Disparities in Cancer Incidence among Chinese Population versus Migrants to Developed Regions: A Population-Based Comparative Study

With increasing incidence and mortality, cancer is the leading cause of death in China and is a major public health concern (1). Much of the changing burden is attributable to population growth, aging, and sociodemographic changes. For example, the persistent decreases of liver cancer and gastric cancer in the last decade were mainly due to the effective control of chronic infections (1). In contrast, prostate cancer in men and thyroid cancer in women experienced significant rises since the beginning of this century (1–3), which might ascribe to the shifting to Western diet and the increased use of new imaging technologies in the assessment of the thyroid gland, respectively.

Understanding the etiology of cancer has kept researchers occupied for centuries, although it has become increasingly clear that cancer can be considered neither purely genetic nor purely environmental (4). Previous Genome-wide Association Studies (GWAS) have suggested that the nuance in gene might have significant impact on cancer development (5, 6). However, growing evidence showed that cancer may be predominantly an environmental disease (7, 8), because only a small proportion of cancers follow a Mendelian pattern of inheritance, and the incidence of cancer changes when people are exposed to different cultures and lifestyles (9–11). Briefly, cancer is the detrimental derivative of interaction between the gene and environment. Nevertheless, it is hard to disentangle the absolute impacts of gene and environment on oncogenesis.

To further understand the environmental influence on cancer, we designed a comprehensive migration study based on the cancer registry data covering the last three decades. Chinese people living in Shanghai, Hong Kong, Macau, Taiwan, Singapore, and Los Angeles were included in this study. Chinese are the largest ethnic group in the world, composing 20% of the entire global human population (12). Chinese people living in Shanghai, Hong Kong, Macau, and Taiwan were either indigenous or immigrated from elsewhere in China, whereas the majority of their counterparts in Singapore and Los Angeles were descendants of immigrants who landed in the 1800s. Compared with Shanghai, the other five regions are inconsistent in many ways, including the social system, diet pattern, and natural environment. However, Shanghai is a metropolis with a higher human development index than the average of China, and Hong Kong, Los Angeles, Singapore, Taiwan, and Macau had similar or slightly higher development index compared with Shanghai. Moreover, the cancer registries in these regions have a long history and wide coverage and high quality in cancer data. All these indicate the equivalence of gross income, health service accessibility, life expectancy, and education among these regions, thereby providing the possibility and reducing the bias to compare the cancer incidence among the selected regions. The results of our study might provide further insights into the understanding of the exogenous factors in the etiology of cancer.

We collected the annual cancer case data of Shanghai, Hong Kong, Taiwan, and Macau from their cancer registry system, respectively. Cancer data of Chinese people living in Los Angeles and Singapore was collected from Cancer Incidence in Five Continents Plus (13) (http://ci5.iarc.fr; the details of cancer registries are presented in Supplementary Table S1). The time coverage of these data is shown in Supplementary Fig. S1. The original cancer data were categorized by sex, age (5-year interval from 0−4 years to ≥85 years) and cancer sites, which were identified via the International Classifications of Diseases, 10th version (ICD-10). For example, the code “C22” denotes primary liver cancer, including liver cell carcinoma, intrahepatic bile duct carcinoma, and so on. However, the histology of cancer was not taken into account due to the paucity of histologic information. Finally, a total of 27 cancers were included in this study (for details, see Table 1). All recorded cases have been validated for their demographic data, information on the topography, and histology, if provided, by the health professional work staff.

To better match the data from different regions, the whole period has been separated into three un-overlapped intervals, 1983 to 1994, 1995 to 2007, and 2008 to 2013 (Supplementary Fig. S1). Cancer data in children (0–14 years) was excluded, and then the age at diagnosis of cancer (≥15 years) was integrated into three categories: 15 to 34, 35 to 64, and ≥65 years, denoting adolescents and young adults, middle-aged people, and elderly people, respectively. Meanwhile, we collected the corresponding population data either from regional statistical bureau or from International Association for Research on Cancer (IARC). Likewise, we processed the population data as the cancer data described above.

