Objective Globally, gastric cancer incidence shows remarkable international variation and demonstrates distinct characteristics by the two major topographical subsites, cardia (CGC) and non-cardia (NCGC). Because global incidence estimates by subsite are lacking, we aimed to describe the worldwide incidence patterns of CGC and NCGC separately.
Design Using Cancer Incidence in Five Continents Volume X (CI5X), we ascertained the proportions of CGC and NCGC by country, sex and age group (<65 and ≥65 years). These derived proportions were applied to GLOBOCAN 2012 data to estimate country-specific age-standardised CGC and NCGC incidence rates (ASR). Regional proportions were used to estimate rates for countries not included in CI5X.
Results According to our estimates, in 2012, there were 260 000 cases of CGC (ASR 3.3 per 100 000) and 691 000 cases of NCGC (ASR 8.8) worldwide. The highest regional rates of both gastric cancer subsites were in Eastern/Southeastern Asia (in men, ASRs: 8.7 and 21.7 for CGC and NCGC, respectively). In most countries NCGC occurred more frequently than CGC with an average ratio of 2:1; however, in some populations where NCGC incidence rates were lower than the global average, CGC rates were similar or higher than NCGC rates. Men had higher rates than women for both subsites but particularly for CGC (male-to-female ratio 3:1).
Conclusions This study has, for the first time, quantified global incidence patterns of CGC and NCGC providing new insights into the global burden of these cancers. Country-specific estimates are provided; however, these should be interpreted with caution. This work will support future investigations across populations.
- CANCER EPIDEMIOLOGY
- GASTRIC CANCER
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Gastric cancer is a major contributor to the global cancer burden, ranking as the fifth most common malignancy worldwide in 2012.
The two major topographical subsites of gastric cancer, cardia and non-cardia, display distinct characteristics in their descriptive epidemiology and risk factor profiles.
Cardia and non-cardia gastric cancer incidence rates have been reported for some countries, with cardia rates typically lower than non-cardia rates.
What are the new findings?
The highest estimated regional age-standardised incidence rates of both gastric cancer subsites occur in Eastern/Southeastern Asia.
For country-specific estimates, cardia gastric cancer rates were highest in countries within Central Asia; non-cardia rates were highest in countries within Eastern/Southeastern Asia.
Men had higher estimated rates than women, particularly for cardia gastric cancer.
For several populations, for example, for men in Australia, the USA and the UK, estimated cardia gastric cancer rates were comparable with or exceeded estimated non-cardia gastric cancer rates.
How might it impact on clinical practice in the foreseeable future?
This work provides further insights into the global burden of gastric cancer and will aid planning and decision making related to gastric cancer control strategies.
Gastric cancer is a major contributor to the global cancer burden. In 2012, there were an estimated 952 000 new cases worldwide, making gastric cancer the fifth most common malignancy after lung, breast, colorectal and prostate cancer.1 There is considerable variation in gastric cancer incidence by sex and geography. On average, gastric cancer incidence rates in men are twice those observed in women. Geographically, almost two-thirds of all gastric cancer cases occur in Asia, primarily concentrated in China (43% of total global cases).1 While incidence rates remain high in countries within Eastern/Southeastern and Central Asia, Eastern Europe and parts of Central and Southern America, gastric cancer incidence rates are comparatively low in many developed regions such as North America and Western Europe.
Although often reported as a single entity, gastric cancers can generally be classified into two topographical categories: cardia gastric cancer (CGC) arising in the area of the stomach adjoining the oesophageal-gastric junction, and non-CGC (NCGC) arising from more distal regions of the stomach. While both CGC and NCGC are thought to be influenced by cigarette smoking2–4 and possibly by low intake of fruits and vegetables,5–7 there are other risk factors that differ between the two subsites. Risk factors for CGC are similar to those for oesophageal adenocarcinoma and include obesity8 ,9 and GORD.10 ,11 Barrett's oesophagus, a metaplastic condition that can result from GORD, is also thought to be a risk factor for CGC, though not in all populations.12 NCGC, however, is strongly associated with Helicobacter pylori infection.13 ,14 Some evidence suggests that H pylori infection may be inversely associated with both oesophageal adenocarcinoma and CGC,15 ,16 though studies in some populations have suggested a positive association between the bacterium and CGC.12 ,17 ,18 The influence of socioeconomic status (SES) also differs. While indicators of low SES such as household crowding, low income, low education and increased number of siblings are positively associated with NCGC, they do not appear to be associated with CGC.19 ,20 Given these differences, there is increasing interest in describing the worldwide burden of CGC and NCGC subsites separately.
