SEARCH

SEARCH BY CITATION

Keywords:

  • cigarette smoking;
  • alcohol consumption;
  • gastric cancer;
  • histologic type;
  • cardia cancer

Abstract

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The effect of cigarette smoking or alcohol consumption on the risk of gastric cancer has not been clarified. We investigated this relationship, considering the anatomic subsite and histologic type of gastric cancer. A total of 19,657 men (aged 40–59 years at baseline), who responded to the baseline questionnaire and reported no serious illness at that time, were followed for 10 years, from January 1990 to December 1999. Gastric cancer was confirmed histologically in 293 men. Smoking was associated with an increased risk of the differentiated type of distal gastric cancer; compared to the group who never smoked, the adjusted rate ratios (RRs) of gastric cancer for past and current smokers were 2.0 (95% CI 1.1–3.7) and 2.1 (95% CI 1.2–3.6), respectively. No association was observed between cigarette smoking and risk of the undifferentiated type of distal gastric cancer except for a suggestive association with cardia cancer. For alcohol consumption, elevated risk was suggested only for cardia cancer of all histologic types, though the relationship failed to reach significance. Among those who drank alcohol at least once per week, RRs for ethanol intake of 2.7–161.0, 162.0–322.0 and 322.5+ g/week compared to those who drank 0–3 times/month were 2.5 (95% CI 0.7–9.5), 3.3 (0.9–11.6) and 3.0 (0.8–11.1), respectively (ptrend = 0.66). In conclusion, our results confirm that smoking is related to gastric cancer of the differentiated type. Further studies with more cases are needed to detect a positive association between cigarette smoking or alcohol consumption and cardia cancer. © 2002 Wiley-Liss, Inc.

It has been suggested that the major cause of gastric cancer is environmental rather than genetic.1 Although infection with Helicobacter pylori and diet have been identified as major risk factors, smoking and drinking have also attracted attention as possible causes of gastric cancer. Several researchers have observed a positive relationship between smoking or alcohol consumption and the risk of gastric cancer, whereas others have failed either to show such an association or to verify a dose–response relationship.1, 2

Previous studies have also indicated that gastric cancer cannot be explained as a single entity.1 In contrast to the decline in the occurrence of distal gastric cancers,3 recent reports have revealed that the incidence of cancer localized to the cardia may be on the rise.4, 5 The observed differences in clinical and pathologic profiles suggest that these 2 tumor types are distinct diseases with different etiologies.6, 7 Gastric cancer may also be divided into 2 major histologic types, diffuse and intestinal, according to Lauren's7 classification. The differential pattern of occurrence by place and time suggests that the intestinal type of gastric cancer is more closely governed by environmental factors than by host-related factors.8, 9 Subsequent studies, however, have not supported this hypothesis,10, 11, 12, 13 and more investigations are needed.

Many epidemiologic studies have evaluated the role of smoking or alcohol use as a risk factor in gastric cancer, but few have considered anatomic and histopathologic subdivisions. We report this relationship over a 10-year follow-up period in a population-based cohort study, the Japan Public Health Center (JPHC)-based prospective study on cancer and cardiovascular diseases (JPHC Study Cohort I).

MATERIAL AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Study population

The JPHC Study Cohort I was partly reported on elsewhere.14 As of 1 January 1990, we established a population-based cohort of 27,063 men and 27,435 women, who registered their addresses in 14 administrative districts supervised by 4 PHC areas: 12,291 from Ninohe City and Karumai Town in the Ninohe PHC area of Iwate Prefecture, 15,782 from Yokote City and Omonogawa Town in the Yokote PHC area of Akita, 12,219 from 8 districts of Minami-Saku County in the Saku PHC area of Nagano and 14,206 from Gushikawa City and Onna Village in the Ishikawa PHC area of Okinawa. All subjects were born between 1930 and 1949 (40–59 years of age at baseline). The JPHC Study area was based on a previous ecologic study of gastric cancer;15, 16, 17 thus, it encompasses areas with representative rates of gastric cancer mortality.

