Cigarette smoking and the risk of gastric cancer: A pooled analysis of two prospective studies in Japan

Authors

  • Yayoi Koizumi,

    Corresponding author
    1. Division of Epidemiology, Department of Public Health and Forensic Medicine, Tohoku University Graduate School of Medicine, Sendai, Japan
    2. Departments of Psychiatry, Tohoku University Graduate School of Medicine, Sendai, Japan
    • Division of Epidemiology, Department of Public Health and Forensic Medicine, Tohoku University Graduate School of Medicine, Seiryou-chou 2-1 Aoba-ku, Sendai city, Miyagi Prefecture, 980-8575, Japan
    Search for more papers by this author
    • Fax: +81-22-717-8125

  • Yoshitaka Tsubono,

    1. Division of Epidemiology, Department of Public Health and Forensic Medicine, Tohoku University Graduate School of Medicine, Sendai, Japan
    Search for more papers by this author
  • Naoki Nakaya,

    1. Division of Epidemiology, Department of Public Health and Forensic Medicine, Tohoku University Graduate School of Medicine, Sendai, Japan
    Search for more papers by this author
  • Shinichi Kuriyama,

    1. Division of Epidemiology, Department of Public Health and Forensic Medicine, Tohoku University Graduate School of Medicine, Sendai, Japan
    Search for more papers by this author
  • Daisuke Shibuya,

    1. Miyagi Cancer Society, Sendai, Japan
    Search for more papers by this author
  • Hiroo Matsuoka,

    1. Departments of Psychiatry, Tohoku University Graduate School of Medicine, Sendai, Japan
    Search for more papers by this author
  • Ichiro Tsuji

    1. Division of Epidemiology, Department of Public Health and Forensic Medicine, Tohoku University Graduate School of Medicine, Sendai, Japan
    Search for more papers by this author

Abstract

To examine the association between cigarette smoking and the risk of gastric cancer, we conducted a pooled analysis of 2 population-based prospective cohort studies in rural northern Japan. Cohort 1 included 9,980 men (≥40 years old) and Cohort 2 included 19,412 men (40–64 years old). The subjects completed a self-administered questionnaire on cigarette smoking and other health habits. We identified 228 cases of gastric cancer among Cohort 1 subjects (9 years of follow-up with 74,073 person-years) and 223 among Cohort 2 subjects (7 years of follow-up with 141,675 person-years). From each cohort, we computed the relative risk (RR) and 95% confidence interval (CI) of gastric cancer associated with smoking using a Cox regression analysis and pooled these estimates to obtain summary measures. The pooled multivariate RRs (95% CIs) for current smokers and past smokers compared to subjects who had never smoked were 1.84 (1.39–2.43) and 1.77 (1.29–2.43), respectively. The higher number of cigarettes smoked per day among current smokers was associated with a linear increase in risk (trend p < 0.05). The significant increase in risk for past smokers remained for up to 14 years after cessation. An increased risk was noted for cancer of the antrum but not for cardia or body lesions. The risk was increased for both differentiated and nondifferentiated histologic subtypes. Our findings support the hypothesis that cigarette smoking is a risk factor for gastric cancer. © 2004 Wiley-Liss, Inc.

An increased risk of gastric cancer associated with tobacco smoking has long been controversial.1 Although the International Agency for Research on Cancer Working Group concluded in 2002 that there was “sufficient” evidence of causality,2 it remains to be clarified if the effects of smoking differ by anatomic subsite or histologic subtype of gastric cancer. Furthermore, the majority of available evidence on smoking and gastric cancer are based on case-control studies3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75 that are susceptible to selection and recall biases, and most prospective studies have methodologic limitations including the use of mortality rather than incidence as an endpoint,76, 78, 79, 81, 82, 84, 85, 87, 88, 89 the use of a relatively small number (<300) of gastric cancer cases77, 79, 80, 81, 82, 83, 85, 86, 87, 89 or no adjustment for dietary variables as potential confounders.78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89

To further examine the association between cigarette smoking and the risk of gastric cancer, we conducted a pooled analysis of 2 population-based prospective cohort studies in rural northern Japan, a high-risk area for gastric cancer. Our study involved 451 incidence cases of gastric cancer, and we adjusted for various potential confounders, including dietary variables. We also examined the association according to the anatomic subsite and histologic subtype of gastric cancer.

