Ethnic Disparities in Gastric Cancer Presentation and Screening Practice in the United States: Analysis of 1997–2010 Surveillance, Epidemiology, and End Results-Medicare Data

This article is featured in Highlights of This Issue, p. 627

Gastric cancer is the fifth most diagnosed cancer and the third most common cancer-related death worldwide, with more than one million cases diagnosed and 700,000 deaths (1). In 2018, of the 1.7 million new cancer cases reported in the United States (U.S.), 26,240 were gastric cancer (2). Among the U.S. elderly (those aged 65 years or older), cancer has an even larger impact as more than half of all new cancer cases and two-thirds of all cancer deaths occur in this group (3). More specifically, of the incident gastric cancer cases in the United States, 60% are diagnosed among individuals over age 65 years, representing 66% of all gastric cancer deaths in the United States (4). The median age at diagnosis is 68 years (4) and the highest incidence occurs at approximately 70 years of age (5).

Globally, more than 50% of all gastric cancer cases are diagnosed in Asia, specifically Japan, South Korea, and China (6). Similarly, within the United States, gastric cancer incidence is highest among Asian/Pacific Islanders (6), a minority group that has seen more than 400% population growth in the United States in the past 30 years (7). Within this Asian American population, Americans of East Asian and Southeast Asian (e.g., Vietnamese) descent have higher incidence and mortality rates when compared with Americans of non-Hispanic white (NHW) origin (8). In addition, in the United States, Hispanics and African Americans are disproportionately affected compared with Caucasians (9).

Helicobacter pylori (H. pylori) infection is a risk factor for gastric cancer, with chronic infection often resulting in chronic gastritis that can lead to gastric atrophy, which in turn leads to intestinal metaplasia, dysplasia, and gastric cancer (10–12). It has been estimated that 89% of noncardia gastric cancer (ncGC) cases are attributable to H. pylori infection (13). Although the prevalence of H. pylori infection is lower in the United States than in other places, such as Asia and South America (14), Hispanic Americans and African Americans experience a higher prevalence than NHWs (14). Thus, the American College of Gastroenterology (14, 15) recommends testing for H. pylori infection in those at higher risk such as those taking long-term low-dose aspirin or initiating NSAID treatment, and in those with gastric cancer–related conditions such as peptic ulcer (to include gastric or duodenal ulcer) or a history of peptic ulcer (10–12).

In this article, we describe ethnic variations in gastric cancer presentation among the elderly population (65 years and older) whose data are in the population-based Surveillance, Epidemiology and End Results (SEER)-Medicare database. Lui and colleagues identified gastric cancer differences in a general SEER population by ethnicity in incidence, stage of disease, and survival (16). Here, we extend this work with more recent data and describe differences in this population in screening services (H. pylori testing and endoscopy) by race/ethnicity, place of birth, and history of gastric cancer–related conditions.

We used the NCI's population-based SEER-Medicare cancer database for gastric cancer. This database included SEER identifiers for gastric cancer cases matched with identifiers contained in Medicare's master enrollment file (for detailed description of this matching process, see Warren and colleagues; ref. 17). The linked database included medical outcomes of gastric cancer cases among the elderly (65 or older) who received fee-for-service Medicare benefits, as well as information regarding costs and utilization (3). In addition, the GLOBOCAN 2012 Fact Sheet (1) (accessed October 1, 2017 at: http://globocan.iarc.fr/old/FactSheets/cancers/stomach-new.asp) and the SEER Stat Facts (4) (accessed October 1, 2017 at: https://seer.cancer.gov/statfacts/html/stomach.html) websites were used to obtain updated incidence and mortality data.

The study sample comprised of the SEER data file and the patient entitlement and diagnosis summary file (PEDSF), which included patients with a diagnosis of gastric cancer from 1997 to 2010 (n = 48,377) for whom Medicare data were available. We excluded those whose race was marked as either American Indian or Other/Unknown (n = 383) due to the small sample size (0.8% of the sample). These SEER records for the 47,994 gastric cancer cases were then merged with the inpatient hospital (Medicare Provider Analysis and Review, MEDPAR) and physician/outpatient services (National Claims History, NCH) claims to examine healthcare utilization.