CGC and NCGC incidence rates have been reported separately for some countries, with CGC rates typically lower than NCGC rates.21–23 However, there are currently no global estimates available by gastric cancer subsite and estimates are lacking for most countries. We aim here to describe the global incidence patterns of gastric cancer for each of the two major topographical subsites. Based on data from Cancer Incidence in Five Continents Volume X (CI5X)24 and GLOBOCAN 2012,1 using methods similar to those employed to estimate global oesophageal cancer incidence by histological subsite,25 we report age-standardised incidence rate (ASR) estimates for CGC and NCGC in 2012 by country, region and Human Development Index (HDI).26
We used CI5X, containing cancer incidence data from 290 cancer registries in 68 countries for the period 2003–2007.24 The CI5X results are presented according to the International Classification of Diseases, 10th Revision.
We first calculated the proportions of CGC (C16.0) cases and NCGC (C16.1-6) cases out of all gastric cancer cases with known topography (C16.0-6) for each country with data included in CI5X (n=53) and stratified these calculations by sex and age group (<65 and ≥65 years). Where data were available from multiple cancer registries within one country, these were aggregated without weighting to obtain national estimates of the subsite proportions. If a dataset for a country contained fewer than two cases of CGC or NCGC within any sex and age group stratification (n=15 countries), a second method was employed (see below).
In the second step, we estimated the proportion of CGC and NCGC for countries with insufficient subgroup data (small numbers) or those without CI5X data (n=128 countries in total). For these, regional proportions based on non-weighted, aggregated country-level data were calculated to estimate country-specific subsite proportions. To estimate regional proportions, all CI5X data were included: this included all countries with a low number of cases within subgroup stratifications based on sex and age group. Regional subsite proportions were calculated based on UN geographical areas27 as was done by Arnold et al25: Sub-Saharan Africa, Northern Africa and Western Asia, Central Asia (including India), Eastern/Southeastern Asia (including China), Central/Southern America and the Caribbean, Northern America, Eastern Europe, Northern and Western Europe, Southern Europe and Oceania (see online supplementary annex I for a listing of countries and corresponding regions).
For the available datasets in the Sub-Saharan Africa region (three registries: PROMEC in South Africa, Harare in Zimbabwe and Kyadondo County in Uganda), approximately 94% of CI5X gastric cancer cases were classified as ‘unspecified’ (C16.8-9, herein referred to as not otherwise specified ‘NOS’) and there were fewer than five cases of CGC or NCGC for any sex and age group stratification. Without stratification, the proportion of CGC and NCGC was approximately 50:50; however, this was based on fewer than 15 cases each. In view of the high burden of cancer attributable to infection in Sub-Saharan Africa,28–30 the regional CGC/NCGC proportions for Sub-Saharan Africa were calculated by redistributing all cases classified as NOS to NCGC cases. These regional proportions were then applied to all countries in this region (47 countries).
The sex-specific and age-specific national or regional gastric cancer subsite proportions estimated using the above methods were then applied to the 2012 gastric cancer incidence estimates for each corresponding country (184 countries) in GLOBOCAN 2012.1 Estimated ASRs (per 100 000 person-years) for CGC and NCGC in both men and women were calculated using the World Standard Population as defined by Segi.31 In addition to national and global estimates of gastric cancer by subsite, we also estimated subsite-specific rates by world region and by HDI category (low, medium, high or very high).26
In the CI5X data used in this analysis, the proportion of gastric cancer cases categorised as NOS (C16.8-9) ranged from 28% to 98% across countries. As a result, sensitivity analyses were performed to evaluate a variety of potential methods to identify the country-specific and sex-specific proportions of CGC and NCGC. For a selection of CI5X countries, the proportions of CGC and NCGC were calculated by redistributing NOS cases under six different scenarios. We also evaluated the influence of these different scenarios on the regional proportions of CGC and NCGC. Our preferred method, applied here, falls in the middle of the sensitivity analysis results (see online supplementary annex II for more information about the methods used and results obtained).
In 2012, according to our estimates, there were 952 000 cases of gastric cancer globally. Two-thirds of these cases were in men (631 000) and one third in women (320 000), corresponding to global gastric cancer ASRs of 17.4 and 7.5 per 100 000 respectively.
Cardia gastric cancer
According to our estimates, there were 260 000 CGC cases in 2012 worldwide, comprising 27% of total gastric cancer cases (192 000 cases, 30% of the global total in men and 69 000 cases, 21% of the global total in women). Geographically, the majority of these cases occurred in Eastern/Southeastern Asia (152 000 cases, 59%), followed by Central Asia (39 000, 15%) and Central/Southern America and the Caribbean (16 000, 6%). More than half of the total cases occurred in China (135 000, 52%; figure 1).