Baseline questionnaire

In 1990, subjects were asked to reply to a lifestyle questionnaire, covering sociodemographic characteristics, medical history, diet as well as history of cigarette smoking and alcohol consumption. A total of 43,149 subjects, 20,665 men (76%) and 22,484 women (82%), returned their questionnaires. Subjects were first asked whether they had ever been smokers. Subjects were assigned to a past-smoker or a current-smoker group according to current smoking status. Age at initiation of smoking, number of cigarettes consumed/day for smokers and age at cessation of smoking for past smokers were also included.

Frequency of alcoholic beverage consumption was determined using a precoded category of 6 levels: none or <1 day/month, 1–3 days/month, 1–2 days/week, 3–4 days/week, 5–6 days/week or daily). Subjects who reported alcohol consumption at least once per week were also asked to report on the usual amount of 4 specified beverages (sake, shochu/awamori, beer and whiskey). Units were go (180 ml) for sake and shochu, large bottles (633 ml) for beer and glasses (30 ml) for whisky. Weekly ethanol consumption was calculated by combining the daily amount of total ethanol consumption and frequency per week. Subjects who consumed alcohol at least once per week were classified by tertiles, and thus subjects were categorized in four groups as follows: 0–3 days/month, 2.7–161.0 g/week, 162.0–322.0 g/week and 322.5+ g/week. Since smoking and alcohol drinking are rare in Japanese women of this age group, we confined the analysis to men.

The weekly intake frequency of 27 food items was reported in 4 categories: rarely, 1–2 days/week, 3–4 days/week, almost daily (5 days or more/week). For rice, soybean paste soup and 9 kinds of beverage, the daily amount consumed was also asked.18 Family history of gastric cancer was regarded as positive if 1 parent or sibling had gastric cancer.

We excluded subjects with a self-reported serious illness (cancer, ischemic heart disease, cerebrovascular disease, chronic liver disease) at baseline, subjects who were not Japanese and subjects who had already moved away at baseline, which we confirmed during the follow-up period. These exclusions left 19,657 eligible men for study.

Follow-up and identification of gastric cancer

All subjects were followed from 1 January 1990 to 31 December 1999. In Japan, all death certificates are submitted to a local government office and forwarded to the PHC in the area of residence. Mortality data are then sent to the Ministry of Health and Welfare and coded for the National Vital Statistics. Registration of deaths in Japan is required by the Family Registration Law and is believed to be complete. Therefore, all deaths of cohort subjects were based on death certificates from each PHC, whenever the subjects stayed in their original area. Changes of residence were identified annually through the residential registry in each area.

Cancer registry for JPHC study

Newly diagnosed cases of cancer were reported by hospitals in and around the study areas when the birth date and residence fulfilled cohort inclusion criteria. Candidate patients were linked by name, address and date of birth and entered in the cancer registry for the JPHC Study Cohort I. In the Ninohe and Ishikawa PHC areas, prefecture-wide cancer registry was available: the Iwate and Okinawa prefecture cancer registries, respectively. The death certificate was used as a supplementary information source for the cancer registry, and 157 cases were first identified by it. Such cases accounted for 10.5% (DCN) of all 1,494 male entries in the cancer registry as of November 2000, which were diagnosed in 1990–1999. Twenty-five of those cases were not confirmed by medical records, and they accounted for 1.7% (DCO) of all entries in the cancer registry.

Identification of gastric cancer

Cases of gastric cancer were extracted from the cancer registry for the JPHC Study, based on site (ICD-O code C160–169)19 and histologic confirmation by biopsy or surgery. A total of 405 cases, 293 in 19,657 men, were documented, with histologically proven diagnosis made in 1990–1999, as of November 2000. DCN accounted for 11 cases (2.7%), and no DCO cases were included because cases were restricted to subjects with histologic diagnosis.