We extracted demographic information: age at diagnosis, sex, marital status, place of residence, income, place of birth, SEER region, and race/ethnicity. Race/ethnicity was grouped into four categories: NHW, Asian American/Pacific Islander, Hispanic, and black. The SEER regions included the following: the Northwest (Connecticut, New Jersey), South (Kentucky, Louisiana, Georgia), Midwest (Detroit, Iowa) and West (Hawaii, New Mexico, Seattle, Utah, California). Place of birth was categorized as U.S.- or foreign-born or unknown; those born in any U.S. territory (e.g., Puerto Rico, Guam) were classified as U.S.-born. Income was based on census tract of residence using the 2000 Census from the PEDSF, and classified by quartiles (<$35,149; $35,150-$43,348; $46,349-$60,306; >$60,307) as previously classified in other SEER-Medicare analyses (18). Place of residence was classified as urban or rural; urban was defined as counties in metro areas (from fewer than 250,000 population to more than 1 million as described in the PEDSF data dictionary). Marital status was classified as married or not married.

Disease presentation factors (location and staging) were assessed. The location of the gastric cancer tumor was classified as cardia or noncardia; noncardia was further broken down into body and fundus, pyloric antrum, pylorus, lesser and greater curvature, and other (classified as other, specified sites of stomach and stomach unspecified) using ICD-9 codes in the dataset. Stage of tumor at diagnosis was classified as unstaged, in situ, local, regional, and distant using the historical SEER staging system (7). Those with no information on location and stage of tumor were classified as missing.

Finally, from the inpatient hospital and physician/outpatient services claims, we obtained information on presence of various gastric cancer–related conditions (i.e., peptic ulcer [ICD533], gastric ulcer [ICD531], duodenal ulcer [ICD532], gastrojejunal ulcer [ICD534], gastritis [ICD535], disorders of function, such as achlorhydria, persistent vomiting, and gastroparesis [ICD536], disorders-other, such as gastric diverticulum, chronic duodenal ileus, and gastroptosis [ICD537]), and screening services received [H. pylori testing (urease activity in blood: CPT83009; or breath: CPT78267, CPT78268, CPT83013, CPT83014; or antibody testing: CPT86677; or enzyme immunoassay: CPT87338 or CPT87339) and endoscopy (Berenson-Eggers Type of Service (BETOS) code: P8B—upper gastrointestinal endoscopy)]. The dates for these conditions' claims were queried to investigate when they were filed in relation to the gastric cancer diagnosis occurring between 1997 and 2010. The gastric cancer diagnosis dates were subtracted from the claims dates. The median time in months was calculated with a negative value indicative of claim being filed before the diagnosis, and a positive value indicative of claim being filed after the diagnosis.

Frequencies of characteristics of cases were calculated for various subgroups. Two-sided χ² tests (and two-sided ANOVA test on age at diagnosis) were used to assess differences in demographic and disease presentation factors, and to assess differences in gastric cancer–related conditions and screening (H. pylori testing, endoscopy) across race/ethnicities. Kruskal–Wallis tests were used to assess differences in the median number of months between date of claim and date of diagnosis across race/ethnicities.

Of the 47,994 eligible gastric cancer cases, 62% were NHW (n = 29,614), 13% were API (n = 6,240), 12% were Hispanic (n = 5,630), and 14% were black (n = 6,510). These records were merged with the inpatient hospital and physician/outpatient services claims. Among the 42,952 unique subjects with claims filed, 62% were NHW (n = 26,707), 13% were API (n = 5,529), 11% were Hispanic (n = 4,783), and 14% were black (n = 5,933).

Demographic data and clinical characteristics are summarized in Table 1. The mean age at diagnosis for all cases was 74.1 years with NHWs the oldest (74.6 years ± 10.5), and Hispanic and black cases the youngest (both 72.5 years). Overall, 60% of the sample was male and 93% lived in an urban community. Within this sample, the highest percentages of subjects in the lowest-income quartile were among black and Hispanic cases (57.6% and 39.8%, respectively), almost double NHW and API (22% in the lowest-income quartile). Over 60% of NHW and black subjects were born in the United States (64.4% and 72.9%, respectively), whereas Hispanics were almost evenly split (34% U.S.-born vs. 32% foreign-born) and the majority of APIs were foreign-born (48%). Across all races/ethnicities, over 24% had an unknown place of birth, with Hispanics having the highest percentage at 34%. The above observed differences across races/ethnicities were statistically significant (P < 0.001; Table 1).