The Standard World Population 2000 was used to estimate age-standardized rates (ASR) of cancers per 100,000 person years for all groups. The estimated annual percentage change (EAPC) was used to quantify the ASRs trends. A regression linear model was fitted to the natural logarithm of the ASRs, that is, y = α + βx + δ, where y = ln(ASRs) and x = calendar year, and the EAPC was calculated as |100{\rm{\ }} \times {\rm{\ }}( {{e^\beta } - 1} )$|⁠.

Instead of the traditional Poisson regression model that previous studies used (14, 15), a rate model was applied to estimate the incidence rate ratios (IRR) of cancers with the Shanghai set as reference, because the number of cancers observed was largely dependent on the population size. The rate model can be termed as the following formula:

Furthermore, this can be rearranged as:

where ln (population) is the log-offset for population at risk in each category. The “population” has been standardized by the World Population 2000 and the “case” was accordingly the expected case to overcome the bias introduced by different population structures in different regions. As a result, the original rate model can be termed as:

where Case_i and Pop_i denote the cancer cases and population in age group i, respectively. The ω_i denotes the weight of this age group and the vector X represents the region variable here.

All statistical tests were implemented with R (R core team, version 3.4.0). A P value of less than 0.05 was considered statistically significant.

As shown in Tables 1A and B, the ASRs of all cancer sites combined (ICD-10 code: C00-96bC44) varied considerably among regions, with the highest rate found in Hong Kong, followed by Singapore, Los Angeles, Taiwan, Shanghai, and Macau in both sexes. In men, the incidence of cancer was annually decreased in Singapore and Hong Kong by 0.32% (95% CI = −0.52% to −0.12%) and 1.49% (95% CI = −1.62% to −1.37%), respectively; whereas the incidence of cancer was increased in Los Angeles and Taiwan by 0.97% (95% CI = 0.54%–1.40%) and 2.87% (95% CI = 1.48%–3.85%), respectively. Cancer in Shanghai and Macau remained stable in the study period (Table 1A; Supplementary Fig. S2). In women, the incidence of cancer increased in all regions except Hong Kong (EAPC = −0.72%; 95% CI = −0.90% to −0.54%) and Macau (EAPC = 0.74%; 95% CI = −3.22%–3.13%; Table 1B; Supplementary Fig. S2).

For specific cancer site, the gastrointestinal cancers, prostate cancer, gynecologic cancers, and thyroid cancer were among those with the most significant temporal trends. In Shanghai, the most striking decreases were detected in upper gastrointestinal carcinomas (stomach, liver, and esophagus). However, colon, rectum−anus, pancreas, and gallbladder cancers experienced significant increases over the last three decades. Of note, the most pronounced increase was found in prostate cancer in men and thyroid cancer in women (EAPC = 9.27, 95% CI = 8.57–9.97; EAPC = 9.46, 95% CI = 8.28–10.65, respectively). Similar changing patterns were found in Singapore and Hong Kong in both sexes, with the most significant increase seen in prostate cancer in men and Hodgkin lymphoma in women. Different to Shanghai, most cancers in Los Angeles and Macau remained stable but increased in Taiwan in both men and women over time. Unexpectedly, in Los Angeles, the ASR of liver cancer increased by 1.30% per year (95% CI = 0.11%–2.51%) in men and by 3.85% per year (95% CI = 0.95%–6.83%) in women, respectively (Table 1A and B). The temporal trends of cancers, by sex and regions, were presented in Supplementary Figs. S3–S29.

Figures 1 and 2 display the rank of cancer sites, by regions and periods, in terms of the log-scale cancer incidence in men and women, respectively. Seven gastrointestinal cancers (stomach, liver, esophagus, colon, rectum−anus, pancreas, and gallbladder) were included to compare the incidence across regions. As shown in Figs. 1 and 2, gastrointestinal cancers ranked high in terms of the ASR regardless of sex, regions, and periods.