In 2012, the estimated global CGC ASR was 3.3 per 100 000. Rates in men were on average three times the rates in women (ASRs: 5.3 per 100 000 and 1.6 per 100 000, respectively); the male-to-female ratio ranged from 1.5 in Sub-Saharan Africa to 3.9 in Northern America (figure 2). The highest regional rates of CGC occurred in Eastern/Southeastern Asia and Eastern Europe; the lowest regional rates occurred in Sub-Saharan Africa, followed by Northern Africa and Western Asia, and Northern America for men and women, respectively (figure 2). A slight gradient across HDI categories was observed when China was excluded for both men and women, with the highest rates found in high HDI countries (ASRs: 4.6 per 100 000 and 1.6 per 100 000, respectively; figure 3).
At the country level, the highest estimated rates of CGC occurred in Central Asian countries: in men, ASRs were 14.7 per 100 000 in Kyrgyzstan and 14.6 per 100 000 in Kazakhstan. High rates in men also occurred in Iran and Tajikistan: ASRs of 13.4 and 12.6 per 100 000, respectively. Similarly, in women, countries with the highest rates included Tajikistan (ASR: 6.1 per 100 000), Turkmenistan (ASR: 5.5 per 100 000), Iran (ASR: 5.4 per 100 000) and Kazakhstan (ASR: 5.2 per 100 000). Amongst non-Sub-Saharan African countries with >10 cases of CGC in both sexes, the lowest rates were observed in Eastern/Southeastern Asian countries such as Indonesia and Cambodia (figures 4A, B; see online supplementary annex I for a full list of country-specific rates).
Non-cardia gastric cancer
According to our estimates, there were 691 000 NCGC cases in 2012, comprising 73% of total gastric cancer cases worldwide (70% of the global total in men, 440 000 cases and 79% of the global total in women, 252 000 cases). Almost two-thirds of all cases occurred in Eastern/Southeastern Asia (434 000, 63%): more than half of these were in China (270 000, 39% of the global total). Central Asia (57 000 cases) and Eastern Europe (55 000) followed with approximately 8% of global cases each (figure 1).
Globally, the estimated ASR of NCGC was 8.8 per 100 000. Rates in men were double when compared with those in women (ASRs 12.1 per 100 000 and 5.9 per 100 000, respectively); the male-to-female ratio ranged from1.3 in Oceania to 2.3 in Eastern/Southeastern Asia (figure 2). As with CGC, incidence was highest in Eastern/Southeastern Asia with ASRs of 21.7 per 100 000 and 8.7 per 100 000 for men and women, respectively. The lowest regional rates of NCGC occurred in Northern America and Oceania. A gradient across HDI categories was observed when China was excluded, with the highest rates of NCGC in very high HDI countries for men (ASR 12.5 per 100 000; figure 4). Although not as distinct a gradient, countries in higher HDI categories also had higher rates in women.
Differences also existed across countries. The highest NCGC rates in men occurred in countries located within Eastern/Southeastern Asia, including the Republic of Korea (ASR 58.6 per 100 000), Mongolia (ASR 42.1 per 100 000), and Japan (ASR 41.0 per 100 000). Similarly, in women the highest rates occurred in the Republic of Korea (ASR 23.6 per 100 000) and Mongolia (ASR 18.5 per 100 000); however, a high rate of NCGC among women also occurred in Guatemala (ASR 19.5 per 100 000). Among countries with >10 cases of NCGC in both sexes in 2012, the lowest rates were observed primarily in African countries (figure 4C, D; see online supplementary annex I for a full list of country-specific rates).
Country-specific relative incidence rates of CGC and NCGC
In most countries, CGC incidence rates were lower than NCGC rates (91% or 167 of 184 GLOBOCAN countries, see online supplementary annex table A1). There were several countries, however, where CGC rates were similar to or higher than NCGC rates. For men, these included countries located in Oceania (n=6) and countries within the regions of Northern/Western Europe (n=4) such as Denmark and Belgium and Southern Europe (n=2) such as Malta. This was also true in Poland and Uruguay. Additionally, there were three countries where CGC rates were higher than NCGC rates for both men and women: Iran, Finland and Serbia.