Cardia was the most diagnosed specific cancer site among all gastric cancer cases (26.5%), although there was variation by race/ethnicity. For NHWs, 35.6% of tumors were diagnosed in the cardia compared with 10% for APIs, 15% for Hispanics, and 11% for blacks. The most common site among APIs was the pyloric antrum (31.4%), whereas the other/unspecified stomach sites were more frequently reported among Hispanics and blacks (26.6% and 29.3%, respectively), followed by the pyloric antrum (22.8% and 25.8%, respectively; Table 1). For stage at diagnosis, over 55% of NHWs, Hispanics, and blacks were diagnosed at the regional or distant stage, compared with 63% of APIs diagnosed at the local or regional stage. While only about 30% of ncGC was diagnosed while still localized (31% in APIs, 29% in blacks and NHWs, and 27% in Hispanics), almost also 30% was not diagnosed until distant spread occurred (30% in NHWs, 29% in blacks and Hispanics, and 25% in APIs). Furthermore, when APIs and Hispanics were stratified by place of birth, foreign-born cases had a higher percentage of gastric cancer diagnosed in noncardia as compared with U.S.-born cases (90.9% vs. 87.7% in APIs, respectively, and 86.2% vs. 82.7% in Hispanics, respectively; Table 2). The majority of both foreign- and U.S.-born API cases had the ncGC diagnosed at regional stage (33% for both groups) compared with 35% for foreign-born Hispanic cases; majority of U.S.-born Hispanic cases were diagnosed at the distant stage (30%). The differences across the races/ethnicities for the disease presentation factors were statistically significant (P < 0.001).

The most common gastric cancer–related conditions are summarized in Table 3. In the total sample, 46.1% of cases had a claim related to gastritis, followed by gastric ulcer (27.8%), with similar ordering when stratified by race/ethnicity. In addition, across all races/ethnicities, over 20% had a “disorder other”; the ICD9 code encompassed other disorders of stomach and duodenum such as gastric diverticulum, gastroptosis, and pylorospasm. Within those with an established gastric condition of a peptic ulcer, gastric ulcer, duodenal ulcer, or a history of gastritis (over 55% across all cases), APIs had the highest percentage of H. pylori testing (22.8%) and black cases had the lowest (6.5%). Within API cases, 72% of foreign-born cases had an established condition compared with 50% of U.S.-born cases (Supplementary Table S2). Furthermore, foreign-born cases had higher frequency of H. pylori testing (26%). Similarly, for Hispanics, foreign-born cases had a higher percentage of established conditions (62% vs. 52% in U.S.-born), with those foreign-born cases having double the H. pylori testing (24% vs. 12% in U.S.-born). Endoscopy percentages within those with established conditions ranged from 86.2% in Hispanics to 90.9% in NHWs (Table 3). Overall, within those with an established condition, the H. pylori testing occurred 4.9 (33.0) months before the gastric cancer diagnosis, while the endoscopy occurred 0.3 (2.10) months after the diagnosis (Supplementary Table S3). A similar pattern was observed across all races/ethnicities with APIs having H. pylori testing occurring the furthest out at 7.10 (38.9) months before diagnosis and black cases having the testing occur 3.15 (30.7) months before diagnosis. Within U.S.-born subjects, APIs had the highest H. pylori testing frequency (7.5%) followed by Hispanics (7.3%). Among foreign-born cases, the testing frequency was highest for APIs (19.7%), followed by 16.6% for Hispanics, 14.3% for NHWs, and 7.3% for blacks. Within those with an unknown place of birth, APIs also had the highest testing frequency (16.8%) followed by Hispanics (11.6%; Fig. 1). Comparing U.S.-born with foreign-born cases, endoscopy testing was generally higher in foreign-born than U.S.-born across race/ethnic groups (aside from blacks) with the biggest difference for APIs (75% in foreign-born vs. 56% in U.S.-born). The differences across races/ethnicities for these conditions were statistically significant (P < 0.001).

Currently, gastric cancer is the fifth most diagnosed cancer in the world and the third most common cause of cancer-related deaths (1). Although the incidence of gastric cancer has decreased over time, racial and ethnic differences in both incidence and survival have remained throughout the United States. For example, even though Asians have the highest gastric cancer incidence, survival in this group is the highest overall (16). Although previous work (16) identified gastric cancer differences by ethnicity in incidence, stage of disease and survival, we expanded the work by using more recent data and describing gastric cancer differences in screening H. pylori (testing and endoscopy) among the SEER-Medicare population by ethnicity, place of birth, and gastric cancer–related conditions.