In both sexes, the ASR of stomach cancer was highest in Shanghai (55.14 per 100,000 in men and 26.24 per 100,000 in women) compared with all other regions and nearly 3-fold higher than that in Los Angeles, with similarly large reductions in risk seen separately in Hong Kong, Singapore, Taiwan, and Macau (Table 1; Fig. 3). Likewise, similar patterns were found in pancreatic cancer in both sexes. The ASR was 10.60 per 100,000 in Shanghai men and 7.57 per 100,000 in Shanghai women, while approximately 20% to 40% reductions in risk was witnessed in other regions (Fig. 3). Different to stomach and pancreatic cancer, reverse pattern, especially in men, was found in colon cancer. Men in Shanghai have a nearly 20% to 40% reduction in cancer risk when compared with their counterparts in other regions (Fig. 3). However, the disparities in women were not as significant as those in men. For example, there was no significant difference between Shanghai and Macau (IRR = 1.17; 95% CI = 0.84–1.39). An analogous pattern was found in rectum−anus cancer when compared with colon cancer, although the differences between Shanghai and Los Angeles in both men and women were nonsignificant (IRR = 1.06, 95% CI = 0.91–1.22; IRR = 0.98, 95% CI = 0.95–1.06, respectively; Fig. 3). For liver cancer in men, significantly higher ASRs were found in Hong Kong and Taiwan, whereas significantly lower ASRs were found in Los Angeles and Singapore, in contrast to Shanghai, with the IRRs 1.11 (95% CI = 1.04–1.17), 1.87 (95% CI = 1.80–1.96), 0.70 (95% CI = 0.58–0.81), and 0.93 (95% CI = 0.90–0.96), respectively. In women, the ASR of liver cancer was the second highest in Shanghai, behind Taiwan (IRR = 2.03, 95% CI = 1.95–2.10). There were also substantial differences between Shanghai and the rest of the regions, with IRRs of 0.62 (Los Angeles), 0.93 (Hong Kong), 0.78 (Singapore), and 0.82 (Macau), respectively (P < 0.001). For esophageal cancer, the ASR was the second highest and highest in Shanghai men and women, respectively (14.93 per 100,000; 5.55 per 100,000). The most pronounced differences were detected between Shanghai and Los Angeles in both sexes, with the IRRs of 0.35 (95% CI = 0.31–0.40) and 0.23 (95% CI = 0.17–0.28), respectively (Fig. 3). Gallbladder cancer was rarely diagnosed in the general population compared with other gastrointestinal cancers. In men, the ASR was highest in Hong Kong, followed by Shanghai (IRR = 1.11, 95% CI = 1.07–1.16). There were also considerable differences between Shanghai and other regions (Fig. 3). In women, the ASR was highest in Shanghai (6.20 per 100,000), and approximately half of that in Los Angeles, Singapore, and Macau (Fig. 3).

Over the past decades, lung cancer was the most common malignancy in men irrespective of region and period (Fig. 1). The lung cancer ASRs were varied across regions, ranging from 49.88 per 100,000 in Los Angeles to 97.01 per 100,000 in Hong Kong. When compared with Shanghai, men in Los Angeles and Taiwan have nearly 40% and 25% reduction in risk, respectively (IRR = 0.58, 95% CI = 0.52–0.67; IRR = 0.75, 95% CI = 0.70–0.80). In contrast, men in Hong Kong and Singapore have an approximately 20% to 30% rise in risk (IRR = 1.29, 95% CI = 1.22–1.36; IRR = 1.21, 95% CI = 1.16–1.26). In women, lung cancer ranked top three in most regions and periods (Fig. 2). The significant difference was only detected between Shanghai and Hong Kong (IRR = 1.42, 95% CI = 1.36–1.47; Fig. 3).

