Gastric cancer screening and subsequent risk of gastric cancer: A large-scale population-based cohort study, with a 13-year follow-up in Japan

Authors

  • Kyung-Jae Lee,

    1. Epidemiology and Prevention Division, Research Center for Cancer Prevention and Screening, National Cancer Center, Tokyo, Japan
    2. Department of Preventive Medicine, Soonchunhyang University College of Medicine, Seoul, South Korea
    Search for more papers by this author
  • Manami Inoue,

    Corresponding author
    1. Epidemiology and Prevention Division, Research Center for Cancer Prevention and Screening, National Cancer Center, Tokyo, Japan
    • Epidemiology and Prevention Division, Research Center for Cancer Prevention and Screening, National Cancer Center, 5-1-1 Tsukiji, Chuo-ku, Tokyo 104-0045, Japan
    Search for more papers by this author
    • Fax: +81-3-3547-8578.

  • Tetsuya Otani,

    1. Epidemiology and Prevention Division, Research Center for Cancer Prevention and Screening, National Cancer Center, Tokyo, Japan
    Search for more papers by this author
  • Motoki Iwasaki,

    1. Epidemiology and Prevention Division, Research Center for Cancer Prevention and Screening, National Cancer Center, Tokyo, Japan
    Search for more papers by this author
  • Shizuka Sasazuki,

    1. Epidemiology and Prevention Division, Research Center for Cancer Prevention and Screening, National Cancer Center, Tokyo, Japan
    Search for more papers by this author
  • Shoichiro Tsugane

    1. Epidemiology and Prevention Division, Research Center for Cancer Prevention and Screening, National Cancer Center, Tokyo, Japan
    Search for more papers by this author

Abstract

We prospectively investigated the association between gastric cancer screening and subsequent risk of gastric cancer in a large-scale population-based prospective cohort study, with a 13-year follow-up in Japan. Data were analyzed from a population-based cohort of 42,150 (20,326 men and 21,824 women) subjects. Approximately 36% of subjects reported that they had undergone screening photofluorography during the preceding 12 months, and were regarded as the screened group. A total of 179 gastric cancer deaths and 636 incident gastric cancers were identified during the follow-up period. We observed a 2-fold decrease in gastric cancer mortality in screened versus unscreened subjects (RR = 0.52; 95% CI = 0.36–0.74). The extent of the reduction in mortality for gastric cancer was greater than in death from all causes excluding gastric cancer (RR = 0.71; 95% CI = 0.65–0.78). A significant decrease in the incidence of advanced gastric cancer was observed in screened subjects (RR = 0.75; 95% CI = 0.58–0.96), although the overall incidence rate did not differ significantly between the screened and unscreened subjects (RR = 1.06; 95% CI = 0.90–1.25). In age-stratified analyses, a significant reduction in gastric cancer mortality was seen in screened subjects aged 40–49 years at baseline, compared with a lesser reduction in screened subjects aged 50–59 (RR = 0.30, 95% CI = 0.13–0.72; and RR = 0.60, 95% CI = 0.40–0.88, respectively). These findings suggest that gastric cancer screening may be associated with a reduced risk of mortality from gastric cancer. © 2005 Wiley-Liss, Inc.

Although the incidence and mortality of gastric cancer in Japan is falling, in line with a worldwide decrease in this condition,1, 2 it remains the second most frequent cause of cancer death among Japanese men and women, with an estimated 50,000 deaths from gastric cancer in 2000.2 The prevention of gastric cancer is therefore still one of the most important cancer control strategies.

In Japan, gastric cancer screening by photofluorography began in Miyagi Prefecture, around 1960.3 It was subsequently expanded through implementation of the National Health and Medical Services Law for the Aged in 1983, and is now an established cancer control strategy nationwide. National statistics indicate that 4.3 million inhabitants (13% of those aged 40 years and over) participated in gastric cancer screening organized by local governments in 2002.4