MATERIAL AND METHODS

Study cohorts

The study designs for the 2 cohort studies have been described in detail elsewhere.90, 91, 92 Briefly, Cohort 1 was started in January 1984, when we delivered a self-administered questionnaire to 33,453 men and women (40 years of age or older) in 3 municipalities of Miyagi Prefecture. Usable questionnaires were returned from 31,345 subjects (93.7%). For Cohort 2, we delivered a self-administered questionnaire from June to August 1990 to 51,921 men and women (40–64 years of age) in 14 municipalities of the Prefecture. Usable questionnaires were returned from 47,605 subjects (91.7%). Study protocols for the 2 cohorts were approved by the institutional review board of the Tohoku University Graduate School of Medicine. We considered the return of self-administered questionnaires signed by the subjects to imply their consent to participate in the study.

Exposure data

In both cohorts, the questionnaires asked if subjects had never smoked, were current smokers or were past smokers. Current smokers were asked about the number of cigarettes they consumed per day. Past smokers were asked about the age at which they initiated smoking and the age at which they quit smoking (Cohort 1) or the number of years since they had quit (Cohort 2).

We divided current smokers into 3 categories according to the number of cigarettes consumed per day (1–19, 20–24 and >25/day). Also, we divided past smokers into 3 categories according to the length of time since they had quit smoking (<5 years, 5–14 years and >15 years).

Follow-up

For both cohorts, we used population registries of the municipalities to obtain information on the vital and residential status of each subject. We ascertained the incidence of cancer using record linkage to the Miyagi Prefectural Cancer Registry covering the study areas.93 In this cancer registry, the proportion of gastric cancer cases for which information was available only from death certificates was 5% in men and 8% in women.93 A follow-up was conducted from 1 January 1984 to 31 December 1992 for Cohort 1 and from 1 August 1990 to 31 March 1997 for Cohort 2.

Since the prevalence of smoking was low among women in the study cohorts (6.2% or 722 women in Cohort 1 and 5.7% or 982 women in Cohort 2), we limited the analysis to men (13,992 in Cohort 1; 22,836 in Cohort 2). Of these men, we excluded cancer cases prevalent at the baseline (258 in Cohort 1 and 424 in Cohort 2). We also excluded the men who did not respond fully to the question on smoking status (3,754 in Cohort 1 and 3,000 in Cohort 2). Consequently, our analyses included 9,980 men with a total of 228 gastric cancer cases in Cohort 1 and 19,412 men with a total of 223 gastric cancer cases in Cohort 2.

We classified the gastric cancer cases according to the anatomic subsite of the primary lesion (cardia, body, antrum and overlapping or unknown), as well as the histologic subtype (differentiated, nondifferentiated and unclassified or unknown). The histologic classification was based on Inoue's method,14 in which “differentiated” and “nondifferentiated” types correspond to intestinal and diffuse types, respectively, by Lauren's classification.94 Information on subsite and histology of gastric cancer cases were based on reports from institutions (to the cancer registry), and no independent review of clinical or histologic records were conducted.

Statistical analysis

We counted the number of person-years of follow-up for each subject from the beginning of follow-up until the date of diagnosis of gastric cancer; the date of emigration outside the study districts; the date of death; or the end of the follow-up period, whichever occurred first. Total person-years accrued were 74,073 for Cohort 1 and 141,675 for Cohort 2. We used a Cox proportional-hazards regression analysis to estimate the relative risks (RRs) and 95% confidence intervals (CIs) of gastric cancer according to the levels of cigarette smoking, with adjustment for potentially confounding variables.

For the 2 cohorts, we considered the following variables as potential confounders: age, body mass index, prior history of peptic ulcer; parental history of gastric cancer; type of health insurance; alcohol consumption; and consumption frequencies of pickled vegetables and bean-paste soup. We adjusted for consumption frequencies of pickled vegetables and bean-paste soup because they were the major food sources of sodium in the study population. We adjusted for types of health insurance as a measure of socioeconomic status because the people in Japan are enrolled into one of several types of public health insurances based on their occupation. For Cohort 1, we further adjusted for consumption frequencies of green or yellow vegetables, other vegetables and fruits. For Cohort 2, we further adjusted for consumption frequencies of spinach, carrot and pumpkin, cabbage and lettuce, Chinese cabbage, oranges and other fruits. We did not adjust for education because this information is available only for Cohort 2 but not for Cohort 1. For Cohort 2, we observed basically the same results whether or not level of education (primary, secondary and tertiary) was adjusted for.