The most common four gynecologic cancers (breast, cervical, corpus, ovary, and other uterine adnexa cancer) were included in this study. As shown in Fig. 2, breast cancer was the leading malignancy in most regions and periods. The ASR of breast cancer was lower in Shanghai compared with Los Angeles, Singapore, Hong Kong, and Taiwan but slightly higher than Macau. Substantial differences between Shanghai and these regions were detected, with the most striking disparity found between Los Angeles (IRR = 1.78, 95% CI = 1.72–1.85; Fig. 3). For cervical and corpus cancer, the ASRs were lowest in Shanghai and varied 3-fold across regions (Table 1B). Significant disparities in risk were found between Shanghai and any other of the regions, with the exception of Macau in corpus cancer (Fig. 3). For ovary and other uterine adnexa cancer, women in Shanghai have a relatively lower ASR than that in Los Angeles, Singapore, and Hong Kong, but a higher ASR compared with that in Taiwan and Macau. The regional differences were significant, with the point estimates of IRR ranging from 0.73 in Macau to 1.28 in Singapore (Fig. 3).

The rank of prostate cancer remained stable in American Chinese from 1983 to 2007. However, in other regions, a dramatic increase in prostate cancer rank was observed. For example, prostate cancer ranked 14th in 1983 to 1994 in Shanghai, and then increased to third in 2008 to 2013 (Fig. 1). The highest incidence was observed in American Chinese, with the overall ASR of 55.20 per 100,000, which was nearly five times higher than that in Shanghai (IRR = 4.78, 95% CI = 4.61–5.13). Moreover, nearly 3-fold higher risks of prostate cancer were found in other regions when compared with Shanghai (Fig. 3).

The rank of oral cavity and pharynx cancer remained relatively stable over the last three decades in both men and women (Figs. 1 and 2). The ASR was lowest in Shanghai, with the highest rates were seen in Taiwan in men (Table 1A), and Hong Kong in women (Table 1B). Subsequently, the strikingly increased risks were found in other regions relative to Shanghai (Fig. 3).

The rank of thyroid cancer remained stable in regions other than Shanghai, in which the rank increased from 19th and 18th in 1983 to 1994 to 13th and 5th in 2008 to 2013 in men and women, respectively (Figs. 1 and 2). In men, the ASR of thyroid cancer was lower in Shanghai than that in Macau, while higher when compared with other regions (Table 1A; Fig. 3). For women, the ASR of thyroid cancer was significantly higher in Shanghai than that in Los Angeles, Singapore, and Hong Kong while lower than that in Taiwan and Macau, despite nonsignificant IRRs (Table 1B; Fig. 3).

The incidence of brain and CNS cancer was highest in Shanghai irrespective of sex (Table 1A and B). An approximate 30% to 70% reduction in cancer risk was resultantly seen in other regions (Fig. 3).

To be more precise, we further elucidated the disparities in cancers across regions by age, period, and sex. The results were detailed in Supplementary Tables S2–S3 and Supplementary Figs. S30–S35.

We divided the study period into three nonoverlapped intervals as described in Materials and Methods to investigate the changes of IRRs over time. The IRRs altered with time due to the changes of cancer incidences among regions, especially in cancer experienced significant trend, despite the heterogeneity among age groups. For example, in elderly people (≥65 years), the ASR of colon cancer was nearly three times higher in Los Angeles than in Shanghai in 1983 to 1994 (IRR = 2.87, 95% CI = 2.46–3.32). This regional gap, however, had narrowed considerably in 1995 to 2007 (IRR = 1.29, 95% CI = 1.17–1.42). The gaps for breast cancer and liver cancer in these two regions were narrowed with time as well. Nevertheless, for lung cancer, the regional gaps were widening, with the cancer risk totally reversed in most regions over time. More details can be seen in Supplementary Figs. S36–S43.

This is the first report, to our knowledge, to compare the cancer incidence among Chinese populations living in different regions in the world. The remarkable geographic disparities in cancers among people sharing similar genetic background might suggest the critical roles of environmental factors such as diet, ambient air pollution, and infections for most cancer development (16–19).

