The pathogenic role of Helicobacter pylori for gastric cancer has been documented in a large number of epidemiological studies^1–5 and basic research investigations.^6–9 In earlier epidemiological studies using H pylori antibody as a marker of infection, various risk ratios of H pylori infection for gastric cancer were reported, ranging widely from none to 10 or above.^{1–3,10–16}

Recently, a follow up study by Uemura et al showed that gastric cancer developed only in patients infected with H pylori when using a full set of diagnostic tests for H pylori infection.⁵ They also reported in the same study that subjects with severe gastric atrophy, corpus predominant gastritis, or intestinal metaplasia were at increased risk for gastric cancer.⁵

We also confirmed that gastric atrophy status was essential for cancer development in our previous cross sectional study.¹⁷ In that study, gastric atrophy was estimated by serum pepsinogen levels, which were determined in serum samples.¹⁷ Pepsinogen I and II, the two main precursors of pepsin, are both produced by chief cells and mucous neck cells of the stomach.^18,19 Pepsinogen II is also produced by pyloric gland cells. Chief cells are replaced by pyloric glands, leading to a decrease in pepsinogen I as gastric atrophy develops. However, a decrease in pepsinogen II is minimal. Therefore, both low serum pepsinogen I and a low pepsinogen I/II ratio are recognised as serological markers of gastric atrophy.^20–22

The combination of serum pepsinogen and H pylori antibody served as a useful marker for the prevalence of gastric cancer in a cross sectional setting.¹⁷ This modality is much simpler and less invasive than those using endoscopy, and therefore suitable for a large general population. On the basis of this premise, we conducted the present prospective study in participants in our health check programme without any specific symptoms. We aimed to estimate the incidence rate of gastric cancer in the general population. The role of H pylori infection and gastric atrophy in cancer development was evaluated in terms of these serological markers.

METHODS

Enrolment

Between March 1995 and February 1997, participants in health examination programmes held by Kameda General Hospital and Makuhari Clinic who underwent upper endoscopy were consecutively enrolled. Blood samples were obtained from each subject. Excluding those with gastric cancer, peptic ulcer, or a past history of surgical resection of the stomach, a total of 9293 participants were candidates for inclusion in this study. Some of these subjects were analysed in a previous cross sectional study.¹⁷ Proton pump inhibitors or H₂ blockers had not been prescribed within one month prior to the examination. None had undergone eradication therapy for H pylori. Patients were encouraged to undergo endoscopic examination annually to check for the development of gastric cancer, and 6983 revisited the programme for follow up endoscopy during the observation period. Data of these participants were analysed in this study.

The protocol was approved by the ethics committees of the respective institutions, and informed consent was obtained from each subject according to the Declaration of Helsinki.

Serum H pylori antibody

Serum anti-H pylori antibody was measured using a commercial ELISA kit (GAP-IgG kit; Biomerica Inc., California, USA). Seropositivity for H pylori antibody was defined by optical density values according to the manufacturer’s protocol. Sensitivity and specificity for H pylori infection in Japanese were reported to be 95% and 83%, respectively, compared with the results of specific culture.²³

Serum pepsinogen level

Serum pepsinogen was measured using a commercial RIA kit (pepsinogen I/II RIA bead kit; Dainabot Co., Tokyo, Japan). Serum pepsinogen status was defined as “atrophic” when the criteria of both serum pepsinogen I level ⩽70 ng/ml and a pepsinogen I/II ratio (serum pepsinogen I (ng/ml)/serum pepsinogen II (ng/ml)) ⩽3.0 were simultaneously fulfilled, as proposed by Miki and colleagues.²² All other cases were classified as “normal”. A sensitivity of 70.5% and specificity of 97.0% for atrophic gastritis compared with histology have been reported in Japan.²⁴ These criteria have been widely applied to mass screening for gastric cancer in Japan.^17,22,24

Classification by anti-H pylori antibody and serum pepsinogen status

Subjects were classified into four groups according to serum pepsinogen status and H pylori status antibody at enrolment. Group A had “normal” pepsinogen and were negative for H pylori antibody. Group B had “normal” pepsinogen and were positive for H pylori antibody. Group C had “atrophic” pepsinogen and were positive for H pylori antibody. Group D had “atrophic” pepsinogen and were negative for H pylori antibody.