Cardia cancer was defined as a tumor located in the esophagogastric junction or upper third of the stomach (ICD-O code C160–161). Until quite recently in Japan, the upper third of the stomach has been called the cardia based on the guidelines for gastric cancer classification.20 Because of difficulty distinguishing real cardia from the upper third of the stomach, we combined them into 1 group for analysis in this study. A tumor located in the lower side of the stomach was classified as distal gastric cancer (ICD-O code C162–167). Subsites that could not be classified because of a diffuse lesion (ICD-O code C168) or no information (ICD-O code C169) were categorized as unclassified. Histologic classification was based on review of the record from each hospital by 1 of the authors (S.S.) in consultation with a pathologist. Subdivisions were made as follows: differentiated type (corresponds to intestinal type in Lauren's7 classification), including papillary adenocarcinoma, tubular adenocarcinoma (well-differentiated type) and tubular adenocarcinoma (moderately differentiated type), and undifferentiated type (corresponds to diffuse type in Lauren's7 classification), including poorly differentiated adenocarcinoma, mucinous adenocarcinoma and signet-ring cell carcinoma. Adenosquamous carcinoma, squamous cell carcinoma, carcinoid tumor, undifferentiated carcinoma and miscellaneous were considered unclassified types.

Statistical analysis

Time at risk for each subject was calculated as the duration from the return of the baseline questionnaire to a histologic diagnosis of gastric cancer, move from the PHC area, death or 31 December 1999, whichever came first. Cox's proportional hazards regression model was used to estimate RRs of gastric cancer according to smoking or drinking habit. Covariates used in the model were age; PHC area; consumption of fruit, green or yellow vegetables, salted cod roe or fish gut; and body mass index, with indicator variables representing the separate levels of each covariate. Age was categorized in 1 of 4 groups: 40–44, 45–49, 50–54 and 55–59 years, based on age at baseline. Fruit consumption was categorized in 3 groups: <2 days/week, 3–4 days/week and almost daily. Consumption of green or yellow vegetables, the sum of the frequencies of intake of green vegetables and yellow vegetables, was classified in 3 groups: <4 times/week, 5–7 times/week and >8 times/week. Salted cod roe or fish gut consumption was expressed as the sum of the frequencies of intake of each and was categorized in 3 groups: none, 1–2 times/week and >3 times/week. Body mass index was classified in 3 categories: <22.0, 22.0–25.0 or ≥25.0. Tertiles were used to categorize smoke-related variables, such as cigarettes/day, age started smoking and pack-years. To evaluate the risk from alcohol consumption, subjects who drank rarely or 1–3 days per month were combined into 1 group.

The trend was assessed by assigning ordinal values for categorical variables and median values for continuous variables. Trends for cigarettes/day, age started smoking and pack-years were calculated among current smokers. For alcohol consumption, trend was calculated among those who drink at least once per week. For the presentation of incidence rates for each type of cancer, person-years at risk were calculated. The effect of interaction was checked by calculating an interaction term, multiplying a dummy variable for smoking (current smoker = 1, never or past smoker = 0) by one for alcohol drinking (drinking at least once per week = 1, 0–3 days per month = 0). p values were 2-sided, and all statistical analyses were done using the Statistical Analysis System (SAS, Cary, NC).21

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Table I shows the distribution of 293 gastric cancers by subsite and histologic type. There were 37 cardia cancers (13%) and 205 distal gastric cancers (70%). Distribution of histologic types did not substantially differ among subsites. Cardia cancer of any histologic type, including unclassified type, and distal gastric cancer with known histologic types were used in the analysis.

Table I. Distribution of Gastric Cancer Cases According to Anatomic Subsite and Histologic Type in Men
 Differentiated (%)Undifferentiated (%)Unclassified (%)Total (%)
Cardia and upper third20 (54.1)13 (35.1)4 (10.8)37 (100.0)
Distal134 (65.4)66 (32.2)5 (2.4)205 (100.0)
Unknown27 (52.9)16 (31.4)8 (15.7)51 (100.0)
Total181 (61.8)95 (32.4)17 (5.8)293 (100.0)

Baseline characteristics of the cohort according to smoking status and drinking habits are shown in Tables II and III, respectively. All listed variables were distributed differentially according to smoking or drinking habits, except for consumption of soybean paste soup and family history for smoking status. Intake of salted cod roe or fish gut increased substantially as drinking level increased.