In this study, we estimated global, regional-specific and country-specific incidence rates of gastric cancer by subsite for the first time. Estimated rates varied across regions and countries. The highest estimated regional rates of both CGC and NCGC occurred in Eastern/Southeastern Asia: in both men and women, incidence rates were 47%–79% higher than the global average. For country-specific estimates, the highest estimated rates of CGC occurred in countries within Central Asia, such as Iran; estimated NCGC rates were highest in countries within Eastern/Southeastern Asia, such as the Republic of Korea. Although CGC rates were typically lower than NCGC rates, for several populations, CGC rates were similar to or higher than NCGC rates: for example, for men in Australia, the USA and the UK, CGC rates were comparable with NCGC rates. Incidence rates also varied by sex; in particular, for CGC, rates in men were three times the rates in women.
Geographic patterns of CGC reported here confirm previous descriptions of elevated rates in some populations. At the country level, the highest CGC rates occurred in countries within Central Asia. In areas within Iran, CGC has been reported to comprise between 40% and 50% of gastric cancer cases,22 ,32 ,33 with incidence rates approximately five times our estimate of the global average.22 Similarly, for men and women in Iran, our estimates of CGC incidence rates were 2.5 times and 3.4 times the global average, respectively. High incidence rates also occurred in other countries within this area. In men, CGC rates were highest in Kyrgyzstan and Kazakhstan, both with rates approximately 2.8 times the global average; in women, the highest rates were in Tajikistan (3.8 times the global average). The reason for a higher incidence of CGCs in countries within this geographic area in Central Asia is unknown; however, it has been noted that high-risk populations may have a high prevalence of certain risk factors for CGC such as obesity and GORD.33 ,34 Furthermore, given a high prevalence of H pylori infection in countries within Central Asia,35 ,36 and the disparate influence of risk factors for CGC across some populations,12 ,17 ,18 it is also possible that H pylori infection may contribute to high rates of CGC in these populations.
In the case of NCGC estimates, the highest rates occurred in the Republic of Korea, Mongolia and Japan. Given that approximately 90% of gastric cancer cases in these populations are NCGC, these results confirm high rates of overall gastric cancer as previously reported.20 ,37 In these countries, for men, estimated NCGC rates ranged from 3.4 to 4.8 times the global average; for women, this range was 2.6–4.0. Environmental factors may play a role: elevated incidence rates may result from a high prevalence of H pylori infection and, in particular, the presence of highly virulent strains of the bacterium in Eastern/Southeastern Asian populations.38 A high prevalence of risk factors such as the consumption of foods high in salt may also contribute to increased risk of NCGC in Asian populations such as Japan.39–41 Furthermore, a high incidence of NCGC in countries such as the Republic of Korea and Japan may be partly due to diagnostic practices that detect small lesions not typically identified as cancerous elsewhere.42
Differences in the relative frequency of CGC and NCGC have also been observed across populations: a recent report noted that CGC is more common in countries with lower overall gastric cancer incidence.43 Here, in some geographic areas where NCGC incidence rates were lower than the global average (for men, on average four times lower), we found a greater proportion of CGC cases relative to NCGC cases. This was particularly true for men in Iran where the CGC rate was double the NCGC rate, unlike what is observed at a global level, where CGC rates were two times lower than NCGC rates. Similar or higher rates of CGC relative to NCGC in some countries may reflect a decreased prevalence of risk factors for NCGC such as H pylori infection and smoking, and increased prevalence of risk factors that contribute to the development of CGC such as obesity.44 ,45 This may also indicate improvements in gastric cancer subsite recording in some populations: in the USA, for example, the topographical classification of gastric cancer cases to an unspecified location decreased between 1974 and 1998 while those categorised as cardia increased.46 Similarly, these results may reflect an overestimation of reported CGC cases as some of these cases may be misclassified as adenocarcinomas of the oesophagus.47 ,48
Differential trends in the incidence of NCGC and CGC have also been reported. In the USA and some European countries such as the Netherlands, NCGC incidence rates have decreased annually since the mid-1970s, while CGC rates have remained stable or increased.21 ,23 ,49 Although the estimation of temporal trends was beyond the scope of the current work, future studies that estimate time trends in global incidence rates by gastric cancer subsite would be greatly beneficial.
Differences in the incidence rates of gastric cancer subsites by sex were also observed: CGC in men was approximately three times the rate in females, whereas for NCGC, rates in men were twice as high.50 The reasons for higher rates of the two subsites of gastric cancer in men are poorly understood. Variations by sex may reflect disparate acquisition of risk factors such as H pylori infection and Barrett's oesophagus.51 ,52 Higher rates in men may also result from differences in cigarette smoking behaviour, though work by Freedman et al53 suggests that smoking might not be an explanatory factor. Furthermore, variations by sex may also reflect the protective effect of oestrogen in women54 or a lag in age of intestinal gastric cancer onset in women compared with that in men.55 The potential contribution of these factors, however, to an increased risk in men and to the differential rates by subsite is unclear.
