Impact of Fasting Plasma Glucose Levels on Gastric Cancer Incidence in a General Japanese Population: The Hisayama Study

The Hisayama study is a prospective epidemiological study of ongoing cardiovascular disease and malignancy in Hisayama Town, a suburban community adjacent to Fukuoka City, a metropolitan area on Kyushu Island in Japan. The study design and characteristics of the subject population have been described in detail elsewhere (2, 13, 14). In 1988, 2,742 residents aged ≥40 years (80.1% of the total population in that age population) underwent a screening examination. After excluding 132 individuals with a prior history of gastrectomy or gastric cancer, 141 individuals who had eaten breakfast on the day of the screening examination, and 3 individuals who died during the examination period, a total of 2,466 subjects were enrolled in the present study.

The population was followed up for 9 years, from December 1988 through November 1997, by repeated health check-ups every 1–2 years. For subjects who did not undergo regular check-ups or who moved out of the town, the health status was checked every year by mail or telephone. In addition, a daily monitoring system was established by the study team and local physicians or members of the Division of Health and Welfare of the town. To identify any occurrence of gastric cancer in the cohort, we reviewed radiographic, endoscopic, and biopsy records for the stomach at local clinics or general hospitals within and around Hisayama Town. We also checked all the records of the annual mass screenings for gastric cancer by means of barium X-ray examination. Further, to find any concealed gastric cancer, autopsy examinations were performed on 212 (79.1%) of a total of 268 subjects who died during the follow-up period. The diagnosis of gastric cancer was confirmed by histological examination of resected specimens obtained by gastrectomy, endoscopic mucosal resection, or autopsy.

During the follow-up period, only 1 subject was lost, and 71 gastric cancers were identified in 66 subjects (48 men and 18 women); 5 subjects (7.6%) each had two gastric cancers, and 2 concealed cases (3.0%) were diagnosed at autopsy. The time interval from the baseline screening to the diagnosis of gastric cancer ranged from 0.5 to 8.7 years (mean 5.0 years).

For the measurement of FPG levels, blood was drawn from an antecubital vein using vacutainer tubes with heparin, EDTA, and fluoridated natrium. The blood sampling was undertaken between 8:00 a.m. and 10:30 a.m. after an overnight fast of at least 12 h. Plasma glucose was determined by the glucose-oxidase method. Diabetes was determined by either a 75-g oral glucose tolerance test (1998 World Health Organization criteria), FPG levels (≥7.0 mmol/l), or a medical history of diabetes. The numbers of subjects with diabetes diagnosed by each type of diagnosis were 294, 4, and 9, respectively. Based on the distribution of FPG levels, subjects were classified into tertile groups: low FPG (<5.3 mmol/l), modest FPG (5.3–5.8 mmol/l), and high FPG (>5.8 mmol/l).

Serum IgG antibodies to H. pylori were measured by means of a quantitative enzyme immunoassay using a commercial kit (HM-CAP; Enteric Products, Westbury, NY). The assay values were interpreted as positive, negative, or indeterminate, based on the manufacturer’s instructions. Serum cholesterol levels were determined by an enzymatic autoanalyzer (TBA-80S; Toshiba, Tokyo, Japan). Height and weight were measured with the subject in light clothes without shoes, and the BMI (kg/m²) was calculated. Dietary factors were obtained by a semiquantitative food frequency method that was previously validated in a prior study (15). A self-administered questionnaire concerning food intake over the previous year, which included 70 food items, was completed before the start of the study by each participant and was checked by experienced dietitians and nutritionists by showing food models of actual size in the survey. The average daily nutrient intakes, including total energy, total fat, salt, vitamin A, vitamin B₁, vitamin B₂, vitamin C, and dietary fibers, were calculated using the 4th revision of the Standard Tables of Food Composition in Japan (16), and the nutritional elements were adjusted for energy intake using the method of Willet and Stampfer (17). Information about smoking habits, alcohol intake, and history of peptic ulcer disease were obtained by means of a questionnaire administered to each subject, and the former two items were categorized as in current use or not in current use.

The SAS computer package (18) was used for all statistical analyses. Mean values of the possible risk factors were adjusted for age by the covariance method and were compared among tertile groups of FPG using Fisher’s least significant difference method. The frequencies of risk factors were compared by the Mantel-Haenszel χ² test after adjusting for age by the direct method. The incidence of gastric cancer was calculated by the person-year method, and its differences among groups with different FPG levels were analyzed by means of Cox’s proportional hazards model (19). The risk factor–adjusted relative risks (RRs) were also estimated using Cox’s proportional hazards model and are expressed together with 95% CIs. In the multivariate analysis, we used a stepwise method, setting the significance level for entry and keeping it at 0.1. Only 19 subjects who did not develop gastric cancer dropped out due to missing values in the covariates, while no case with gastric cancer dropped out of the analysis. For age adjustment, all study subjects were used as a standard population.

This study was conducted with the approval of the ethics committee of Kyushu University, and written informed consent for medical research was obtained from the study participants.