To obtain a summary measure of the results from Cohort 1 and Cohort 2, the general variance-based method was used to combine each RR and 95% CI.95 The p-values for the analysis of linear trends were also calculated by treating the number of cigarettes consumed per day as continuous variables (never smokers were coded as 0, and past smokers were excluded). All reported p-values are 2-tailed.

RESULTS

Table I compares the characteristics of men according to their smoking status. In Cohort 1, compared to subjects who had never smoked, current smokers were more likely to have a history of peptic ulcer, drink alcohol and were less likely to consume fruits and vegetables. We observed a similar tendency for men in Cohort 2.

Table I. Characteristics of the Subjects According to Smoking Status
 Smoking status
NeverCurrentPast
Characteristics for Cohort 1   
 No. of subjects2,3405,4632,177
 Mean age ± SD56.8 ± 12.454.7 ± 9.958.3 ± 11.3
 Mean body mass index ± SD23.6 ± 3.022.9 ± 2.923.3 ± 3.1
 History of peptic ulcer (%)8.415.515.3
 Parental history of gastric cancer (%)8.49.58.6
 History of peptic ulcer (%)8.415.515.3
 Current use of alcohol (%)71.481.372.7
 Daily dietary consumption (%)   
  Pickled vegetables66.369.461.5
  Bean-paste soup89.287.084.6
  Green-yellow vegetables56.047.153.5
  Other vegetables69.363.566.9
  Fruits64.045.759.4
Characteristics for Cohort 2   
 No. of subjects3,99712,1183,297
 Mean age ± SD51.7 ± 7.151.0 ± 7.653.0 ± 7.7
 Mean body mass index ± SD24.0 ± 2.823.3 ± 2.824.0 ± 2.7
 History of peptic ulcer (%)14.621.822.6
 Parental history of gastric cancer (%)9.38.18.9
 Health insurance (the national health insurance) (%)54.651.151.3
 Current use of alcohol (%)72.079.075.2
 Daily dietary consumption (%)   
  Pickled vegetables51.356.553.8
  Bean-paste soup87.486.088.6
  Spinach21.619.122.2
  Carrot and pumpkin10.48.510.3
  Cabbage and lettuce17.216.417.8
  Chinese cabbage14.514.913.8
  Oranges21.516.019.8
  Other fruits25.119.225.4

Table II presents the RRs and 95% CIs for gastric cancer according to smoking status. We found a significant increase in risk for both current smokers and past smokers. The pooled multivariate RRs (95% CI) for current smokers and past smokers, in comparison to subjects who had never smoked, were 1.84 (1.39–2.43) and 1.77 (1.29–2.43), respectively. We observed similar results between Cohort 1 and Cohort 2, and the results did not change materially between age-adjusted and multivariate-adjusted analyses. We also observed similar results after excluding gastric cancer cases diagnosed in the first 3 years of follow-up (73 cases in Cohort 1; 92 cases in Cohort 2).

Table II. Relative Risks (RRs) and 95% Confidence Intervals (95% CIs) of Gastric Cancer According to Smoking Status1
 Smoking status
NeverCurrentPast
  • 1

    The multivariate relative risk (RR1) has been adjusted for age (in years); body mass index (<18.5, 18.5–25 or >25); history of peptic ulcer; parental history of gastric cancer; type of health insurance (5 categories); alcohol consumption (never drank alcohol, drank in the past, currently drink less often than daily or currently drink daily); daily consumption of pickled vegetables and consumption of bean-paste soup (less than 1 day/week, 1–2 days/week, 3–4 days/week or daily). The multivariate RR1 for cohort 1 has also been adjusted for consumption of green or yellow vegetables, other vegetables and fruits (less than 1 day/week, 1–2 days/week, 3–4 days/week or daily). The multivariate RR1 for cohort 2 has also been adjusted for consumption of spinach, carrot and pumpkin, cabbage and lettuce, Chinese cabbage, oranges and other fruits (less than 1 day/week, 1–2 days/week, 3–4 days/week or daily). RR1 denotes the relative risk with all cases of gastric cancer included in the multivariate analysis, and RR2 has been calculated after the exclusion of gastric cancer cases diagnosed in the first 3 years of follow-up. Values in parentheses are 95% confidence intervals.

