Stomach Carcinoma Incidence Patterns in the United States by Histologic Type and Anatomic Site

Stomach cancer is the fourth most common cancer diagnosed and the second most frequent cause of cancer death worldwide (1, 2). Although stomach cancer rates are generally lower in the United States than worldwide, substantial numbers are affected. The American Cancer Society estimated that 21,500 people in the United States (13,190 men and 8,310 women) would be diagnosed with stomach cancer, and 10,880 would die from the disease during 2008 (3). Incidence and mortality rates for stomach cancer in the United States have decreased steadily for many years (4-6).

Stomach cancer may be classified into intestinal and diffuse types based on histopathology, as initially described by Lauren (7). The two biological entities are different with regards to epidemiology, etiology, pathogenesis, and tumor behavior. The diffuse type occurs in relatively younger individuals and has a poorer prognosis compared with the intestinal type (8). Using the Surveillance, Epidemiology, and End Results (SEER) data from 1973 to 2000, Henson et al. reported that rates for the intestinal type decreased by ∼50%, whereas rates for the diffuse or signet type increased by >400% (9).

Stomach cancers could also be classified by subsite within the stomach: cardia, fundus, body, distal (antrum and pylorus), and lesser or greater curvature. Some studies have shown that there has been a striking increase in gastric cardia cancer in the United States since the 1970s, although the incidence of stomach cancer as a whole has decreased (10, 11). Using SEER-9 data from the 1974-1976 to the 1992-1994 time periods, Devesa et al. (11) reported that the age-adjusted (1970 U.S. standard) incidence rates of gastric cardia cancer increased in both white males (from 2.1 to 3.3/100,000 person-years) and black males (from 1.0 to 1.9/100,000 person-years). However, a more recent report using SEER-11 data for cases diagnosed during 1992 to 1998 found that gastric cardia rates did not significantly increase during that time period among any ethnic or gender group (12).

Expanding on previous studies and adding cancer cases diagnosed through 2005, we used SEER data to analyze stomach carcinoma incidence patterns by histologic type, anatomic site, race, gender, and age.

The SEER Program of the National Cancer Institute has collected data for cases diagnosed since the early 1970s among residents of nine population-based cancer registries (San Francisco, Connecticut, Detroit, Hawaii, Iowa, New Mexico, Seattle, Utah, and Atlanta) that include ∼10% of the United States population (13).⁴

SEER Program registries routinely collect data on patient demographics, primary tumor site, morphology, stage at diagnosis, first course of treatment, and follow-up for vital status (13). This analysis used data from the November 2007 SEER data submission, released in April 2008 (14).

In SEER, histologic type and anatomic site were classified according to the Manual of Tumor Nomenclature and Coding (15) during the early and mid-1970s, the International Classification of Disease for Oncology (ICD-O; ref. 16) during the late 1970s through 1991, the second edition (ICD-O-2; ref. 17) for cases diagnosed from 1992 to 2000, and the third edition (ICD-O-3; ref. 18) for cases from 2001 forward. All cases have been recoded using ICD-O-3. We excluded nonepithelial cancers (such as sarcomas and myomatous neoplasms; M8800-M9759) and poorly specified neoplasms (M8000-M8004) in order to focus on the epithelial malignancies. The histologic classification of gastric carcinoma into the intestinal type and diffuse type is based on the criteria proposed by Lauren (7) and used by Henson et al. (9): diffuse types include signet ring cell carcinoma (M8490), diffuse carcinoma (M8145), and linitis plastica (M8142); intestinal type includes carcinoma (not otherwise specified; M8010), adenocarcinoma (not otherwise specified; M8140), tubular (M8211), and intestinal type (M8144). The remaining epithelial types formed the other type category. Based on anatomic site, we divided stomach cancer into three groups: cardia (C16.0), specified noncardia (fundus, body, antrum, pylorus lesser curvature, and greater curvature: C16.1-C16.6), and overlapping/nonspecified site (C16.8-C16.9).

