast, ovarian, and prostate) except for liver cancer which was higher among AAs (4). Comparing the death rates from our study with the death rates from this study, which didn't stratify by nativity, we see that the death rates among Asians is driven by cancer-related deaths in the foreign-born population. Cancers often take years to develop, and the U.S.-born Asian-American population in our study was smaller and younger than the foreign-born, which may indicate that the full extent of U.S.-born AA cancer mortality burden is yet to be realized. However, compared with native-born AAs, foreign-born AAs face additional barriers to cancer screening attendance and receipt of guideline concordant treatment—they have higher endemic chronic infection rates that are often undiagnosed and undertreated, and often adopt a higher-fat Western diet upon immigration to the United States—potentially driving the comparatively higher cancer-related death rates we observed among foreign-born AAs.

Our study has several limitations. While we have captured every death in the United States, states adopted the National Vital Statistics 2003 Standardized Death Certificate unevenly over the last 18 years with some states. Wyoming is still not reporting disaggregated AA deaths (20). Additionally, some immigrants return to their country of origin when severely ill for traditional family support and lower cost of care, which may artificially decrease recorded mortality rates amongst foreign-born AAs resulting in a bias sometimes referred to as the “salmon bias”. Furthermore, U.S. death certificate data is completed by physicians, medical examiners, or funeral home directors who may inadvertently misclassify an individual's ethnicity if family is unavailable. U.S. death certificate data does not include other demographic or health information and is unlinked to patient electronic health records. Finally, multiracial AAs were not included in our study, and we relied solely on the primary underlying cause, which may underestimate cancer burden in some decedents (53). By not counting multiracial AAs, we may be undercounting the cancer-related deaths in the U.S.-born AAs as there may be more second or third generation Asian immigrants who identify as multi-race. Additionally, we collapsed death counts over a 10-year period to account for the small sample sizes of (especially U.S. born) decedents and provide results that would otherwise be suppressed if we presented 1-year or 5-year rates. However, this does assume that the foreign-born Asian population is in a steady state over the 10 years and while this assumption might be violated with the recent increases in immigration, most recent immigrants tend to be of working age (18–64 years old; ref. 54) and thus are less likely to be affected by cancer-related deaths. Another limitation is that we used the ACS as our denominator data for the AAMR calculation instead of the Census as it contains information on birthplace—the ACS is a sample of the United States and suffers from sampling error. We conducted sensitivity analyses to check if our estimates suffered from sampling error bias; after bias correction, our AAMRs and 95% CI did not differ substantially from our original AAMRs—this may be due to the pooling of 10 years of data with a large denominator, as well as small numerator counts in our data.

Our study illustrates pressing need for disaggregation of AA groups to understand and improve cancer outcomes for Asians in America. We observed substantial heterogeneity in Asian cancer outcomes and higher mortality among AAs born outside of the United States. This heterogeneity points to the need for precise culturally tailored health care and awareness by healthcare providers, policy makers, and Asian communities, to improve the health of Asians across the United States.

No disclosures were reported.

O. Tripathi: Conceptualization, data curation, formal analysis, visualization, methodology, writing–original draft, writing–review and editing. Y. He: Data curation, formal analysis, writing–review and editing. B.Y. Han: Data curation, writing–original draft, writing–review and editing. D.G. Paragas: Data curation, writing–original draft, writing–review and editing. N. Sharp: Project administration, writing–review and editing. S. Dan: Formal analysis, methodology. M. Srinivasan: Supervision, writing–original draft, writing–review and editing. L.P. Palaniappan: Resources, supervision, writing–review and editing. C.A. Thompson: Conceptualization, resources, supervision, methodology, writing–review and editing.

This study was considered not-human subjects research by the Stanford University Institutional Review Board (Protocol #53429). L.P. Palaniappan was supported by NIH grants (5UL1TR003142 and 5R01MD007012). We would like to acknowledge Dr. Jin Long for his assistance with the sensitivity analysis.

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.