Table 1 compares the age-adjusted mean values or frequencies of possible risk factors for gastric cancer among the three FPG groups by sex. In men and women, the mean age increased significantly with an increase in FPG levels, and diabetes was found most frequently in the high-FPG group. Similarly, mean values of BMI and total cholesterol and frequency of alcohol intake in both sexes increased significantly with increases in FPG levels. The frequency of smoking habits in men decreased significantly with elevated FPG levels. The frequency of H. pylori infection and history of peptic ulcer disease and mean values of dietary factors were not found to be related to FPG levels in either sex.

As shown in Table 2, the age-adjusted incidence of gastric cancer of 5.6 per 1,000 person-years for men was significantly higher than that of 1.3 per 1,000 person-years for women. In men, the age-adjusted incidence was significantly higher in the modest-FPG (7.0, P < 0.05) and high-FPG (7.2, P < 0.05) groups than in the low-FPG (2.2) group. In women, the high-FPG group also had a significantly higher age-adjusted incidence of gastric cancer (2.5, P < 0.05) compared with that of the low-FPG group (0.8). The age- and sex-adjusted risks of gastric cancer in the modest-FPG (RR 2.3 [95% CI 1.1–5.1]) and high-FPG (3.0 [1.5–6.4]) groups were significantly higher than those in the low-FPG group (Fig. 1). These associations remained unchanged even after adjustment for age, sex, BMI, serum cholesterol, H. pylori seropositivity, smoking habits, alcohol intake, history of peptic ulcer disease, and dietary factors, including intake of total energy, total fat, salt, vitamin A, vitamin B₁, vitamin B₂, vitamin C, and dietary fibers. In addition, we performed the same analysis with all subjects except for those who developed gastric cancer in the first 2 years of the follow-up period to decrease the influence of the concealed gastric cancers at baseline. As a result, the age- and sex-adjusted RR of gastric cancer was 2.2 (95% CI 0.9–4.9) in the modest-FPG group and 2.9 (1.3–6.3) in the high-FPG group. The magnitude of the cancer risk in the modest- and high-FPG groups was almost the same as that obtained in the analysis of all subjects, although no statistically significant difference was found in the modest-FPG group, probably due to the small number of cases.

The seroprevalence of H. pylori was 66.6% for all subjects, 77.3% for those with gastric cancer, and 66.3% for the subjects who did not develop gastric cancer. We then compared the risk of gastric cancer among FPG groups under stratification by H. pylori status (Fig. 2). Among H. pylori–positive subjects, the age- and sex-adjusted risks of gastric cancer were significantly higher in the modest-FPG (RR 3.5 [95% CI 1.3–9.5]) and high-FPG (4.2 [1.6–11.1]) groups than in the low-FPG group, whereas no such differences were found in H. pylori–negative subjects.

Our findings indicate a positive association between elevated FPG levels and gastric cancer incidence in men and women, an association that remains significant even after adjusting for other risk factors such as age, sex, BMI, serum cholesterol, H. pylori seropositivity, smoking habits, alcohol intake, history of peptic ulcer disease, and dietary factors. The risk of gastric cancer was found to be increased not only in the high-FPG group, of which approximately one-third was diagnosed as diabetic, but also in the modest-FPG group, in which only a small number of subjects fulfilled the diagnostic criteria of diabetes. These results suggest that subjects with high FPG levels may have an increased risk of developing gastric cancer, even if they have not developed diabetes. In addition, a stratified analysis showed increased FPG levels to be an independent risk factor for gastric cancer only among H. pylori–seropositive subjects; no such risk was observed among H. pylori–seronegative subjects.

In this study, the age-adjusted incidence of gastric cancer was 5.6 per 1,000 person-years for men and 1.3 for women, which is higher than that found in previous studies in Japan (0.7–2.0 per 1,000 person-years for men and 0.3–0.7 for women) (20–23). This discrepancy seems to be due to differences in the study design as well as in the age structures or regions examined. The previous studies were registration studies, while ours was a prospective cohort study in which we carried out an intensive and accurate survey of gastric cancer, including autopsy examination of 79% of the deceased subjects to find any concealed gastric cancer. It is therefore supposed that our study results reflect the actual cancer incidence in the Japanese population.

The mechanisms for increased risk of gastric cancer in hyperglycemia remain obscure. One possible explanation is that hyperglycemia and its related conditions act directly as a carcinogenic factor. Dandona et al. (24) have demonstrated in a clinical study with diabetic subjects and healthy volunteers that diabetes is associated with increased production of reactive oxygen species and greater oxidative damage to DNA. In an experimental study, high glucose itself was shown to induce DNA damage (25). Thus, it is possible that increased production of reactive oxygen species or high glucose itself contributes to DNA damage, which may lead to mutational changes in oncogenes and tumor suppressor genes, and thereby to the development of gastric cancer.

Another possible explanation is that hyperinsulinemia is related to gastric carcinogenesis. Patients with hyperglycemia are prone to insulin resistance, which leads to higher levels of blood insulin. McKeown-Eyssen (26) and Giovannucci (27, 28) showed in epidemiological and experimental studies that hyperinsulinemia is involved in colonic carcinogenesis. These investigators independently hypothesized that well-specified risk factors for colorectal cancer, such as obesity, physical inactivity, alcohol consumption, or a typical western diet, contribute to insulin resistance. Ogihara et al. (29) have demonstrated that insulin enhances the stimulatory effects of epidermal growth factor on the proliferation of cultured gastric epithelial cells obtained from the guinea pig. It can be speculated that an increase in cell proliferation predisposes the gastric mucosa to genetic or epigenetic alterations and, therefore, to carcinogenesis (30, 31).