Dietary exposures, including foods, individual nutrients, methods of preparation, and habits of consumption (20–22), have been proposed to protect against or increase risk for cancers, especially the gastrointestinal cancers. For instance, esophageal cancer, gastric cancer, and some premalignant conditions of the upper gastrointestinal tract are all negatively associated with fruit and vegetable intake (23, 24). Consumption of red or processed meat has been linked to increased risk of gastrointestinal cancers (25, 26). Subsequently, the dietary patterns were deemed to be significantly associated with cancer risk. For example, epidemiologic data are concordant in suggesting that the Mediterranean Diet (MD) decreases the risk of a variety of cancers, although the underpinning mechanisms are still unclear (27–29). China is vast in territory and varied in dietary patterns across 31 provinces (30, 31). Diet in Shanghai is various but dominated by a traditional southern dietary pattern, characterized by high intakes of rice, fresh leafy vegetables, pork, poultry, sodium and fish/seafood, and low intakes of beef, processed meat, wheat buns/breads, cakes/cookies, deep-fried grains, fruits, milk, and instant noodles (31). On the contrary, American Chinese were prone to Western foods such as butter, lunchmeats, and snack chips (32). The dietary patterns among Chinese people living in Hong Kong, Macau, Singapore, and Taiwan might be the mix of that in Shanghai and Los Angeles (33–35). Diet influences the cancer development through many potential ways such as interaction with gut microbiota via regulation of host metabolism and immune (36, 37) and can explain the majority of the disparities in gastrointestinal cancers as well as breast and prostate cancer across regions. But the underlying mechanism also remains unclear and therefore warrants further investigation.

The relatively higher incidences of stomach and liver cancers in Shanghai were mainly due to the high prevalence of infections, including Helicobacter pylori for stomach cancer and HBV and HCV for liver cancer (38, 39). By the enormous efforts to control infections, striking decreasing trends were found in these two cancers in Shanghai, and the gaps across regions were subsequently narrowed over time. Interestingly, consistent with other Asian American groups, minor increases of liver cancer among American Chinese in Los Angeles were detected in both sexes (40), which underscore the need for improving HBV vaccination rates and HBV and HCV screening in the at-risk population.

Lung cancer, the leading malignancy in men, was mainly ascribed to smoking and ambient air pollution (41). In our study, the highest lung cancer incidence was found in Hong Kong, followed by Singapore, in both men and women, which might be the result of the relatively high smoking prevalence (42, 43). However, a significant decrease in lung cancer incidence was observed in these regions, which might be majorly attributed to the effective control of smoking. Of note is that a slight increase in lung cancer ASR was detected in women in Shanghai, albeit lower smoking rate of 3.5% was found in this population (44). Possible explanations for this increase include: (i) women suffered much more from cooking oil fumes, a defined risk factor for lung cancer (45); and (ii) women benefitted less from smoking control than men.

For cervical cancer, a cancer closely related with human papillomavirus (HPV) infection, the lowest incidence was found in Shanghai compared with other regions, but a significant increasing trend has been seen in Shanghai while decreasing trends seen in other regions. In 2017, Shanghai initiated its HPV vaccination programming; the accumulative protective effect on population level can be expected to emerge in the next few decades (46). For other gynecologic cancers, despite having lower rates compared with Chinese women in other regions, incidence rates were significantly increased in Shanghai women, for whom breast cancer has surpassed lung cancer and ranked the first one in recent years. Known risk factors for gynecologic cancers include obesity, postmenopausal estrogen therapy, nulliparity, early menarche, and late menopause (47). Changes in the prevalence of these risk factors, especially obesity, in Shanghai women may explain some of the observed increases (48, 49).