Along with variations by geography and sex, CGC and NCGC incidence rates varied by HDI. Although 58% of all gastric cancer cases occurred in countries with a low or medium HDI, countries with some of the highest rates such as the Republic of Korea and Japan are within the very high HDI category. This in turn elevates rates in this HDI group for both CGC and NCGC. The resulting apparent association between higher HDI and NCGC is likely to reflect an ecological bias: previous studies have indicated an association between low SES and NCGC at the individual level.15 ,16 Similarly, high-risk groups within low-risk countries may not be represented and the region-level and country-level estimates may mask this within-country variation. Risk factors for NCGC such as low SES and H pylori infection status may vary across populations within countries.56 ,57 Likewise, there are variations in both CGC and NCGC rates across ethnic groups: ethnic minorities such as black[S2] or Asian Americans populations have higher rates of NCGC and lower rates of CGC compared with US white populations,58 ,59 a pattern also observed in Indigenous populations worldwide compared with non-Indigenous counterparts.60
The estimation of CGC and NCGC incidence was limited by the extent to which topographical information identifying the exact anatomical location of gastric cancer cases was reported by each registry. In the CI5X data used in the current analysis, an average of 41% of gastric cancer cases were categorised as NOS (C16.8-9); however, this ranged from 28% to 98% across countries. Difficulties distinguishing between CGC and NCGC may result from the necessity to rely on surgical judgement for these classifications and, in some cases, it is difficult to determine whether a tumour originated in the stomach or in the oesophagus.47 ,48 Although some of these challenges are likely to persist, it would be valuable to further understand the barriers and potential solutions to identifying or reporting the topography of gastric cancer cases in order to reduce under-reporting and misclassification by gastric subsite.
There were additional data limitations that should be considered as sources of uncertainty regarding the accuracy of country-specific estimates. First, GLOBOCAN numbers of incident gastric cancer cases used to calculate rates are based on estimates.1 Furthermore, population-based cancer registry data were used within the analysis. Although of high quality, data aggregated from multiple cancer registries within one country as a proxy of national data may inadequately represent country-level rates. The use of these data also limited the number of countries that could be included directly in the derivation of proportions. Wherever country-specific proportion estimations were not possible, the information used to calculate country rates was based on regional information; as such, heterogeneity across those countries may not be represented. In particular, proportions for countries in Sub-Saharan Africa were calculated based on regional information only. Moreover, given the generally low burden of gastric cancer reported from registries in this region, all NOS cases were reclassified as NCGC which may overestimate the NCGC rates in Sub-Saharan Africa.
While we believe the estimates presented in this paper are sufficiently reliable to provide an indication of the burden of gastric cancer subsites at the global and regional level, because of the data deficiencies, limitations and assumptions made for individual countries, we think it is prudent to regard the country-level CGC and NCGC incidence rate estimates presented in online supplementary annex I as rough approximations, especially for those countries for which the proportion of CGC/NCGC was based on regional estimates.
This is the first comprehensive attempt to estimate global incidence patterns of gastric cancer subsites. While the highest estimated regional rates for both subsites occurred in Eastern/Southeastern Asia, country-specific estimates identified high rates of CGC in countries such as Iran located within Central Asia, and high rates of NCGC in countries within Eastern/Southeastern Asia such as the Republic of Korea. Furthermore, while CGC rates were typically lower than NCGC, for several countries such as Australia, the USA and the UK, the rates of CGC were comparable or higher. The CGC and NCGC incidence estimates reported here provide new insights into the global burden of gastric cancer and support future ecological investigations across populations. This work will also aid in planning and decision making related to gastric cancer control strategies.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors Study concept and design and drafting the manuscript: AC, MA, DF and IS. Analysis and interpretation of data: AC, MA, JF, KJG, DF and IS. Critical revision of the manuscript for important intellectual content: JF, KJG, DF and IS. Study supervision: DF and IS.
Funding AC is supported by a Vanier Canada Graduate Scholarship and a Michael Smith Foreign Study Supplement. KJG is a Health Senior Scholar supported by Alberta Innovates Health Solutions.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The estimated age-standardized incidence rates per 100 000 person-years and estimated number of cases of cardia and non-cardia gastric cancer by country and sex in 2012 are available online through this journal as an annex to this work.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.