Table II. Baseline Characteristics According to Smoking Status in 19,576 Men, Japan, 1990–1999
 Smoking status1p for difference
NeverPastCurrent
  • 1

    Figures are means (SE) unless otherwise specified.

  • 2Based on 1-way ANOVA or χ2 test.

Number4760437210444 
Age (years)49.6 (0.1)50.2 (0.1)48.9 (0.1)0.0001
Alcohol drinking, at least once/6 week (%)23.536.142.20.001
Fruit, daily (%)30.026.620.40.001
Green or yellow vegetables, daily (%)34.934.228.50.001
Salted or dried fish, 3+/week (%)26.827.030.80.001
Salted cod roe or fish gut, 3+/week (%)35.837.846.10.001
Soybean paste soup, daily (%)76.877.278.10.17
Pickled vegetable, daily (%)39.041.346.50.001
Family history of gastric cancer (%)6.37.36.80.18
Body mass index24.0 (0.04)24.0 (0.04)23.2 (0.03)0.0001
Table III. Baseline Characteristics According to Alcohol Consumption in 19,227 Men, Japan, 1990–1999
 Alcohol consumption1p for difference3
0–3 days/month−161.0 g2162.0–322.0 g2322.5 g+2
  • 1

    Figures are means (SE) unless otherwise specified.

  • 2

    Weekly ethanol intake among those who drink at least once per week.

  • 3

    Based on 1-way ANOVA or χ2 test.

Number6074481741874149 
Age (years)49.6 (0.1)49.2 (0.1)49.5 (0.1)49.2 (0.1)0.0001
Current tobacco use (%)47.346.458.163.80.001
Fruit, daily (%)26.927.222.218.50.001
Green or yellow vegetables, daily (%)32.032.531.229.00.002
Salted or dried fish, 3+/week (%)22.527.434.034.40.001
Salted cod roe or fish gut, 3+/week (%)28.839.849.654.00.001
Soybean paste soup, daily (%)73.277.580.680.80.001
Pickled vegetable, daily (%)35.142.450.449.80.001
Family history of gastric cancer (%)5.56.68.07.80.001
Body mass index23.7 (0.04)23.6 (0.04)23.5 (0.04)23.5 (0.04)0.001

Table IV shows adjusted RRs and 95% CIs for gastric cancer at all sites by smoking and alcohol drinking. Smoking was related to an elevated risk of gastric cancer; RRs and 95% CIs for former smokers and current smokers compared to subjects who never smoked were 1.6 (1.1–2.4) and 1.7 (1.2–2.4), respectively. Smoking-related variables such as cigarettes/day, age started smoking and pack-years also appeared to be related to gastric cancer risk, though dose-response was not clear. Total alcohol consumption was not associated with gastric cancer risk.

Table IV. Adjusted RRs and 95% CIs by Smoking Habit and Alcohol Consumption for Gastric Cancer at All Sites
 Gastric cancer (all sites)
Number of casesIncidence rate (per 100,000 person-years)Adjusted RR1 (95% CI)
  • 1

    Adjusted for age, area, smoking habit (for analysis of alcohol drinking), alcohol drinking (for analysis of smoking habit), consumption of fruit, green or yellow vegetables, salted cod roe or fish gut, and body mass index.

  • 2

    Weekly ethanol intake among those who drank at least once/week.

  • 3

    Trend calculated among those who drank at least once/week.

Smoking status   
 Never41961.0
 Former691761.6 (1.1–2.4)
 Current1631771.7 (1.2–2.4)
Cigarettes/6 day   
 1–19401601.5 (0.9–2.3)
 20792202.0 (1.4–3.0)
 21+431431.5 (1.0–2.4)
 ptrend  0.63
Age started smoking (years) 
 21+572081.9 (1.3–2.9)
 20741731.6 (1.1–2.4)
 10–19321501.6 (1.0–2.6)
 ptrend  0.38
Pack-years   
 0.05–20.9541191.4 (0.9–2.1)
 21–33781811.9 (1.3–2.8)
 33.15+912321.7 (1.2–2.5)
 ptrend  0.26
Total alcohol   
 0–3 days/month681241.0
 0–161.0 g/week2541240.8 (0.6–1.2)
 162.0–322.0 g/week2772021.1 (0.8–1.5)
 322.5 + g/week2741971.1 (0.8–1.6)
 ptrend3  0.66