Within this SEER-Medicare gastric cancer population, the average age at diagnosis was 74 years old, similar to other studies (5). Of note, Hispanic and black cases were diagnosed at a younger age. The majority of NHWs, Hispanics, and blacks were U.S.-born, in contrast to APIs. Foreign-born NHWs were diagnosed at the oldest age, while foreign-born Hispanic and black cases were diagnosed at younger ages. The majority of the cases was male, as seen in global ncGC rates where the male-to-female ratio is approximately 2:1 (19). The majority of Hispanic and black cases presented in the lowest-income quartile, which is of interest as lower socioeconomic status has been associated with lifestyle behaviors that increase cancer risk (20).

We observed notable racial and ethnic variation in both screening practices and disease presentation. ncGC is strongly associated with H. pylori infection (21). Although, overall, the majority of all cases were diagnosed in the noncardia area, there were important variations by race/ethnicity. Over 35% of NHWs were diagnosed in the cardia region versus less than 15% for blacks, Hispanics, and APIs. For Hispanics, the majority of cases was diagnosed as either regional tumor, meaning the cancer extended beyond organ of origin into surrounding organs, or with distant tumor (metastatic disease). When stratified by place of birth, the majority of foreign-born Hispanic noncardia tumor cases were diagnosed at regional stage, whereas U.S.-born Hispanics were diagnosed as distant. On the other hand, majority of NHW and black cases were diagnosed as either local or distant, and majority of API cases were diagnosed as either local or regional. Although 50% of patients with localized gastric cancer can be cured, the 5-year survival rate for disseminated gastric cancer is lower than 20% (22). In addition, the majority of Hispanics and blacks had the location of their noncardia tumor as “other.” Thus, either their tumor was located in areas of the stomach that could not be classified as body and fundus, pyloric antrum, pylorus and the curvatures; or location could not be determined; or location was not completely evaluated.

Screening for H. pylori is recommended by the American College of Gastroenterology among individuals with gastric cancer–related conditions such as peptic ulcer, gastric ulcer, duodenal ulcer, or a history of peptic ulcer (14, 15). Despite a high proportion of ncGC with late staging (∼30% regional), in the SEER-Medicare population, testing for H. pylori was low (7.7%) even among those with gastric cancer–related conditions (11.6%). However, the H. pylori testing both overall and for those with established conditions, occurred about 4 months before gastric cancer diagnosis. This is indicative that screening did occur before diagnosis. Furthermore, testing was particularly low among black gastric cancer cases with established conditions (6.5%), though this testing occurred within the shortest time period before gastric cancer diagnosis at 3.2 months. H. pylori testing was highest in APIs with established conditions (23%), but occurred about 7 months before gastric cancer diagnosis. A large proportion of foreign-born API cases also had established conditions (72%), but only 26% were tested for H. pylori. Finally, an important disparity became apparent among Hispanics with established conditions, who had the second highest H. pylori testing (17.9%), but the lowest endoscopy testing (86.2%); Hispanics also had the lowest endoscopy frequency across all individuals (62.5%). Endoscopy rates were highest in NHWs with established conditions (91%), and unlike H. pylori testing, endoscopies were performed, both overall and for those with established conditions, 0.3 months after gastric cancer diagnosis. This low level of testing and potential screening is contrary to the findings of a recent survey of gastroenterologists, which indicated that 97% of practitioners regularly test for H. pylori infection among patients with established conditions and 85% of practitioners report testing patients with gastritis (10). Whether these SEER-Medicare data are representative of broader testing and reporting practices, they indicate a potential source for intervention. Crew and colleagues (19) reported decreased mortality in high-risk H. pylori areas as a result of regular screening and early detection; for example, a 50% reduction in mortality in men in Japan due to implementation of mass screening programs.

Others have also suggested that individuals who migrate from high-risk areas (such as Japan) to low-incidence regions experience a decreased risk in developing gastric cancer (19). This finding is borne out in this study, where API cases, the majority of whom were foreign-born, were the only group with a majority of cases diagnosed at local or regional stage of noncardia tumor. In addition, H. pylori testing was higher in foreign-born than U.S.-born cases; cases with unknown place of birth had higher testing rates than U.S.-born cases as well.