Our analysis includes cases diagnosed during 1978 to 2005 because the coding system (Manual of Tumor Nomenclature and Coding) used prior to 1978 did not include codes for intestinal or diffuse carcinoma. There were a total of 56,803 stomach cancers diagnosed among residents of the nine SEER areas during 1978 to 2005. We excluded the 2,613 cases that were not carcinomas and restricted our analysis to the 54,190 cases specified as carcinomas. We then also excluded the 91 cases with race unspecified or unknown, and thus restricted our analysis to the 54,099 cases with race specified as white, black, or other specified. In 2000, the other specified race category in the nine SEER areas was 87% Asian/Pacific Islander and 13% American Indian/Alaska Native; population data for these racial categories were not available prior to 1992, so our analysis was limited to the categories white, black, and other specified race.

We calculated age-specific and age-adjusted (2000 U.S. standard population) incidence rates, expressed per 100,000 person-years, for 1978 to 2005 and for the three 6-year and two 5-year time periods from 1978 to 1983, 1984 to 1989, 1990 to 1995, 1996 to 2000, and 2001 to 2005. Temporal trends and age-specific rates were plotted so that a slope of 10 degrees represented a change of 1% per year (i.e., 40 years on the horizontal axis is the same length as one logarithmic cycle on the vertical axis; ref. 19). Only rates based on at least 10 cases were presented.

There were 54,099 stomach carcinomas diagnosed during 1978 to 2005 among residents of the nine SEER areas (Table 1). Of these, 40,040 (74%) were intestinal type, 8,558 (16%) were diffuse type, and 5,501 (10%) were other epithelial carcinomas. By anatomic site, 13,022 (24%) arose in the cardia, 25,002 (46%) were specified as noncardia, and 16,075 (30%) were of the overlapping or nonspecified site.

From the 1978-1983 to the 2001-2005 time periods, total stomach carcinoma rates declined by 34%. Regardless of site, rates for the intestinal type rate decreased by 44% from 8.6 to 4.8, whereas the diffuse type rate increased by 62% from 1.1 in 1978-1983 to 1.7 in 1996-2000 before declining to 1.5 in 2001-2005 (Fig. 1). Regardless of type, cardia carcinoma rates increased by 23% from 1.8 in 1978-1983 to 2.2 in 1996-2000 before decreasing slightly to 2.1, with most of the increase during the 1980s. Rates declined for specified noncardia and overlapping/nonspecified sites.

When considered by type and site, diffuse type rates increased for all the specified subsites from the 1978-1983 to the 1996-2000 time periods, after which they decreased; rates for overlapping/nonspecified subsites declined 17% overall (Fig. 2). Diffuse type rates increased most rapidly for the cardia: 377% from 0.1 in 1978-1983 to 0.3 in 1996-2000. Intestinal type rates decreased more than 40% for all subsites except for the cardia, which increased from 1.5 in 1978-1983 to 1.8 in 1984-1989 before plateauing. Rates for the other carcinomas all declined.

Among all six race/gender groups, carcinoma rates declined markedly for the intestinal type and other types while they generally increased for the diffuse type (Fig. 3). They also decreased consistently for the noncardia and overlapping/nonspecified sites but not for the cardia. Rates for all types and sites were higher among males than females, especially for the diffuse type and the cardia site. Intestinal type and noncardia site rates were highest among the other races and black males and lowest among white females. Diffuse type rates were consistently highest among the other races, intermediate among blacks, and lowest among whites. In contrast, cardia carcinoma rates were notably highest among white males and lowest among females of all three racial groups. Patterns for the other types and the nonspecified sites were similar to those for the intestinal type and the noncardia site. Male/female rate ratios were ∼2 for all types combined and for both intestinal and other types, but they were 1.1 to 1.7 for the diffuse type and 3.3, 3.9, and 5.0 for cardia cancers among blacks, other races, and whites, respectively (Table 2). Noncardia rates among blacks and other races were twice and triple those among whites, respectively, whereas cardia rates were highest among whites.

Stomach carcinoma incidence rates for each type and site increased exponentially with age among all six race/gender groups; the curves were steeper for intestinal compared with diffuse type carcinomas and for noncardia than cardia site (Fig. 4). Rates increased less rapidly at older ages in several instances, especially for cardia cancer among white and black males. The racial/gender patterns generally mirrored those apparent in the age-adjusted rates, although male/female differences were much less pronounced at younger than older ages for diffuse carcinomas and noncardia sites. The male predominance was apparent across all ages for cardia carcinomas, and the male/female rate ratio peaked at 7 among whites ages 55 to 59 years.