Finally, it is postulated that gastric cancer and hyperglycemia share common genetic or environmental risk factors. However, no common genetic background or common provisional risk factor other than age has been identified to date. Furthermore, that hypothesis cannot be supported by our results; we failed to show any significant correlation of FPG levels with H. pylori status or with dietary factors. Further, although smoking habits have been presumed to be a risk factor for gastric cancer (32), the frequency of smoking habits in men was rather low in the high- and modest-FPG groups relative to that in the low-FPG group.

Based on numerous epidemiologic and experimental studies, H. pylori has been regarded to be a definite risk factor for gastric cancer (2, 33). Although the precise pathogenetic role of H. pylori in gastric carcinogenesis remains unclear, it has been clarified that this organism contributes to modifications in epithelial cell proliferation (34, 35), which may be the initiating event in a cascade culminating in the development of gastric cancer. However, an increased risk of gastric cancer by H. pylori infection notwithstanding, the majority of H. pylori–infected subjects do not develop gastric cancer. As such, H. pylori is not the absolute oncogenic factor for gastric cancer, and there must be other critical cofactors contributing to the risk posed by H. pylori infection. Our stratified analysis showed increased FPG levels to be a significant risk factor for gastric cancer only among H. pylori–seropositive subjects; this link was not observed among H. pylori–seronegative subjects. This result indicates that hyperglycemia is a possible cofactor increasing the risk posed by H. pylori infection. In a clinical study, Acbay et al. (36) demonstrated that H. pylori gastritis enhances glucose- and meal-stimulated insulin release, probably by increasing gastrin secretion. Thus, the enhanced effect of hyperglycemia on the H. pylori–cancer association may be explained partially by hyperinsulinemia. Another possible explanation for this phenomenon may be that hyperglycemia affects H. pylori and its infection status or stimulates its carcinogenic effects. However, the association between H. pylori infection and diabetes is controversial in the literature. A higher prevalence of H. pylori infection in diabetic than in control subjects has been reported in some studies (37, 38), whereas other studies have found no association between H. pylori and diabetes (39, 40). In this study, we found no significant correlation between FPG levels and H. pylori status. It may be speculated that increased reactive oxygen–related damage to DNA and genetic or epigenetic alterations in gastric mucosa induced by hyperglycemia or associated hyperinsulinemia encourage a modifying effect of H. pylori on epithelial cell proliferation, which may be the initial step in a cascade of gastric carcinogenesis. Given the range of findings, this hypothesis requires further consideration.

Several limitations of our study should be discussed. The primary limitation of our study, which is typical of most prospective studies, is that changes in other potentially confounding factors for the development of gastric cancer were not reassessed over time in our subjects. It is therefore possible that as a result of treatment for diabetes, greater modification of other risk factors occurred in diabetic than in nondiabetic subjects. In our subjects, however, the risk of gastric cancer was increased even in association with pre-diabetic hyperglycemia, which is not subject to medical treatment. In addition, the carcinogenic effects of risk factors usually continue for a long period (41–43). Thus, bias of this kind was considered to be small in the present study.

A second limitation is that an average follow-up time of 5 years does not account for the much longer latency or induction period of gastric cancer. Accordingly, we cannot eliminate the possibility that there were concealed gastric cancers that had already developed by the time of the baseline examination, though this limitation is a common problem for a large majority of other registration studies of gastric cancer. However, the prevalence of gastric cancer in healthy subjects has been found to be low (0.12%) in a nationwide mass screening in Japan (44). In addition, our analysis of all subjects except for those who developed gastric cancer in the first 2 years of the follow-up period produced results similar to those obtained from our analysis of all subjects. We therefore believe that concealed cancers were rare at the baseline examination and that the influence of this bias is small.

The final limitation is that only a small number of gastric cancer cases developed in our cohort, indicating a high possibility of bias in the results. Nonetheless, we believe that the findings of our study represent an accurate incidence of gastric cancer and its association with hyperglycemia, since we performed the study using a highly accurate method for determining all gastric cancer cases.

In conclusion, we found the elevation of FGP levels to be a significant risk factor for gastric cancer in men and women. The contribution of FPG to the subsequent occurrence of gastric cancer was significant in H. pylori–seropositive subjects and not in H. pylori–seronegative subjects. These findings suggest that some interaction between hyperglycemia and H. pylori infection contributes to the development of gastric cancer or that hyperglycemia is a possible cofactor increasing the risk posed by H. pylori infection. Further study is necessary to clarify the pathogenetic role of hyperglycemia as well as of H. pylori infection and their interaction in gastric carcinogenesis.

This study was supported in part by a grant-in-aid for scientific research (C:13670370) from the Ministry of Education, Culture, Sports, Science and Technology of Japan.