Among gastric cancer cases, Lui and colleagues found the incidence of localized cancer had increased for black cases and stayed the same for Hispanics between 1992 and 2009 (16). Our results showed that the most common stages at diagnosis in this elderly population were regional (28.3%) and local (29.1%). The top two stages for black cases were distant (28.2%) and local (29.2%), which is consistent with other cancer studies showing that African Americans experience the highest mortality and shortest survival (20). Similar to the findings of Torre and colleagues (23), which stated that APIs in a non-Medicare sample were more likely to be diagnosed with gastric cancer at localized or regional stage than NHWs, we found the majority of APIs (63%) were diagnosed at the local or regional stage.

This study had several strengths. First, it utilized 14 years of incidence data from the large population-based SEER cancer registry data that had been merged with Medicare claims data. The results of this study would likely be generalizable to the elderly U.S. population that receives Medicare benefits. Second, the findings described disparities in screening and gastric cancer precursor conditions. This study also expanded upon the classic race/ethnicity categorizations by classifying race/ethnicity by place of birth. There were also several limitations. The SEER-Medicare dataset included only cases whose data (or healthcare) were in the fee-for-service Medicare claims data. As the data only capture the time a person was enrolled in Medicare, it is possible that H. pylori testing could have occurred before that person was eligible for Medicare. This analysis did not analyze, in-depth, the time interval between testing for H. pylori and endoscopy services, and gastric cancer diagnosis. Future analyses will focus more on determining screening and diagnostic testing both in relation to the various conditions, such as ulcers, and by race/ethnicity. In addition, these future analyses will also investigate sociodemographic factors and relevant conditions to determine what might predict whether individuals receive an endoscopy. Another limitation of our study is related to the information on race/ethnicity and place of birth. The information on race/ethnicity came from medical records, death records, and registration information (24), which may be more subject to misclassification, especially for Hispanic and NHWs (7). Furthermore, we used a combination of the Hispanic origin variable as determined by the North American Association of Central Cancer Registries (NAACR) Hispanic/Latino Identification Algorithm (NHIA), and the SEER race/ethnicity variable. The Hispanic surname algorithm that is part of the NAACR tool may not distinguish between Hispanic/Latino and Portuguese, Italian, or Filipino (25). In addition, responses to Hispanic/Latino origin questions have been inconsistent in self-report, which may be problematic with medical records that use patient self-report (25). Finally, place of birth information also came from medical records, as well as death records (26) and was missing from 28% of the cases. Because of more complete data on death certificates, the information on place of birth might be more complete for deceased patients (7, 26, 27).

This study is the first to describe demographic differences among gastric cancer cases in the SEER-Medicare population by race/ethnicity. Tumor location suggests that H. pylori infection still plays a potential causal role for many gastric cancer cases. Importantly, location and stage of tumor at diagnosis differed by race/ethnicity and H. pylori testing was higher in foreign-born than U.S.-born cases. H. pylori testing was low despite a high proportion of cases exhibiting gastric conditions for which H. pylori testing is recommended. Future studies can investigate the reasons for the low H. pylori testing rates in elderly patients with gastric cancer–related conditions.

A. Florea is a senior research specialist at the University of Arizona, College of Medicine, Department of Medicine, Division of Nephrology. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Florea, H.E. Brown, R.B. Harris, E. Oren

Development of methodology: A. Florea, H.E. Brown, R.B. Harris, E. Oren

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Oren

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Florea, H.E. Brown, E. Oren

Writing, review, and/or revision of the manuscript: A. Florea, H.E. Brown, R.B. Harris, E. Oren

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Florea, E. Oren

Study supervision: A. Florea, E. Oren

The authors would like to acknowledge the work put in by past and present members of the Helicobacter pylori working group at the Mel & Enid Zuckerman College of Public Health, Tucson AZ. Special thanks to Elizabeth Jacobs, PhD who provided outside review of the working manuscript, and to Jose M. Guillen-Rodriguez for his statistical assistance and advice. An oral abstract presentation related to this work was given at the American Society of Preventive Oncology annual meeting in New York City, New York, March 10–13, 2018. This study used the linked SEER-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the NCI; the Office of Research, Development, and Information, CMS; Information Management Services (IMS), Inc.; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database. The work and research reported in this publication was supported by the Chapa Foundation, Tucson AZ., and by the NCI of the NIH under award number P30 CA023074.

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.