Our study found that in the United States, during the years 1978 to 2005, stomach carcinoma incidence rates for the intestinal histologic type decreased whereas those for the diffuse type increased through 2000 and then declined in recent years. By anatomic site, cardia rates increased during the earlier years and then plateaued, whereas the rates for all other sites decreased. When considered jointly by histologic type and anatomic site, the rates decreased for intestinal type cancer of all sites except for gastric cardia; rates increased for diffuse types of all sites through 2000 and then declined except for overlapping/nonspecified sites. Diffuse and intestinal rates were lowest among whites, intermediate among blacks, and highest among others, primarily of Asian race, with only modest gender differences for the diffuse type. In contrast, cardia carcinoma rates were highest among whites and were notably higher among males, especially in whites among whom the cardia male/female rate ratio was five. Age-specific rates increased exponentially for all types, sites, races, and genders. The male predominance was notably absent at younger ages for diffuse carcinomas and noncardia sites, and the male/female rate ratio reached 7 for cardia carcinoma among whites ages 55 to 59 years.

Previous studies (9, 20, 21) have indicated that the decreases in stomach cancer over the past seven decades may be attributed to declines primarily in intestinal type histology, as we also show here. The intestinal type is related to atrophic gastritis and intestinal metaplasia, whereas the diffuse type is usually related to nonatrophic gastritis (22-25). Typically, the intestinal type affects older age groups and occurs notably more frequently among males, whereas the diffuse type is more prevalent among younger age groups and exhibits more modest gender differences (24), which our data also show. The dominant risk factors for stomach cancer overall and especially for intestinal/noncardia type cancer are diet, cigarette smoking, and Helicobacter pylori infection, whereas those for cardia cancer are obesity, gastroesophageal reflux, and Barrett's esophagus (25, 26). Some, but not all, studies have found that aspirin and other nonsteroidal anti-inflammatory drug use decreased the risk of both cardia and noncardia gastric cancers (27).

The declining prevalence of H. pylori infection likely has contributed to the downward trends of intestinal type stomach cancer at noncardia sites (i.e., at distal sites) seen in our study. Based on a meta-analysis of prospective studies, H. pylori infection was associated with the risk of noncardia stomach cancer but not cardia cancer (28). It has also been reported that H. pylori gastritis is a universal precursor condition for both intestinal and diffuse type stomach cancers (23, 24), although it is not clear how this could account for the divergent incidence trends. A recent meta-analysis found that fruit and vegetable intake reduced risk both sites and for both types, although the protective effects were more pronounced for the intestinal type (29); improvements in diet likely have contributed to the long-term trends in intestinal cancer. Noncardia gastric cancer risk is inversely associated with intake of several micronutrients, especially among H. pylori/Cag A–positive cases of the intestinal type (30). Cigarette smoking was found to increase the risk of both cardia and noncardia cancer in another recent meta-analysis (31); declines in cigarette smoking may have contributed to the decreases in intestinal cancer in recent years but could not account for the long-term trends.

It has been suggested that cardia tumors share demographic and pathologic features with Barrett's associated esophageal adenocarcinoma and are more likely to occur in men (11, 32). This parallels the male predominance in the increasing incidence of lower esophageal adenocarcinoma (22, 33, 34). However, in contrast to the rapid increases in esophageal adenocarcinoma (34), total cardia carcinoma rates have not been rising. The rates for diffuse carcinomas of the cardia have, however, been increasing; these trends are similar to the rising rates of reflux disease apparent among male veterans, and reflux rates also have been higher among whites than in blacks (35). Adenocarcinoma of both the esophagus and gastric cardia have been associated with obesity, although the relationship for cardia cancer seemed restricted to U.S. and European studies but not to Chinese studies (36). Furthermore, in contrast to the rapid increases suggested here for diffuse carcinomas of the cardia, diffuse carcinomas of the esophagus are exceedingly rare, and rates have not increased in concert with those for other adenocarcinomas (data not shown). Although we found that the incidence of gastric cardia cancer did not continue to increase in recent years, the trends are still in contrast with the long-term declines in stomach cancer at other sites.