Of particular concern are cancers having dramatic differences in incidence across populations such as prostate cancer. In this study, we found that the prostate cancer incidence in Chinese people in Los Angeles was nearly half the incidence in non-Hispanic whites in the United States (50), while 5 times higher than that in Shanghai. In addition, prostate cancer has been experiencing an unexpected sharp increase in incidence during the last three decades regardless of region. Consequently, the regional gap between Shanghai and Los Angeles in prostate cancer incidence was narrowed with time in older people (≥65 years), but widened in people aged 35 to 64 years. These results were derivatives of changing cancer incidence and might suggest that the lack of emphasis in cancer prevention among middle-aged Chinese people in Los Angeles and the impending heavy burden posed by prostate cancer in Shanghai in the near future. The factors driving the increase in prostate cancer are not entirely clear; however, they may include gradual implementation of prostate-specific antigen (PSA) screening and improved biopsy techniques or the impact of an increasing western lifestyle (51).

Another noted geographic difference was observed in incidence of oral cavity and pharynx cancer, which was rare in Shanghai and Los Angeles but occurred at relatively high rates in Taiwan, Hong Kong, Singapore, and Macau. We speculated that this disparity was mostly ascribed to the difference in incidence of nasopharyngeal carcinoma, a malignancy related with Epstein–Barr virus (EBV) infection (52). It has been well documented that nasopharyngeal carcinoma commonly occurs in Southern China and Southeast Asia (53). In our study, oral cavity and pharynx cancer experienced a decreasing trend in Singapore and Hong Kong, but an increasing trend in Taiwan, which was calling for further investigation and more effective prevention.

Our findings highlight the substantial impact of environmental factors on cancer development, although germline mutations have emphasized their role in cancer development, such as BRCA1 and BRCA2 in breast cancer (54–56). However, such mutations are rare occurrence among populations, and taken together account for a very small proportion of cancer cases. On the other hand, growing evidence suggested that the oncogenesis was mainly driven by the somatic mutations in driver genes (47), but how environmental exposures influence these critically carcinogenic mutations is still not entirely clear.

There are several caveats worth noting when interpreting these results. First, data used in this study lacked histology information and thus cannot compare the cancer incidence by its histology, such as the differences in esophageal adenocarcinoma and esophageal squamous cell carcinoma, respectively. Second, no information about the immigration status, an indicator for exposure time, was available. Third, no personal data were available to investigate the association between risk factors and cancer risk at the individual level. Finally, it is hard to measure the equivalence in cancer diagnostic criteria across regions due to the changing proportion of different diagnostic criterion over time. All limitations listed here might introduce bias into our results, and therefore one should be cautious while interpreting them.

In summary, in this study, we first reported the geographical disparities of cancer incidence among Chinese populations. The cancer profiles and cancer risks were distinct across regions, which further demonstrated the significant impact of environmental exposures on cancer development, and called for more targeted and precise strategies of cancer prevention for Chinese people living in different regions. For instance, much more priority should be placed on prostate cancer, especially in the Chinese population in Los Angeles, which indicates that the PSA screening should be widely adopted in these populations. Moreover, recommended interventions at the population level include policy regulation to decrease cancer risks, such as HBV vaccination among populations in Los Angeles and EBV monitoring among people in Singapore and Taiwan.

No potential conflicts of interest were disclosed.

Conception and design: C. Suo, T. Zhang, X. Chen

Development of methodology: Z. Liu, O. Shi, X. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Z. Liu, O. Shi, N. Cai, Y. Jiang, K. Zhang, Z. Zhu, H. Yuan, Q. Fang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Z. Liu, O. Shi, H. Yuan, Q. Fang, C. Suo, T. Zhang

Writing, review, and/or revision of the manuscript: Z. Liu, O. Shi, Y. Jiang, S. Franceschi, T. Zhang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z. Liu, H. Yuan, Q. Fang

Study supervision: C. Suo, T. Zhang, X. Chen

This work was supported by the National Natural Science Foundation of China (grant numbers: 81772170 and 81502870); the National Key Research and Development Program of China (grant numbers: 2017YFC0907002, 2017YFC0907501, and 2017YFC211700); the Key Basic Research Grants from Science and Technology Commission of Shanghai Municipality (grant number: 16JC1400500); the International S&T Cooperation Program of China (grant number: 2015DFE32790); and Shanghai Municipal Science and Technology Major Project (2017SHZDZX01).

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.