Table V shows adjusted RRs and 95% CIs of gastric cancer by subsite and histologic type according to smoking status. Compared to subjects who never smoked, both past smokers and current smokers showed an elevated risk for the differentiated type of distal gastric cancer, whereas no relationship was seen for the undifferentiated type of distal gastric cancer. For cardia cancer, current smokers had an adjusted RR of 2.4 (95% CI 0.8–7.1) compared to the group who never smoked. Findings on 2 histologic types of cardia cancer were comparable, so these types were combined for further analysis. Similar results were also seen for the number of cigarettes smoked per day or pack-years of smoking; when current smokers were separated by numbers of cigarettes smoked per day or pack-years of smoking, the differentiated type of distal gastric cancer showed an elevated risk, though neither trend was statistically significant. For the undifferentiated type of distal cancer, a negative association was observed for smoke-related variables, though it was not statistically significant.

Table V. Adjusted RRs and 95% CIs by Smoking Habit for Gastric Cancer According to Subsite and Histologic Type in Men
 Cardia and upper-third gastric cancerDistal gastric cancer
All histologic typesDifferentiated typeUndifferentiated type
Number of casesIncidence rate (per 100,000 person-years)Adjusted RR1 (95% CI)Number of casesIncidence rate (per 100,000 person-years)Adjusted RR1 (95% CI)Number of casesIncidence rate (per 100,000 person-years)Adjusted RR1 (95% CI)
  1. Adjusted for age, area, alcohol drinking, consumption of fruit, green or yellow vegetables, salted cod roe or fish gut and body mass index. Three and four subjects for cardia cancer and differentiated type of distal gastric cancer, respectively, had no information on pack-years and were deleted from the calculation of RRs automatically.

Smoking status         
 Never491.016381.018421.0
 Former7181.6 (0.5–5.5)33852.0 (1.1–3.7)18460.9 (0.5–1.8)
 Current24262.4 (0.8–7.1)75822.1 (1.2–3.6)27300.6 (0.3–1.1)
Cigarettes/day         
 1–196242.0 (0.6–7.2)19761.9 (1.0–3.7)11440.9 (0.4–1.8)
 2012343.0 (0.9–9.6)31872.2 (1.2–4.1)11310.6 (0.3–1.3)
 21+6202.2 (0.6–8.0)25832.6 (1.3–4.9)5130.3 (0.1–0.8)
   ptrend= 0.83  ptrend = 0.31  ptrend = 0.09
Age started smoking (years)         
 21+13484.2 (1.3–13.1)25922.2 (1.2–4.3)6220.4 (0.2–1.1)
 206141.2 (0.3–4.5)35822.2 (1.2–4.0)16380.7 (0.4–1.5)
 10–195242.4 (0.6–9.5)15702.2 (1.1–4.5)5240.5 (0.2–1.4)
   ptrend = 0.06  ptrend = 0.81  ptrend = 0.50
Pack-years         
 0.05–20.97151.7 (0.5–5.9)20441.3 (0.7–2.6)19421.0 (0.5–2.0)
 21.0–33.012282.8 (0.9–8.9)42982.7 (1.5–4.9)10230.5 (0.2–1.1)
 33.15+9231.6 (0.5–5.5)421082.2 (1.2–4.0)16410.6 (0.3–1.3)
   ptrend = 0.85  ptrend = 0.16  ptrend = 0.20
 

Table VI shows the risks associated with a combined estimate of alcohol consumption compared to a rate of alcohol consumption of 0–3 times/month. When adjustments were made for smoking status and other variables, cardia cancer appeared to be associated with alcohol drinking, though it failed to reach significance; adjusted RRs for weekly ethanol intake of 2.7–161.0, 162.0–322.0 and 322.5+ g compared to 0–3 days of drinking/month were 2.5 (95% CI 0.7–9.5), 3.3 (95% CI 0.9–11.6) and 3.0 (95% CI 0.8–11.1), respectively (ptrend = 0.66). For distal gastric cancer, no relationship with alcohol consumption was seen for either histologic type.