Total stomach cancer rates among both males and females have been highest among Asian/Pacific Islanders, somewhat lower among blacks, and much lower in whites. These patterns are in contrast with the cardia rates that we found to be highest among whites and lowest among blacks. These variations in racial patterns were also apparent in the 1996-2000 data reported from 24 registries (37). Furthermore, the male predominance of 3-fold to 5-fold for cardia cancer overall, and 7-fold at ages 55 to 59 among whites, contrast sharply with the male/female rate ratios of <2 apparent for total stomach cancer (13) and seen here for noncardia sites. Based on stomach adenocarcinoma cases diagnosed in 37 U.S. registries during 1998-2002, a larger male/female rate ratio for cardia (2.8) versus noncardia (1.7) was also apparent among Hispanics (38). Total stomach cancer rates among Hispanics have been similar to the rates among blacks, lower than those among Asian/Pacific Islanders, and higher than those among whites (13, 38). This pattern was also apparent for noncardia cancers, whereas cardia rates among Hispanics were higher than those among black males and females and white females, but not white males (38). It would be of interest to evaluate the temporal trends in stomach cancer by site among Hispanics and Asian/Pacific Islanders using the data available in SEER since 1992.

The steeper increases in rates with age for intestinal than diffuse type and for noncardia than cardia sites, apparent for each race/gender group, may also reflect differing diseases. An analysis of SEER-11 data for cases diagnosed during 1992 to 2001 found that the age-specific curves were consistently steeper for intestinal than for diffuse cancers across genders and races (Asian, non-Asian), which was interpreted as supporting the notion that these cancers differ in their etiologies or pathways from premalignancy to malignancy (39). These authors included only M8144 as the intestinal type; it seems that these observations hold when the much larger group of histologic types is included, as was done in our analysis and presented in Fig. 4. The changes in slope at older ages apparent for cardia cancer among males likely is due to period or cohort effects, with rates rising more rapidly over time among those middle-aged than older.

The incidence rate for overlapping and unspecified site declined from 4.2 in 1978-1983 to 1.8 in 2001-2005. Some studies have suggested that improved site classification may largely account for the observed increase in cardia cancer incidence in recent decades and that noncardia incidence may be decreasing more rapidly than previously reported (40). However, although improved site classification may have played a role, the cardia incidence trends still differ from those for other sites.

The decrease in the diffuse carcinoma rate and increase in the rate for the other types in 2001-2005 compared with 1996-2000 seen overall, for each site, and for each race/gender group suggest possible coding differences between ICD-O-2 and ICD-O-3. In fact, several new histologic type codes were added in ICD-O-3. In particular, a specific code for adenocarcinoma with mixed subtypes (M8255) appears in ICD-O-3 but not in ICD-O-2. A case described as signet ring cell and other types of carcinoma may have been assigned the higher signet ring cell code M8490 in prior years (included with diffuse type in our analysis) and the mixed type coded M8255 in recent years (included with other types in our analysis). When cases coded to the mixed type code M8255 are included with diffuse rather than with the other types, the decline in the diffuse rate from the 1996-2000 to the 2001-2005 time period is cut by 1/3 and the increase in the rate for the other types virtually disappears.

The differing patterns for gastric cardia cancer compared with those for the distal sites and the rising rates of diffuse type in contrast to declines in intestinal type both suggest etiologic heterogeneity. The substantially larger male/female rate ratios for cardia cancer and the predominance among whites may suggest clues to pursue. Thus, it seems that stomach cancer arising at the cardia versus distal sites and of different histologic types may be distinct diseases, and future studies should attempt to clarify the role of risk factors for these cancers separately as we attempt to learn more about the etiologies of these diseases and the means for their prevention.

No potential conflicts of interest were disclosed.

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.

We thank the SEER registry and NCI staff for their invaluable work in developing the database; Drs. Wong-Ho Chow and Sandy Dawsey of the DCEG, NCI for manuscript review and discussion of pathologic classification; and John Lahey of IMS, Inc., and David Check of the Biostatistics Branch, DCEG, NCI for figure development.