Endoscopic and clinicopathological examinations

Gastrointestinal endoscopy was performed with electronic panendoscopes (type XQ200 or P230; Olympus, Tokyo, Japan), carefully observing the bulbar portion of the duodenum, the entire stomach, and the oesophagus. Experienced endoscopists performed each examination without knowledge of the serological data of the study subjects. Histopathological assessment of gastric cancer was conducted using surgically resected or endoscopically biopsied samples, categorised as intestinal-type or diffuse-type, according to Lauren’s classification.²⁵ Samples were classified as cardiac or non-cardiac in terms of location.

Statistical analysis

All statistical analyses were performed using SAS software (SAS Institute Inc., North Carolina, USA). Differences in mean age were evaluated by analysis of variance (ANOVA) with Fisher’s correction. Difference in sex distribution was evaluated by the Kruskal-Wallis test with Bonferroni’s correction. Incidence of gastric cancer was calculated using the Kaplan-Meier method. Independent risk factors for gastric cancer were assessed by Cox proportional hazard regression. A two sided p value of less than 0.05 was considered statistically significant.

RESULTS

Baseline characteristics of study subjects

Baseline clinical characteristics of the study subjects are summarised in table 1. Of 6983 subjects, 3324 (47.6%) were categorised as group A, 2134 (30.6%) as group B, 1082 (15.5%) as group C, and 443 (6.3%) as group D. Each subject underwent 5.1 (0.05) sessions of endoscopy during a follow up period of 4.7 (0.04) years.

Gastric cancer development

Among 6983 subjects analysed, 43 (37 men and six women) developed gastric cancer during the follow up period. The annual incidence rate of gastric cancer development, as calculated by the person-year method, was 0.13% (95% confidence interval (CI) 0.10%–0.18%). Histopathological features of gastric cancer were intestinal in 34 and diffuse in nine cases. Gastric cardia was involved in two cases. All of the cancers were localised within the submucosa except for one invading the muscularis propria (group B). Twenty three cases were treated by endoscopic resection and 20 cases underwent surgical operation. All were alive in August 2004.

Antibody-pepsinogen status and gastric cancer development

Of 43 cases with gastric cancer, seven were from group A, six from group B, 18 from group C, and 12 from group D. The annual incidence rate was 0.04% (95% CI 0.02%–0.09%), 0.06% (0.03%–0.13%), 0.36% (0.23%–0.57%), and 0.60% (0.34%–1.05%) in groups A, B, C, and D, respectively. The cumulative incidence of gastric cancer by Kaplan-Meier analysis is shown in fig 1, as stratified by group. Groups C and D had a significantly higher incidence of gastric cancer than groups A and B (fig 1). Histopathological features of gastric cancer are shown in table 2. Two cases were found in the gastric cardia and the other 41 elsewhere. Both cardiac cancers occurred in group A. In contrast, no association was found between the groups and histopathological differentiation of cancer.

Risk factors for gastric cancer and establishment of super high risk group

Age, sex, and “group” were revealed to be independent risk factors by the Cox proportional hazard model (table 3). Hazard ratios (95% CI) compared with group A were 1.1 (0.4–3.4; p = 0.81) in group B, 6.0 (2.4–14.5; p<0.0001) in group C, and 8.2 (3.2–21.5; p<0.0001) in group D.

Incidence rates of gastric cancer stratified by age, sex, and “group” are shown in fig 2. Males older than 60 years in group D showed the highest annual incidence of 1.8% (95% CI 0.81%–3.82%). The incidence rate in the same age group was much lower in groups A and B, being less than 0.5% per year.