Table VI. Adjusted RRs and 95% CIs by Alcohol Consumption for Gastric Cancer According to Subsite and Histologic Type in Men
 Cardia and upper-third gastric cancerDistal gastric cancer
All histologic typesDifferentiated typeUndifferentiated type
Number of casesIncidence rate (per 100,000 person-years)Adjusted RR1 (95% CI)Number of casesIncidence rate (per 100,000 person-years)Adjusted RR1 (95% CI)Number of casesIncidence rate (per 100,000 person-years)Adjusted RR1 (95% CI)
  • 1

    Adjusted for age, area, smoking habit, consumption of fruit, green or yellow vegetables, salted cod roe or fish gut and body mass index.

  • 2

    Amount of ethanol intake (g/week) among those who drank at least once per week.

  • 3

    Trend calculated among those who drank at least once/week.

Total alcohol consumption         
 0–3 days/month361.032591.017311.0
 0–161.028182.5 (0.7–9.5)27620.9 (0.5–1.5)11250.7 (0.3–1.4)
 162.0–322.0213343.3 (0.9–11.6)381001.1 (0.7–1.8)15400.9 (0.5–1.9)
 322.5+211303.0 (0.8–11.1)27730.9 (0.5–1.5)20541.3 (0.7–2.6)
   ptrend3 = 0.66  ptrend3 = 1.00  ptrend3 = 0.07
 

Major results did not change in magnitude when the first year of follow-up was deleted from the analysis. We also examined the interaction of cigarette smoking and alcohol drinking and found no measurable effect.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

In our study, current smokers appeared to have an elevated risk for differentiated types of distal gastric cancer. Few studies have evaluated the association of smoking and gastric cancer risk, taking histologic type into account, and all of them have been case-control studies.22, 23, 24, 25 Of 2 studies from Japan in which gastric cancers were stratified by histologic type only, 1 showed a slightly higher risk in differentiated type of gastric cancer among male habitual smokers;22 in the other study, several environmental factors, such as intake of green or yellow vegetables, fruit and meat, and a family history of gastric cancer were associated with the differentiated type of gastric cancer among women, while the difference between both histopathologic types was not clear for men. As for smoking habits, results among men showed higher risk estimates for the undifferentiated type than for the differentiated type in 2 of 3 smoking categories; results among women were essentially null for both histopathologic types.23 Studies from Poland24 and Sweden25 were stratified by both tumor subsite and histologic type. The former found no association between smoking and either histologic type, and the latter actually reported higher cigarette smoking risk estimates for the undifferentiated type than for the differentiated type of noncardia tumor. The clear difference in risk pattern by histologic type supports the hypothesis derived from former studies that the differentiated type of gastric cancer may be more closely related to environmental factors.8, 9 We cannot rule out the possibility that chance could explain the difference in association by histologic type observed in this study. The CIs overlap considerably, and the precision of relative risk estimates is only moderate. The observed statistically significant negative association between heavy smoking (21+ cigarettes/day) and undifferentiated type of distal gastric cancer may be a chance finding because we cannot explain why smoking reduced the risk of cancer. Cigarette smoking also tended to be associated with an elevated risk of cardia and upper-third gastric cancer, in line with several studies26, 27, 28, 29, 30, 31, 32 but not all.33, 34, 35 The inconsistencies among studies may be due, to some extent, to different levels of misclassification of cardia cancers, such as the recent introduction of a separate diagnostic code, the lack of consensus for a definition of cardia and an increased interest in cardia cancer.3, 36

Several mechanisms are suggested to explain the relationship between cigarette smoking and gastric cancer. Nitrosamines and other nitroso compounds present in cigarette smoke may be involved in the process of gastric carcinogenesis.37 Cigarette smoking was related to an elevated risk of transition to dysplasia and intestinal metaplasia,38 which are important in the process of gastric carcinogenesis in the model postulated by Correa.39 A randomized crossover study showed that smoking delayed both stomach emptying and alcohol absorption.40 Although it is still inconclusive, cigarette smoking was also reported to have an association with gastroesophageal reflux disease, which is highly related to the occurrence of cardia cancer.41 The lack of an obvious dose–response relationship may be due to the relatively small sample size for differentiated types of distal gastric cancer that can be truly associated with smoking. Another possible explanation is that cigarette smoke does not require a large dose to be carcinogenic in the presence of genetic damage. The residual confounding effect of smoking may still exist even when controlling for potential confounding factors.