No. of cases/person-years   
 Cohort 137/17,770132/40,62759/15,676
 Cohort 228/29,389151/88,33044/23,956
Age-adjusted RR (95% CI)   
 Cohort 11.001.83 (1.27–2.65)1.78 (1.18–2.68)
 Cohort 21.001.84 (1.23–2.76)1.66 (1.03–2.67)
 Pooled1.001.84 (1.40–2.41)1.73 (1.26–2.35)
Multivariate RR1 (95% CI)   
 Cohort 11.001.91 (1.31–2.79)1.89 (1.24–2.87)
 Cohort 21.001.76 (1.17–2.66)1.63 (1.01–2.63)
 Pooled1.001.84 (1.39–2.43)1.77 (1.29–2.43)
Multivariate RR2 (95% CI)   
 Cohort 11.002.07 (1.31–3.28)1.91 (1.15–3.19)
 Cohort 21.002.58 (1.41–4.75)1.95 (0.97–3.95)
 Pooled1.002.24 (1.55–3.24)1.93 (1.27–2.91)

For current smokers, we observed a linear increase in risk associated with the higher number of cigarettes smoked per day (Table III): The pooled multivariate RRs (95% CIs) for current smokers who smoked 1–19, 20–24 and >25 cigarettes per day in comparison to never smokers were 1.41 (1.00–1.98), 1.98 (1.45–2.71) and 2.15 (1.53–3.02), respectively (trend p < 0.05). For past smokers, a significant increase in risk remained for up to 14 years after cessation: The pooled multivariate RRs (95% CIs) for past smokers who had quit <5, 5–14 and >15 years previously compared to subjects who had never smoked were 1.72 (1.12–2.64), 2.08 (1.41–3.07) and 1.31 (0.77–2.21), respectively. We observed a linear increase in risk associated with pack-years of smoking: The pooled multivariate RRs (95% CIs) for <25, 25–39 and >40 pack-years of smoking in comparison to never smokers were 1.55 (1.07–2.25), 2.20 (1.56–3.11) and 2.26 (1.61–3.18), respectively (trend p < 0.001).

Table III. Pooled Relative Risks (RRs) and 95% Confidence Intervals (CIs) of Gastric Cancer According to Smoking Status1
 Cohort 1Cohort 2Age-adjusted RR (95% CI)Multivariate RR (95% CI)
No. of cases/person-yearsNo. of cases/person-years
  • 1

    The multivariate relative risk (RR) has been adjusted for age (in years); body mass index (<18.5, 18.5–25 or >25); history of peptic ulcer; parental history of gastric cancer; type of health insurance (5 categories); alcohol consumption (never drank alcohol, drank in the past, currently drink less often than daily or currently drink daily); daily consumption of pickled vegetables and consumption of bean-paste soup (less than 1 day/week, 1–2 days/week, 3–4 days/week or daily). The multivariate RR for cohort 1 has also been adjusted for consumption of green or yellow vegetables, other vegetables and fruits (less than 1 day/week, 1–2 days/week, 3–4 days/week or daily). The multivariate RR for cohort 2 has also been adjusted for consumption of spinach, carrot and pumpkin, cabbage and lettuce, Chinese cabbage, oranges and other fruits (less than 1 day/week, 1–2 days/week, 3–4 days/week or daily). Values in parentheses are 95% confidence intervals.