DISCUSSION

Gastric cancer is the second (in males) and fourth (in females) lethal cause of malignancy in the world.²⁶ It still remains the most common malignancy in many countries.²⁷H pylori has been established as a definite carcinogen for gastric cancer.²⁸ However, the magnitude of the association in reports has been diverse, especially in Eastern countries suffering high prevalence rates of gastric cancer.^10,12–15 Uemura et al claimed from the results of their follow up study that all gastric cancers developed from patients with H pylori infection, and that the risk was highly associated with gastric atrophy status induced by H pylori.⁵ The result is epoch making and revealing in terms of understanding and preventing gastric carcinogenesis. However, their results were based on hospitalised patients with gastrointestinal diseases, as well as other follow up studies.^29,30 It should be validated in other settings, in particular in the general population.

Ours is the first large scale prospective follow up study using serum pepsinogen and anti-H pylori antibody to estimate the incidence of gastric cancer in the general population. Subjects in our study were consecutive participants in a general health checkup programme, a very common activity in Japan.^17,22,24,31 Participants were symptom free, and those with peptic ulcers or gastric cancers were excluded from the cohort, as they were receiving treatments such as gastric acid suppression, H pylori eradication, or surgery. It is likely that our subjects represent the healthy Japanese population, with fewer biases than hospitalised patients. Moreover, as was shown by the average number of endoscopic examinations, gastric cancer development was closely and evenly surveyed in each group. Thus gastric cancer development could be accurately detected with minimum delay or aberration. The present study would no doubt estimate precise incidence rates of gastric cancer in the general population.

Serological markers were used in this study for gastric atrophy status induced by H pylori. Subjects were stratified according to H pylori antibody and pepsinogen status into groups A, B, C, and D. Group A (negative for H pylori and normal pepsinogen normal) was assumed to have no H pylori infection whereas the other groups were infected with H pylori. As was discussed in our previous study, group D was assumed to have the most advanced gastric atrophy due to H pylori infection in spite of being negative for H pylori antibody.¹⁷ Pepsinogen levels indicated the most severe gastric atrophy in group D.^4,20,32 It is generally known that the H pylori burden decreases dramatically in such situations,³³ and H pylori antibody spontaneously disappears.³⁴ In fact, our preliminary data from the same cohort of the present study showed a small but significant progression of gastric atrophy and reduction of serum pepsinogen at eight year intervals in groups B and C, leading to group advancement in some patients.³⁵

In the present study, among 6983 subjects analysed, 43 developed gastric cancer during the follow up period. The annual incidence rates of groups A–D steadily increased in this order. Our results are in agreement with those of Uemura et al, irrespective of the difference in study population and diagnostic method for H pylori.⁵ In addition, we are able to define a super high risk group for the development of gastric cancer (group D). Group D comprised 25.7% of subjects older than 60 years, and gastric cancer developed at the highest rates of 1.8%/year in males and 1.5%/year in females from this group. In contrast, group B (H pylori positive and pepsinogen normal) showed the same low risk as group A without H pylori infection. Approximately 58% of those with H pylori infection could be regarded as having a negligible risk for at least five years.

In terms of histopathological features, cardiac cancers, which have been suspected to have little association with H pylori,^2,16 all developed in group A. Both intestinal- and diffuse-type gastric cancers were highly associated with H pylori infection, as has been reported in previous studies.^12,16

In the present study, all of the gastric cancers detected by endoscopic follow up were resectable and most are expected to be curative. Although it is still to be confirmed by longer observation, close endoscopic follow up could be valuable for subjects in the high risk group. Furthermore, eradication of H pylori may be recommended in the population, even in low risks group who are infected with H pylori, if steady progression of gastric atrophy is assumed.

In conclusion, we prospectively observed the natural course of gastric cancer development in the Japanese general population. We found H pylori antibody and serum pepsinogen to be good predictive markers for the development of gastric cancer. There is an increasing tendency for gastrocarcinogenesis with progression of H pylori infection. We believe this study provides definitive baseline data for future prevention studies in gastric cancer.