Although randomized controlled trials represent the most reliable method for evaluating the impact of screening on cancer risk, no such intervention studies are available in Japan. The efficacy of gastric cancer screening has instead been assessed using other epidemiological methods5, 6, 7, 8, 9, 10, 11, 12 over the last 20 years, but results have been controversial (Table I). Although most case-control studies5, 6, 7 suggested a 40–60% decrease in gastric cancer mortality in screened versus unscreened subjects, previous prospective studies have shown inconsistent results.8, 9, 10, 11 Moreover, although more recent results suggested that the screened subjects had a healthier lifestyle than the unscreened subjects,10, 11 most previous studies did not adjust for lifestyle factors that may be associated with gastric cancer.5, 6, 7, 8, 9

Table I. Summary of Epidemiologic Studies, Evaluating The Impact of Gastric Cancer Screening on Mortality and Incidence of Gastric Cancer
ReferenceStudy populationFollow-upAgeNo. of subjectsEndpointResults
  • 1

    Conducted in Japan.

  • 2

    Odds ratio (OR) for men.

  • 3

    OR for women.

  • 4

    Conducted in Venezuela.

  • 5

    OR for both sexes.

  • 6

    Ratio of observed to expected number of cases (O/E ratio) for mortality.

  • 7

    O/E ratio for incidence.

  • 8

    Standardized mortality rate for men.

  • 9

    Standardized mortality rate for women.

  • 10

    Standardized incidence rate for men; NS, Not significant.

  • 11

    Standardized incidence rate for women.

  • 12

    Relative risk and 95% confidence interval (RR (95% CI)) for men.

  • 13

    RR (95% CI) for women.

  • 14

    Incident cases not available.

Case-control study
 Oshima, 1986 (Ref.5)Osaka Prefecture140+91 cases0.60 (0.34–1.05)2
261 controls0.38 (0.19–0.79)3
 Pisani, 1994 (Ref.12)State of Tachira435+241 cases1.26 (0.83–1.91)5
2,410 controls 
 Fukao, 1995 (Ref.6)Miyagi Prefecture150+198 cases0.41 (0.28–0.61)5
577 controls 
 Abe, 1995 (Ref.7)Chiba Prefecture130+820 cases0.37 (0.24–0.57)2
2,413 controls0.46 (0.26–0.80)3
Cohort study
 Oshima, 1979 (Ref.8)Osaka Prefecture11967–1975 (6 years)All ages32,789  
179 deathsDeath0.916
335 casesIncidence1.377
 Hisamichi, 1984 (Ref.9)Miyagi Prefecture11960–1977 (18 years)40–697,008  
81 deathsDeath61.9 (p < 0.005)8
28.1 (p < 0.01)9
165 casesIncidence187.7 (NS)10
58.1 (NS)11
 Inaba, 1999 (Ref.10)Gifu Prefecture11992–1995 (40 months)41+24,134  
40 deathsDeath0.72 (0.31–1.66)12
1.46 (0.43–4.90)13
N/A14  
 Mizoue, 2003 (Ref.11)JACC study11988–1997 (8 years)40–7987,312  
480 deathsDeath0.65 (0.45–0.95)12
0.75 (0.42–1.34)13
N/A14  

The aim of the present study was to investigate prospectively the association between gastric cancer screening and subsequent risk of gastric cancer death in the Japanese population, using a large-scale population-based prospective cohort study, with a 13-year follow-up period.

Material and methods

Study population and baseline survey

The basis for this investigation lies with the Japan Public Health Center-based prospective study (JPHC Study) Cohort I, which was established in January 1990 and questioned about screening experience in the baseline survey. The Cohort covered 5 prefectural public health center (PHC) areas: Ninohe (Iwate Prefecture), Yokote (Akita Prefecture), Saku (Nagano Prefecture), Chubu (Okinawa Prefecture) and Katsushika (metropolitan Tokyo). Details of the study design have been provided elsewhere.13 The study protocol was approved by the institutional review board of the National Cancer Center, Japan. For the present analysis, the Katsushika PHC area was excluded, since its study population was defined differently to the others, and no data for cancer incidence were available. The study population was defined as all registered Japanese inhabitants in 14 administrative districts, supervised by the 4 PHC areas, aged 40–59 years at the beginning of the baseline survey. The Japanese inhabitants were identified from population registries maintained by the local municipalities. Initially, 54,498 subjects were identified as the study population. Of these, 121 subjects were found to be ineligible during the follow-up period and were excluded for any of several reasons (non-Japanese nationality (n = 29), late report of emigration occurring before the start of the follow-up period (n = 86), incorrect birth date (n = 3), and duplicate data (n = 3)). As a result, a population-based cohort of 54,377 subjects (26,988 men and 27,389 women) was established.

A baseline self-administered questionnaire survey, which included socio-demographics, personal medical history, screening experience, smoking and alcohol drinking history and diet, was conducted in 1990. A total of 43,140 subjects responded, giving a response rate of 79%. We further excluded 990 subjects, with a present or past history of cancer (i.e., gastric, lung, colorectal, liver, breast, uterine and other cancers) at baseline. Finally, 42,150 eligible subjects (20,326 men and 21,824 women) were included in the analysis.

Follow-up

Subjects were followed from the baseline survey until December 31, 2003. Residence status was confirmed annually through the residential registry maintained by the respective study area municipalities; for those who moved out of the area, residence status was confirmed through the municipal office of the area to which they had moved. In Japan, residency and death registration is required by law, and the registries are believed to be complete. Inspection of the resident registry is available to anyone under the Resident Registration Law. Information on the cause of each death was supplemented by checking against death certificate files, provided by the Ministry of Health, Labor and Welfare, after permission was obtained from the Ministry of Internal Affairs and Communications. We classified the causes of death, according to the International Classification of Diseases (ICD), 10th Revision (ICD-10).14 Those already coded according to the ICD 9th Version (1990–1994) were converted to the ICD-10 code. During the study period, 2,678 subjects died, 2,791 moved out of the study area (7%) and 58 were lost to follow-up (0.1%). A total of 1,141 died of cancer during the follow-up period. Lung cancer was the most common cause of death from cancer (n = 215; 19%), followed by gastric cancer (n = 179; 16%), colorectal cancer (12%) and liver cancer (9%).

The occurrence of cancer was identified by active patient notification from major local hospitals in the study area and data linkage with population-based cancer registries, with permission from each of the local governments responsible for the cancer registries. Death certificate information was used as a supplementary information source. The site and histology of each case were coded using the ICD for Oncology, Third Edition (ICD-O-3).15 As of December 2003, a total of 3,035 newly diagnosed cases of cancer were identified. Gastric cancer was the most common (n = 636; 21%), followed by colorectal cancer (n = 597; 13%), lung cancer (10%) and breast cancer (9%). Finally, 42,150 subjects (20,326 men and 21,824 women) were used for analysis, including 179 gastric cancer deaths and 636 incident gastric cancers.

Exposure and outcome

Information on screening photofluorography was obtained from the subjects' self-reported screening experience, using the questionnaire in the 2 categories of yes (i.e., has undergone screening) and no. Subjects were also asked about the screening tests they had taken. The conceptual definition of the screened group was those who had undergone screening photofluorography at baseline. On this basis, those who reported to undergo screening photofluorography during the 12 months preceding the baseline survey were defined as the screened group. Gastric cancer death (ICD10: C16)14 was the main outcome. Additional outcome variables were incident gastric cancer (ICD-O-3: C16),15 death from all cancers (ICD10: C00–C97) except gastric cancer, all causes of death except gastric cancer and incident cases of all cancers excluding gastric cancer.

Statistical analysis

Person-years were counted from the date of response to the baseline survey until one of the following endpoints: for the analysis of gastric cancer deaths, the date of emigration from the study area, the date of death or the end of the study period (December 31, 2003), whichever came first; and for the analysis of gastric cancer incidence, the date of occurrence of gastric cancer, the date of emigration from the study area, the date of death or the end of the study period, whichever came first. Persons who were lost to follow-up were censored on the last confirmed date of their presence in the study area.

Gastric cancer death was the main outcome measure. To evaluate potential selection bias, death from all cancers except gastric cancer, all causes of death except gastric cancer, and incident cases of gastric cancer and of all cancers excluding gastric cancer were used as additional outcome variables. To estimate the incidence rate of advanced gastric cancer, incident cases of early gastric cancer (i.e., cases limited to the mucosa, or mucosa and submucosa)1, 16 were excluded. Hazard ratio (HR) and 95% confidence intervals (95% CI) were used to describe the relative risk of gastric cancer death and incidence associated with gastric cancer screening. Additional age-stratified analyses were done to investigate the effect of gastric cancer screening in subjects aged 40–49 and 50–59 years, and to estimate the respective HRs.

Differences in proportions and mean values between the screened and unscreened subjects were assessed using the χ2 test and unpaired t-test, respectively. The Cox proportional hazards regression model was used to control potential confounding factors. We adjusted for age at baseline (continuous) and study area (4 PHC areas), and then further adjusted for educational background (up to high school, college or higher), smoking status (never, former or current), drinking status (none, occasional or regular), salty foods (salted fish roe or gut, or dried or salted fish at least 3 times per week, or less), rice (at least 4 bowls per day, or less), green–yellow vegetable intake (everyday, less than everyday), miso soup intake (everyday, less than everyday), green tea intake (everyday, less than everyday) and family history of gastric cancer (yes or no). These variables are either known or suspected risk or preventive factors for gastric cancer, based on previously reported results.17, 18, 19, 20, 21, 22, 23, 24, 25, 26 All analyses were repeated after the subjects diagnosed with or dying of gastric cancer, during the first 3 years of follow-up, were excluded. The significance of interactions between the screening and these risk factors for gastric cancer was determined by comparing the model of the screening × respective variable, with the model that assessed the main effect only. Statistical analyses were performed using SAS (SAS Institute, Cary, NC).

Results

During the 551,459 person-years of the follow-up period (average follow-up period: 13.1 years) for the 42,150 subjects (20,326 men and 21,824 women), a total of 179 gastric cancer deaths and 636 cases of newly diagnosed gastric cancer were included in the analyses.

Baseline characteristics of subjects according to gastric cancer screening are shown in Table II. Overall mean age was 49.5 years. At baseline, approximately 36% had undergone gastric cancer screening during the preceding 12 months. The educational level of the screened group was higher than that of the unscreened group. Current smokers were more common in the unscreened group, but regular drinkers were more common in the screened group. The screened subjects were more likely to favor miso soup, green–yellow vegetables and green tea than do the unscreened subjects, but were also more likely to report a family history of gastric cancer and to consume salty foods such as salted fish roe or gut and dried or salted fish.

Table II. Baseline Characteristics of The Study Subjects According to Gastric Cancer Screening
 Total (n = 42,150)Unscreened (n = 26,961)Screened (n = 15,189)p
  • 1

    by unpaired t-test.

  • 2

    by χ2 test.

Proportion (%)100.0063.9636.04 
Mean age (years) ± SD49.5 ± 5.949.2 ± 5.950.2 ± 5.8<0.00011
Sex
 Male48.2247.2749.92<0.00012
 Female51.7852.7350.08 
Educational level
 ≤12 years of schooling86.8787.9584.95<0.00012
 >12 years of schooling13.1312.0515.05 
Smoking status (%)
 Never59.5759.3260.01<0.00012
 Former11.8410.9413.42 
 Current28.5929.7426.57 
Alcohol drinking status (%)
 None50.3852.0447.43<0.00012
 Occasional11.4511.0912.09 
 Regular38.1836.8740.49 
Family history of gastric cancer (%)
 No93.1794.3591.09<0.00012
 Yes6.835.658.91 
Salted fish roe or gut intake (%)
 <3 times per week83.1684.0181.66<0.00012
 ≥3 times per week16.8415.9918.34 
Dried or salted fish intake (%)
 <3 times per week67.9469.5665.07<0.00012
 ≥3 times per week32.0630.4434.93 
Miso soup intake (%)
 Less than everyday23.7725.5520.61<0.00012
 Everyday76.2374.4579.39 
Rice intake (%)
 <4 bowls per day70.2870.4969.920.222
 ≥4 bowls per day29.7229.5130.08 
Breakfast (%)
 Less than everyday17.0418.6614.13<0.00012
 Everyday82.9681.3485.87 
Green–yellow vegetable intake (%)
 Less than everyday61.1262.5258.63<0.00012
 Everyday38.8837.4841.37 
Green tea intake (%)
 Less than everyday34.5937.0730.20<0.00012
 Everyday65.4162.9369.80 
Exercise (%)
 Less than 1 time per month72.4375.5766.87<0.00012
 1–3 times per month11.6210.1514.21 
 More than 1 time per week15.9514.2818.92 
Stress level (%)
 Low11.1511.7110.16<0.00012
 Medium65.3765.8664.52 
 High23.4722.4325.32 

The relative risk for the subsequent risk of gastric cancer according to gastric cancer screening is shown in Table III. The overall number of gastric cancer deaths was 49 in the screened group and 130 in the unscreened group. The screened subjects showed a 48% reduced risk of death from gastric cancer compared with the unscreened subjects (RR = 0.52; 95% CI = 0.36–0.74). Screening participation was associated with a 21% reduced risk of death from all cancers excluding gastric cancer and of 29% from all causes excluding gastric cancer (RR = 0.79; 95% CI = 0.69–0.91; RR = 0.71; 95% CI = 0.65–0.78, respectively). This trend did not change when cases of death from gastric cancer occurring within the first 3 years of the follow-up period were excluded (RR = 0.43; 95% CI = 0.29–0.64; RR = 0.79; 95% CI = 0.69–0.92; RR = 0.71; 95% CI = 0.65–0.78, respectively). Although the overall incidence rate of gastric cancer did not differ significantly between the screened and the unscreened group, a significant decrease in the incidence of advanced gastric cancer was seen in the screened subjects, after adjustment for potential confounding factors (RR = 0.75; 95% CI = 0.58–0.96). Similarly, this trend was not changed after the exclusion of gastric cancer cases occurring within the first 3 years of follow-up (RR = 0.73; 95% CI = 0.55–0.97). The rate of decline in mortality rate was greater among the screened subjects.

Table III. Relative Risk and 95% Confidence Interval for Gastric Cancer Screening and Subsequent Risk of Gastric Cancer in 42,150 Subjects in The JPHC Study1
 No. of casesRate (per 100,000 person-years)RR2(95% CI)RR3(95% CI)
  • 1

    RR, relative risk; CI, confidence interval.

  • 2

    Adjusted for age at baseline, sex and study area.

  • 3

    Additionally adjusted for smoking status; alcohol consumption; educational level; intake of salty foods, rice, miso soup, green–yellow vegetables, and green tea; and family history of gastric cancer.

  • 4

    Excluding gastric cancer.

  • 5

    Overall incidence rate of gastric cancer.

  • 6

    Incidence of advanced gastric cancer.

Cancer deaths
Gastric cancer
 Unscreened13036.91.00 1.00 
 Screened4924.60.60(0.43–0.83)0.52(0.36–0.74)
All cancers4
 Unscreened645183.21.00 1.00 
 Screened317158.90.78(0.68–0.89)0.79(0.69–0.91)
All causes of death4
 Unscreened1744495.41.00 1.00 
 Screened755378.50.69(0.64–0.76)0.71(0.65–0.78)
Cancer incidence
Gastric cancer5
 Unscreened374111.41.00 1.00 
 Screened262139.01.13(0.96–1.32)1.06(0.90–1.25)
Gastric cancer6
 Unscreened20561.51.00 1.00 
 Screened10757.40.84(0.67–1.06)0.75(0.58–0.96)
All cancers4
 Unscreened1492444.51.00 1.00 
 Screened907481.21.01(0.93–1.09)1.00(0.92–1.09)

Additional investigation of the interaction between gastric cancer screening and different age groups showed significant interaction effects (p < 0.0001), leading us to investigate the association between gastric cancer screening and subsequent risk of gastric cancer, using stratified analysis by age groups (Table IV). After adjustment for potential confounding factors, the screened subjects showed a 70% reduced risk of death from gastric cancer, among subjects aged 40–49 years at baseline (RR = 0.30; 95% CI = 0.13–0.72). This decrease in mortality for gastric cancer was greater than in death from all causes excluding gastric cancer (RR = 0.77; 95% CI = 0.65–0.91). Among subjects aged 50–59 years at baseline, in contrast, the screened subjects showed a 40% reduced risk of death from gastric cancer (RR = 0.60; 95% CI = 0.40–0.88). Screening participation was also associated with a 26% reduced risk of death from all cancers other than gastric cancer and of 31% from all causes other than gastric cancer (RR = 0.74; 95% CI = 0.63–0.88; RR = 0.69; 95% CI = 0.62–0.77; respectively). In these age-stratified analyses, gastric cancer incidence did not differ significantly between the screened and the unscreened subjects (for subjects aged 40–49, RR = 0.58, 95% CI = 0.33–1.02; for subjects aged 50–59, RR = 0.80, 95% CI = 0.61–1.06). Further, the trends in different age groups did not change after the exclusion of gastric cancer deaths and incident cases occurring within the first 3 years of the follow-up period (data not shown).

Table IV. Relative Risk and 95% Confidence Interval for Gastric Cancer Screening and Subsequent Risk of Gastric Cancer in Subjects Aged 40–49 Years and 50–59 Years at Baseline in The JPHC Study1
 No. of casesRate (per 100,000 person-years)RR2(95% CI)RR3(95% CI)
  • 1

    RR, relative risk; CI, confidence interval.

  • 2

    Adjusted for sex and study area.

  • 3

    Additionally adjusted for smoking status; alcohol consumption; educational level; intake of salty foods, rice, miso soup, green–yellow vegetables, and green tea; and family history of gastric cancer.

  • 4

    Excluding gastric cancer.

  • 5

    Overall incidence rate of gastric cancer.

  • 6

    Incidence of advanced gastric cancer.

40–49 years
Cancer deaths
Gastric cancer
 Unscreened3820.81.00 1.00 
 Screened89.10.41(0.19–0.88)0.30(0.13–0.72)
All cancers4
 Unscreened16288.61.00 1.00 
 Screened7686.60.95(0.72–1.24)0.98(0.74–1.29)
All causes of death4
 Unscreened528288.91.00 1.00 
 Screened199226.70.74(0.63–0.87)0.77(0.65–0.91)
Cancer incidence      
Gastric cancer5
 Unscreened10258.71.00 1.00 
 Screened5768.81.08(0.78–1.50)1.03(0.74–1.45)
Gastric cancer6
 Unscreened5531.81.00 1.00 
 Screened1923.10.66(0.39–1.12)0.58(0.33–1.02)
All cancers4
 Unscreened506291.01.00 1.00 
 Screened272328.51.13(0.97–1.31)1.13(0.97–1.32)
50–59 years
Cancer deaths
Gastric cancer
 Unscreened9254.41.00 1.00 
 Screened4136.70.67(0.46–0.96)0.60(0.40–0.88)
All cancers4
 Unscreened483285.41.00 1.00 
 Screened241215.80.74(0.63–0.86)0.74(0.63–0.88)
All causes of death4
 Unscreened1216718.61.00 1.00 
 Screened556497.90.68(0.61–0.75)0.69(0.62–0.77)
Cancer incidence
Gastric cancer5
 Unscreened272168.11.00 1.00 
 Screened205194.01.14(0.95–1.37)1.07(0.88–1.29)
Gastric cancer6
 Unscreened15093.71.00 1.00 
 Screened8884.40.90(0.69–1.17)0.80(0.61–1.06)
All cancers4
 Unscreened986609.51.00 1.00 
 Screened635600.80.97(0.88–1.07)0.96(0.87–1.07)

Discussion

The JPHC study is an ongoing prospective population-based cohort study designed to investigate the relation of lifestyle, diet and cancer, and has been conducted in Japan since 1990. The present results are the first from the JPHC study on cancer screening.

With a follow-up period of 13 years, the study revealed a 2-fold decrease in gastric cancer mortality in screened versus unscreened subjects. The extent of mortality reduction was greater for gastric cancer than for death from other causes. No statistically significant difference in gastric cancer incidence was seen between the screened and the unscreened subjects. To our knowledge, the only outcome variable known to be valid in the evaluation of cancer screening is mortality rate.27, 28 If the incidence rate were measured as an alternative to mortality rate, it could increase with screening, because it could be artificially inflated by length bias or overdiagnosis bias.27 In place of overall incidence rate, therefore, we also investigated the incidence rate of advanced gastric cancer as another outcome for evaluating gastric cancer screening. Results after adjustment showed a significant decrease in the incidence of advanced gastric cancer in screened versus unscreened subjects. The greater decrease in mortality than in incidence may be related to the improved survival from gastric cancer, resulting from the higher possibility of early detection and early treatment afforded by gastric cancer screening. It is possible, however, that incidence has been somewhat inflated by the detection of gastric cancer by screening. This tendency did not substantially change further, after cases that were diagnosed with or who died early during the follow-up period were excluded.

In the age-stratified analysis, the incidence of gastric cancer did not differ significantly between the screened and unscreened subjects. However, a significant reduction in gastric cancer mortality of 70% was seen in screened subjects aged 40–49 years at baseline, compared with a reduction of 40% in those screened aged 50–59. This finding may suggest that the benefit of screening was greater in subjects aged 40–49 years than in those aged 50–59. However, the size of the benefit and the optimum age range for screening are as yet unknown. Further studies are needed to answer these questions.

Although the effect of gastric cancer screening on mortality from gastric cancer has been evaluated previously,5, 6, 7, 8, 9, 10, 11, 12 few reports have prospectively investigated the association between gastric cancer screening and the subsequent risk of death and incidence from gastric cancer. Further, those prospective studies which have appeared give inconsistent results.8, 9, 10, 11

Most case-control studies5, 6, 7 have shown a 40–60% reduced risk of death from gastric cancer in screened subjects, although one case-control study, conducted in Venezuela, showed no significant decrease.12 More significantly, however, selection bias could not be controlled in any of these studies, because no information on potential confounding factors was provided. Further, with few exceptions,10, 11 no adjustment was made for lifestyle factors.

We therefore considered lifestyle factors and dietary habits that may have been associated with gastric cancer in previous reports as potential confounding factors,17, 18, 19, 20, 21, 22, 23, 24, 25, 26 with the expectation that statistical adjustment might be to some degree possible, if information on associated variables were available. Results showed that even though the screened subjects were more likely to report risk factors such as a family history of gastric cancer and a higher intake of salty foods, they tended to have healthier lifestyles such as a lower prevalence of smoking and higher intake of dietary items like green–yellow vegetables and green tea, all of which may be associated with a reduced risk of gastric cancer.17, 18, 19, 20, 21, 22, 23, 24, 25, 26 Similar differences between the screened and unscreened subjects have been observed in previous reports.10, 11

A previous report6 showed that the effect of past screening remains over at least 3 years of follow-up. Additionally, to avoid the potential effects of past screening, we repeated the analyses after excluding deaths and incident cases during the first 3 years of follow-up. No substantial change in trends after these exclusions was seen, indicating that the effects of such bias were negligible.

One concern suggested in previous studies was the presence of gender differences on the magnitude of effect of gastric cancer screening.5, 6, 11 Findings from previous studies have been inconsistent. No significant interaction effect was seen in the present study, however, nor in the pooled analysis of previous case-control studies, suggesting that there was no marked gender difference in the efficacy of gastric cancer screening.3

Another concern in this study was the validity of the exposure assessment. We determined the screened and the unscreened subjects on the basis of self-reports. We assumed that subjects who had undergone screening photofluorography during the preceding 12 months were also screened during the follow-up period, whereas those who had not undergone screening were not. The subjects were followed on the fifth year after the baseline survey, and we evaluated this assumption using the data from the follow-up survey. The validity of subjects' information on screening experience was assessed by comparing the screening history between baseline and follow-up in the 35,175 subjects participating in both. Among those screened at baseline, the proportion who responded that they had also participated in gastric cancer screening at follow-up was 60% (7,763/13,038). The respective proportion among those unscreened was 75% (16,541/22,137). Similar proportions were observed when the subjects were age-stratified. This finding suggests that recent screening participation can be a predictor of future participation. Further, the screened individuals seemed to have a greater awareness of health and health-promoting behavior than those unscreened. Similar concerns about this health behavior have been expressed elsewhere.10, 11 The misclassification of self-reported screening status resulting from modification during the follow-up period is possible, but would likely be nondifferential. Moreover, it would likely result in an underestimation of true relative risk. We therefore consider that the extent of distortion from this misclassification might be minor, and in any case tend toward underestimation.

Notwithstanding the known limitations of observational studies, the present study has the following strengths: it was a prospective population-based study, with a 13-year follow-up period. Information on screening experience was collected before the subsequent onset of cancer, eliminating the exposure recall bias that is inherent to case-control studies. Study subjects were collected from the general population, and response rate to the baseline questionnaire was high, while the proportion of losses to follow-up was negligible. Finally, adjustment was made for lifestyle factors and dietary habits that may have been associated with gastric cancer in previous reports.17, 18, 19, 20, 21, 22, 23, 24, 25, 26

In conclusion, from long-term follow-up of a large-scale population-based cohort study, we observed a significant reduction in gastric cancer mortality in screened subjects. Our findings suggest that gastric screening may be associated with a reduction in mortality from gastric cancer in the Japanese population. However, further evaluation of this screening by prospective observation or by a randomized controlled trial is warranted, to assess whether the present results are representative of or can be generalized to other populations.

Acknowledgements

K-J. Lee was awarded a Visiting Scientist Fellowship from the Foundation for Promotion of Cancer Research in Japan. The authors thank all staff members in each study area for their unfailing efforts in conducting the baseline and follow-up surveys. We also express our gratitude to the Iwate, Aomori and Okinawa Cancer Registries, for providing the incidence data. We thank Dr. Y. Tsubono for his helpful advice about this article. This work was supported by a Grant-in-Aid for Cancer Research and by the Second- and Third-Term Comprehensive 10-Year Strategy for Cancer Control from the Ministry of Health, Labor and Welfare, Japan.

Appendix

Members of the Japan Public Health Center-based Prospective Study (JPHC Study) Group are S. Tsugane, M. Inoue, T. Sobue, T. Hanaoka, National Cancer Center, Tokyo; J. Ogata, S. Baba, T. Mannami, A. Okayama, National Cardiovascular Center, Suita; K. Miyakawa, F. Saito, A. Koizumi, Y. Sano, I. Hashimoto, Iwate Prefectural Ninohe Public Health Center, Ninohe; Y. Miyajima, N. Suzuki, S. Nagasawa, Y. Furusugi, Akita Prefectural Yokote Public Health Center, Yokote; H. Sanada, Y. Hatayama, F. Kobayashi, H. Uchino, Y. Shirai, T. Kondo, R. Sasaki, Y. Watanabe, Nagano Prefectural Saku Public Health Center, Saku; Y. Kishimoto, E. Takara, T. Fukuyama, M. Kinjo, M. Irei, Okinawa Prefectural Chubu Public Health Center, Okinawa; K. Imoto, H. Yazawa, T. Seo, A. Seiko, F. Ito, Katsushika Public Health Center, Tokyo; A. Murata, K. Minato, K. Motegi, T. Fujieda, Ibaraki Prefectural Mito Public Health Center, Mito; K. Matsui, T. Abe, M. Katagiri, Niigata Prefectural Kashiwazaki Public Health Center, Kashiwazaki; M. Doi, A. Terao, Y. Ishikawa, Kochi Prefectural Chuo-higashi Public Health Center, Tosayamada; H. Sueta, H. Doi, M. Urata, N. Okamoto, F. Ide, Nagasaki Prefectural Kamigoto Public Health Center, Arikawa; H. Sakiyama, N. Onga, H. Takaesu, Okinawa Prefectural Miyako Public Health Center, Hirara; F. Horii, I. Asano, H. Yamaguchi, K. Aoki, S. Maruyama, M. Ichii, Osaka Prefectural Suita Public Health Center, Suita; S. Matsushima, S. Natsukawa, Saku General Hospital, Usuda; S. Watanabe, M. Akabane, Tokyo University of Agriculture, Tokyo; M. Konishi, K. Okada, Ehime University, Matsuyama; H. Iso, Y. Honda, Tsukuba University, Tsukuba; H. Sugimura, Hamamatsu University, Hamamatsu; Y. Tsubono, Tohoku University, Sendai; M. Kabuto, National Institute for Environmental Studies, Tsukuba; S. Tominaga, Aichi Cancer Center Research Institute, Nagoya; M. Iida, W. Ajiki, Osaka Medical Center for Cancer and Cardiovascular Disease, Osaka; S. Sato, Osaka Medical Center for Health Science and Promotion, Osaka; N. Yasuda, Kochi Medical School, Nankoku; S. Kono, Kyushu University, Fukuoka; K. Suzuki, Research Institute for Brain and Blood Vessels Akita, Akita; Y. Takashima, Kyorin University, Mitaka; E. Maruyama, Kobe University, Kobe; M. Yamaguchi, Y. Matsumura, S. Sasaki, National Institute of Health and Nutrition, Tokyo; and T. Kadowaki, Tokyo University, Tokyo.

Ancillary