Never smokers37/17,77028/29,3891.001.00
Current smokers    
 Total132/40,627151/88,3301.84 (1.40–2.41)1.84 (1.39–2.43)
 No. of cigarettes consumed/day    
  1–1938/13,33536/22,3151.42 (1.02–1.99)1.41 (1.00–1.98)
  20–2459/15,87064/35,2782.03 (1.50–2.75)1.98 (1.45–2.71)
  ≥2535/11,42251/30,7372.18 (1.56–3.04)2.15 (1.53–3.02)
 p for trend  <0.05<0.05
Past smokers    
 Total59/15,67644/23,9561.73 (1.26–2.35)1.77 (1.29–2.43)
 Years since quitting    
  <519/6,53416/7,4771.64 (1.08–2.49)1.72 (1.12–2.64)
  5–1428/5,77320/11,2922.07 (1.43–3.01)2.08 (1.41–3.07)
  ≥1512/3,3698/5,1871.26 (0.76–2.11)1.31 (0.77–2.21)

Table IV shows the pooled multivariate RRs and 95% CIs of gastric cancer by anatomic subsite and histologic subtype. Compared to subjects who had never smoked, current smokers had a significant increase in risk for cancer of the antrum but not for cancer of the cardia or body. With regard to the histologic subtype, we observed a significant increase in risk associated with current smoking for both differentiated and nondifferentiated cancers.

Table IV. Pooled Multivariate Relative Risks (RRs) and 95% Confidence Intervals (CIs) of Each Anatomic Subsite and Histologic Subtype of Gastric Cancer According to Smoking Status1
 Smoking status
NeverPastCurrentp for trend
Total1–19 cigarettes/day20–24 cigarettes/day≥25 cigarettes/day
  • 1

    The multivariate relative risk (RR) has been adjusted for age (in years); body mass index (<18.5, 18.5–25 or >25); history of peptic ulcer; parental history of gastric cancer; type of health insurance (5 categories); alcohol consumption (never drank alcohol, drank in the past, currently drink less often than daily or currently drink daily); daily consumption of pickled vegetables and consumption of bean-paste soup (less than 1 day/week, 1–2 days/week, 3–4 days/week or daily). The multivariate RR for Cohort 1 has also been adjusted for consumption of green or yellow vegetables, other vegetables and fruits (less than 1 day/week, 1–2 days/week, 3–4 days/week or daily). The multivariate RR for Cohort 2 has also been adjusted for consumption of spinach, carrot and pumpkin, cabbage and lettuce, Chinese cabbage, oranges and other fruits (less than 1 day/week, 1–2 days/week, 3–4 days/week or daily). Values in parentheses are 95% confidence intervals.

Anatomic subsite       
 Cardia       
  No. of cases of gastric cancer121240131314 
  RR (95% CI)1.001.17 (0.52–2.64)1.29 (0.64–2.61)1.42 (0.63–3.23)1.09 (0.45–2.62)1.74 (0.70–4.33)0.27
 Body       
  No. of cases of gastric cancer172662152819 
  RR (95% CI)1.001.62 (0.84–3.13)1.22 (0.69–2.15)1.15 (0.54–2.47)1.49 (0.79–2.82)1.47 (0.74–2.95)0.10
 Antrum       
  No. of cases of gastric cancer152877153527 
  RR (95% CI)1.002.08 (1.10–3.95)2.00 (1.13–3.55)1.19 (0.57–2.48)2.34 (1.24–4.40)2.79 (1.42–5.48)<0.05
Histologic type       
 Differentiated       
  No. of cases of gastric cancer3860147376347 
  RR (95% CI)1.001.77 (1.17–2.67)1.61 (1.11–2.32)1.20 (0.76–1.91)1.70 (1.12–2.58)1.97 (1.26–3.09)<0.05
 Nondifferentiated       
  No. of cases of gastric cancer111656132815 
  RR (95% CI)1.001.56 (0.72–3.41)2.10 (1.08–4.08)1.66 (0.73–3.78)2.61 (1.27–5.35)2.05 (0.92–4.59)<0.05

In the analyses stratified by age; body mass index; alcohol consumption; type of health insurance; parental history of gastric cancer; history of peptic ulcer; and consumption of pickled vegetables and bean-paste soup, we did not observe a substantial modification by these variables on the association between cigarette smoking and the risk of gastric cancer.

DISCUSSION

We found a significant positive association between smoking and the risk of gastric cancer in current smokers and past smokers in comparison to subjects who had never smoked. We also found a significant dose-response association between the number of cigarettes consumed per day and the risk. The significant increase in risk among past smokers remained for up to 14 years after smoking cessation.

Of 72 case-control studies we identified that examined the association between cigarette smoking and gastric cancer,3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75 40 studies observed a significant increase in the risk among current smokers or subjects who had once smoked in comparison to subjects who had never smoked.3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43 Of 15 prospective cohort studies,76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 96 10 studies observed a significant increase in risk among current smokers in comparison to subjects who had never smoked.76, 77, 78, 79, 80, 81, 82, 83, 84, 96 The results of our study are consistent with the 40 case-control and 10 cohort studies.

With regard to anatomic subsite, we observed a significant increase in risk associated with smoking only for cancer of the antrum but not for cancer of the cardia or body. Of 22 studies examining the association with gastric cancer according to anatomic subsite (20 case-control studies3, 5, 6, 7, 9, 11, 13, 15, 17, 18, 20, 21, 23, 24, 31, 32, 48, 52, 54, 58 and 2 cohort studies77, 96), 14 studies3, 7, 9, 11, 13, 15, 17, 18, 21, 23, 24, 31, 32, 96 found a significant increase in risk for cardia lesions, whereas 13 studies5, 6, 7, 9, 13, 15, 17, 20, 23, 24, 31, 77, 96 observed a significant increase in risk for noncardia lesions. Although the studies are not consistent, our findings suggest that the antrum may be more susceptible to the carcinogenic effects of smoking than the more proximal areas of the stomach.

Of 8 studies examining the association with gastric cancer according to histologic subtype (7 case-control studies6, 7, 14, 15, 29, 54, 60 and 1 cohort study77), 6 studies6, 7, 14, 15, 29, 77 observed a significant increase in risk for intestinal or differentiated cancers, whereas 5 studies6, 7, 14, 15, 29 found a significant increase in risk for diffuse or nondifferentiated cancers. We observed a positive association for both histologic subtypes.

The methodologic advantages of our study over prior studies include the use of a prospective study design, the use of incidence (rather than mortality) as an endpoint and the large number of cases accrued that allowed for analyses according to anatomic subsite and histologic subtype. We also calculated the risks with adjustment for various covariates, including dietary variables.

Our study has several potential limitations. First, we did not obtain information on subjects' history of infection with Helicobacter pylori, an established risk factor for gastric cancer.97 However, a recent study showed that current smoking was associated with a similar extent of increased risk of gastric cancer for subjects with or without Helicobacter pylori infection.8 Further study is necessary to examine potential confounding and effect modification by Helicobacter pylori on the association between smoking and the risk of gastric cancer.

Second, information on subsite and histology of gastric cancer cases were based on reports from institutions to the cancer registry, and no independent review of clinical or histologic records were conducted. The lack of independent review would have resulted in some misclassification of cases. Nevertheless, the proportion of cardia cancers in men observed in our study (14.1%) was comparable to the proportions (12.6% and 18.8%) reported in Japanese studies examining the association between smoking and the risk of gastric cancer according to subsite.23, 77 Similarly, the proportion of differentiated cancers in men observed in our study (54.3%) was comparable to the proportions (53.6% and 61.8%) reported in the Japanese studies.14, 77 These findings suggest that the distributions of anatomic subsite and histologic subtype were generally comparable to those reported in other studies conducted in Japan.

Third, because information on smoking and other variables were based on self-administered questionnaires, some misclassification of subjects was inevitable. Nevertheless, because this information was collected before subjects developed gastric cancer and other serious diseases, the misclassification of smoking status would likely have been nondifferential and resulted in conservative estimates for the association between smoking and the risk of gastric cancer.

Fourth, we excluded from the analyses 28.7% of men in Cohort 1 (4,012 of 13,992) and 15.0% of men in Cohort 2 (3,424 of 22,836) who responded to self-administered questionnaires but provided incomplete information on smoking or had prevalent cancers. However, we observed little differences between men who remained in the analytic cohorts and men who formed the total cohorts. Specifically, for men in Cohort 1, the mean ages (SDs) were 56.6 + 11.1 and 56.7 + 11.1, the prevalence of past history of peptic ulcer were 13.8% and 13.5%, and the prevalence of family history of gastric cancer were 9.0% and 8.3%, respectively. We observed similar tendencies for men in Cohort 2. These findings indicate that the men remaining for the current analyses were comparable in characteristics to the men in the total cohorts.

In conclusion, our pooled analyses of 2 population-based prospective cohort studies in rural northern Japan support the hypothesis that cigarette smoking is a risk factor for gastric cancer.

Ancillary