Intake of alcoholic beverages was not related to any increased risk of gastric cancer of all subsites combined, in line with many studies.42, 43, 44, 45, 46, 47, 48, 49, 50, 51 When the lesion was restricted to the cardia or upper third of the stomach, a positive, though nonsignificant, association between alcohol consumption and cancer was observed. Because alcohol consumption is strongly related to an elevated risk of upper digestive tract cancer (oral cavity, pharynx and esophagus), the finding of cancer in the most proximal part of the stomach in our study appears to be reasonable. Few studies have evaluated the relationship between alcohol consumption and gastric cancer risk by anatomic subsite. Although some studies have shown a significant26, 52 or a nonsignificant53, 54 elevated risk for cardia cancer, most have failed to show such an association.31–33, 35, 54, 55 Thus far, including studies focusing on cardia cancers, alcohol use has been unrelated to the risk of gastric cancer in most studies. Our results for cardia cancer contradict these previous findings, though much more attention is needed to interpret the present findings. For detecting the role of a specific type of alcoholic beverage, more study subjects are needed. Besides the misclassification of cardia cancer mentioned above, the relatively small number of cardia cancer cases may not be adequate to detect precise results.

In addition to misclassification of cardia cancer, the possibility of histologic misclassification cannot be ruled out. Histologic typing is inherently subjective, lacking “gold standard” measures, and depends on the judgment of the pathologist.56 In Japan, although gastric cancer has generally been classified according to a Japanese system,57 the present study had no central review system and thus may have misclassified the tumors, potentially resulting in inconclusive or erroneous findings. However, a review of pathologic slides by even 1 pathologist in a subsample was not practical in this large multicenter study. Although the resulting misclassification would lead to the reduction in differences in relative risk estimates between 2 histologic types, the difference in relative risks of smoking observed between 2 histologic types in our study was distinct.

Accumulating data show that H. pylori infection is closely associated with increased risk of gastric cancer.58, 59 Prevalence rates of H. pylori IgG antibody among randomly selected men aged 40–49 years were 76% in Ninohe (n = 131), 86% in Yokote (n = 133), 72% in Saku (n = 118) and 63% in Ishikawa (n = 128) PHC areas in our previous ecologic study in 1989–1990,60 which were almost parallel with age-adjusted incidence rates of gastric cancer in each area. Moreover, in the cross-sectional analysis of our ecologic study, smoking status and alcohol drinking were not associated with H. pylori infection, while frequent intake of pickled vegetable and miso soup were associated. Thus, we consider that the need to treat H. pylori infection as a confounding factor is less in this cohort.

Among the several advantages of our current study is the prospective design, which diminishes the probability of recall bias and probably reduces nondifferential misclassification of exposures in the time window relevant for the induction of gastric cancer.

To conclude, in this population-based prospective study, we hypothesized that smoking was related to gastric cancer, especially the differentiated type. There was no indication that alcohol consumption has an association with gastric cancer, except cardia cancer. The results support the hypothesis that the differentiated type of gastric cancer may be more affected by environmental factors than the undifferentiated type. Further studies are needed before the relationship between cigarette smoking or alcohol consumption and cardia cancer can be delineated.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

We thank Ms. Y. Iwatare, Ms. M. Takahashi and Ms. M. Hosokawa, who managed the database of this project. We also thank Drs. S. Yamamoto, T. Sobue and A. Ochiai for advice. Our study was supported by grants-in-aid for cancer research and for the Second-Term Comprehensive Ten-Year Strategy for Cancer Control from the Ministry of Health and Welfare of Japan. S.S. received a Research Resident Fellowship from the Foundation for the Promotion of Cancer Research in Japan.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES