VEGETABLES, FRUIT AND RISK OF GASTRIC CANCER IN JAPAN: A10-YEAR FOLLOW-UP OF THE JPHC STUDY COHORT I

Transcription

1 Int. J. Cancer: 102, (2002) 2002 Wiley-Liss, Inc. Publication of the International Union Against Cancer VEGETABLES, FRUIT AND RISK OF GASTRIC CANCER IN JAPAN: A10-YEAR FOLLOW-UP OF THE JPHC STUDY COHORT I Minatsu KOBAYASHI 1,Yoshitaka TSUBONO 2,Shizuka SASAZUKI 1,Satoshi SASAKI 1,and Shoichiro TSUGANE 1 * for the JPHC Study Group 1 Epidemiology and Biostatistics Division, National Cancer Center Research Institute East, Kashiwa, Japan 2 Division of Epidemiology, Department of Public Health and Forensic Medicine, Tohoku University, Sendai, Japan Theassociationbetweenvegetablesandfruitconsumption andgastriccancerriskwasinvestigatedinapopulation-based prospective study in 4public health center areas in Japan. Dietary and other exposure data were obtained in 1990 from acohort of 19,304 men and 20,689 women with aself-administered questionnaire. After 10 years of follow-up, atotal of 404 cases of gastric cancer were documented among them. After adjustment for age, gender, areas and other potential confounding factors and after exclusion of the cases diagnosed in first and second follow-up years, the relative risk associated with intake 1or more days per week compared to less than 1day per week was 0.64 (95% CI ) for yellowvegetable,0.48(95%ci )forwhitevegetable and0.70(95%ci )forfruit.Relativerisksassociated with quintile of total vegetable consumption were 1.00, 0.86, 0.75, 0.90 and 0.75 (p for trend 0.17). In the differentiated type of gastric cancer, the association became clearer: 1.00, 0.96,0.78,0.88and0.53(pfortrend 0.03).Thisprospective study suggests that vegetable and fruit intake, even in low amounts, is associated with alower risk of gastric cancer. Although no striking differences in the association were seen between cardia and noncardia cancer, an inverse association was higher in differentiated rather than in undifferentiated types of gastric cancer Wiley-Liss, Inc. Keywords:gastriccancer;histologicclassification;vegetables;fruit; prospective studies Although gastric cancer mortality has been declining during the last few decades, it remains the second most common cancer worldwide. 1 High consumption of vegetables and fruit has been hypothesized to have protective effects against gastric cancer. Although in many previous case-control studies ahigh consumptionofvegetablesandfruithasbeenassociatedwithareducedrisk of gastric cancer, 2 4 the evidence from 11 prospective cohort studies has not been consistent Trends of gastric cancer incidence reportedly differ depending onthesubsitesandhistologictypes.theincidenceofcardiacancer has recently increased, while the rate of distal cancer has been stableforintestinaltypeordecreasingfordiffusetype Onthe other hand, recent studies have indicated that infection by Helicobacter pylori (H. pylori), an established risk factor for gastric cancer, is associated with an increased risk for noncardia cancer 20 but not for cardia cancer. 21 These findings indicate that discrepanciesinincidencetrendsbetweenthetwosubsites(noncardiaand cardia) and 2 histologic types (intestinal and diffuse) may be linked with not only H. pylori infection but also dietary habits or nutrient factors including the consumption of vegetable and fruits. However, only afew case-control studies of fruits, vegetables and gastric cancer have analyzed the data according to subsites and histology, and no prospective study has been conducted. To further examine the association between the consumption of vegetables and fruit and the risk of gastric cancer, we conducted a population-based prospective study in four public health center areas as part of the Japan Public Health Center-based prospective study on cancer and cardiovascular disease (JPHC Study). Furthermore, we examined the association between the risk of different gastric cancer subsites and histologic types and vegetables and fruit consumption. MATERIAL AND METHODS Study cohort JPHC Study Cohort Iis aprospective cohort study that began in January The study design has been described previously. 26 Briefly, the cohort included 27,063 men and 27,435 women, aged years at baseline who registered their addresses in 4public health center (PHC) areas: Ninohe PHC area of Iwate Prefecture, Yokote PHC area of Akita, Saku PHC area of Nagano and Ishikawa PHC area of Okinawa. These 4PHC areas were selected to represent the extent of variation in the mortality rate of gastric cancer based on our previous ecologic study. 27,28 After the initiation of the study, 65 men and 38 women were found to be ineligible and were thus excluded (28 persons of non-japanese nationality, 73 with delayed reports of out-migration before the start of the follow-up and 2with mistakenly recorded birthdays), leaving 26,998 men and 27,397 women eligible for the study. Exposure data Aself-administered questionnaire was distributed to all registered residents in 1990 and was collected from 20,658 men (76%) and 22,482 (82%) women. It included afood frequency questionnaire (FFQ), which asked about the average consumption of 44 food items during the previous month. The questionnaire included 4items on vegetables ( greenleafyvegetablessuchasspinach, yellowvegetablessuch as carrot, pumpkin, white vegetables such as chinese cabbage, Grant sponsor: Cancer Research and the Second Term Comprehensive Ten-Year Strategy for Cancer Control from the Ministry of Health and Welfare of Japan. The investigators and institutions participating in the JPHC Study Cohort I, apart of the JPHC Study Group, were as follows: S. Tsugane, S. Sasaki, Epidemiology and Biostatistics Division, National Cancer Center Research Institute East, Kashiwa; J. Ogata, S. Baba, National Center for Circulatory Diseases, Suita; K. Miyakawa, F. Saito, A. Koizumi, Iwate Prefectural Ninohe Public Health Center, Ninohe; Y. Miyajima, N. Suzuki, S. Nagasawa, Akita Prefectural Yokote Public HealthCenter,Yokote;H.Sanada,Y.Hatayama,F.Kobayashi,H.Uchino, Y. Shirai, T. Kondo, Nagano Prefectural Saku Public Health Center, Saku; Y. Kishimoto, E. Takara, M. Kinjo, T. Fukuyama, Okinawa Prefectural Ishikawa Public Health Center, Ishikawa; S. Matsushima, S. Natsukawa, Saku General Hospital, Usuda; S. Watanabe, M. Akabane, Tokyo University of Agriculture, Tokyo; M. Konishi, Ehime University, Matsuyama; S. Tominaga, Aichi Cancer Center Research Institute, Nagoya; M. Iida, S. Sato, Center for Adult Diseases, Osaka; the late M. Yamaguchi, Y. Matsumura, National Institute of Health and Nutrition, Tokyo; Y. Tsubono, Tohoku University, Sendai; H. Iso, Tsukuba University, Tsukuba; H. Sugimura, Hamamatsu University, Hamamatsu; M. Kabuto, National Institute for Environmental Studies. *Correspondence to: Epidemiology and Biostatistics Division, National Cancer Center Research Institute East, Kashiwanoha, Kashiwa, Chiba, , Japan. Fax: ; stsugane@east.ncc.go.jp Received 22 January 2002; Revised 23 May, 19 July 2002; Accepted 23 July 2002 DOI /ijc Published online 5 September 2002 in Wiley InterScience (www. interscience.wiley.com).

2 40 KOBAYASHI ET AL. radish, tomato, cucumber and pickled vegetables ), 1 item on total fruit and 2 on juices (vegetable and fruit). The consumption frequency of vegetables and fruit was asked using 4 categories: less than 1 day per week, 1 2 days per week, 3 4 days per week and almost daily. For the 2 juices, we used 6 frequency categories (less than 1 day per week, 1 2 days per week, 3 4 days per week, almost daily and 1 2 cups per day, almost daily and 3 4 cups per day and almost daily and 5 or more cups per day). For calculating the amount of vegetables and fruit intakes, we determined the portion size and the content of each food item based on the observed median values on day diet record data. 29 This diet record was conducted to assess the validity of the questionnaire. The Spearman correlation coefficients between the diet record and the questionnaire in men (n 94) and women (n 107), respectively, were 0.26 and 0.36 for the amount of total vegetable intake (g/day), which is defined as the sum of intake of green, yellow and white vegetables, 0.52 and 0.41 for the amount of fruit intake (g/day), 0.39 and 0.40 for frequency of vegetable intake (times/day) and 0.59 and 0.44 for frequency of fruit intake (times/day). For the amount of vegetables and fruit intake (g/day), we adjusted for total energy intake by the residual method. 30 Participants also provided information on their frequency consumption of highly salted fish roe and gut (sodium chloride content: 7 12%), use of vitamin supplements, smoking status (never, past smoker and current smoker), alcohol consumption (g/day), educational level (up to primary school, high school and college or higher) history of peptic ulcer (yes or no) and family history of gastric cancer (yes or no). Of 43,140 subjects who responded to the questionnaire, 1,086 men and 1,195 women who reported extreme total energy intake (upper 2.5% or lower 2.5%) and 268 men and 598 women who reported a past history of cancer were excluded, leaving 19,304 eligible men and 20,689 eligible women for the analysis. Follow-up We followed the subjects from January 1, 1990, through December 31, Incident cases of cancer occurring in the cohort have been identified through continuous surveillance of hospital records, population-based cancer registries and death certificates. In Ninohe and Ishikawa PHC areas, prefecture-wide cancer registries were available. Of the 2,610 cancer cases diagnosed in , 245 (9.4%) were first identified by death certificate as a supplementary information source for the cancer registry (Death Certificate Notification, DCN). Among them, 55 cases not confirmed by other medical records accounted for 2.1% (Death Certificate Only, DCO) of all entries, reflecting a reasonably high quality of cancer registration in this cohort. Each case was confirmed by a histologic diagnosis, based on biopsy or surgery. The histologic classification was based on the available pathology records, and no collection and reclassification of original specimens were done. As of November 2000, a total of 404 cases of gastric cancer, 294 in 19,304 men and 110 in 20,689 women, were documented with a histologically confirmed diagnosis in Cardia cancer has been defined as a tumor located in the esophagogastric junction or upper third of the stomach (International Classification of Disease for Oncology [ICD-O] code C ). 31 Until quite recently in Japan, the upper third of the stomach has been called the cardia based on the guidelines for gastric cancer classification. 32 Because of the difficulty in distinguishing the actual cardia from the upper third of the stomach, we combined them into one site for analysis in our study. A tumor located on the lower side of the stomach was classified as distal gastric cancer (ICD-O code C ). Those subsites that could not be classified for its diffuse lesion (ICD-O code C168) or those with no information (ICD-O code C169) were categorized as unclassified. The following histologic subdivisions were made in our study: the differentiated types (corresponding to intestinal types of Lauren s classification) were composed of papillary adenocarcinoma, tubular adenocarcinoma (well-differentiated type) and tubular adenocarcinoma (moderately differentiated type); the undifferentiated types (corresponding to diffuse types of Lauren s classification) included poorly differentiated adenocarcinoma, mucinous adenocarcinoma and signet-ring cell carcinoma. Adenosquamous carcinoma, squamous cell carcinoma, carcinoid tumor, undifferentiated carcinoma and miscellaneous others belonged to unclassified types. In accordance with this classification, 47 cardia cancer, 289 noncardia cancer and 64 unclassified subsite cancer patients were identified. The noncardia cancers included 111 undifferentiated types (corresponding to diffuse types of Lauren s classification), 164 differentiated (corresponding to intestinal types of Lauren s classification) and 14 unclassified types. Statistical analysis All analyses were conducted for men and women combined because of the relatively small number of gastric cancer cases among women. We computed the gastric cancer incidence rate for a frequency category of vegetables and fruit and a quintile category of vegetable consumption by dividing the number of gastric cancer cases by person-years of follow-up. We calculated personyears of follow-up for each subject from January 1, 1990 until the dates of diagnosis of gastric cancer, death or moving from a PHC area or the end of follow-up (December 31, 1999), whichever occurred first. We used the Cox proportional-hazards regression model to estimate the rate ratios (RRs) using the SAS PHREG procedure. 33 RRs were adjusted for the following variables: age (5-year age groups), gender, PHC areas, educational level (up to primary school, high school, college or higher), smoking status (never, former, current), BMI (less than 22, 22 and 25, 25), alcohol consumption (none, 250 g/week, 250 g/week), use of supplements for vitamin A, C or E, total energy intake (in quintiles), highly salted food (sum of salted fish roe and salted fish gut intake category), history of peptic ulcer and family history of gastric cancer (yes or no). We repeated all the analyses after excluding the 66 patients with gastric cancer diagnosed in the first 2 years of follow-up (41 men and 25 women). The p-values for the test of linear trend were 2-sided. RESULTS The distribution of the baseline characteristics according to total vegetable consumption is shown in Table I. Persons with high vegetable consumption were older and had higher gastric cancer rates in their family history. Vegetable consumption was not clearly associated with a history of peptic ulcer or the educational level. Participants with high vegetable consumption also reported a higher prevalence of healthy behavior as indicated by lower rates of current smoking and lower alcohol consumption. However, participants with high vegetable consumption also reported a higher consumption of highly salted food. No appreciable difference was observed in body mass index, rates of any vitamin supplement use and total energy intake. Vegetable consumption was positively correlated with vegetable juice or fruit juice consumption in men but not in women. Table II presents RRs of gastric cancer by frequency of vegetable and fruit consumption. Compared to those who consumed less than 1 day per week of vegetables (except pickled vegetables) and fruit, subjects who consumed them more frequently tended to have a lower risk of gastric cancer. However, the risk did not decline in a stepwise manner as the consumption frequencies increased from 1 2 days per week through almost daily except for yellow vegetables. When the upper 3 categories were combined and compared to the lowest category (less than 1 day per week), fruit consumption was associated with a significantly lower risk, while yellow and white vegetables were associated with a nonsignificantly lower risk. Pickled vegetables were not associated with either a higher or lower risk of gastric cancer. After exclusion of the gastric cancer cases diagnosed in the first and second follow-up

3 VEGETABLES, FRUIT AND GASTRIC CANCER 41 TABLE I CHARACTERISTICS 1 OF SUBJECTS ACCORDING TO QUINTILES OF VEGETABLE CONSUMPTION Men Women 1 (low) (high) 1 (low) (high) Age (years) Body mass index (kg/m 2 ) History of peptic ulcer (%) Family history of gastric cancer (%) Educational level (%) Junior school High school College or higher Smoking status (%) Never Past Current Alcohol consumption (g/day) Energy intake (KJ/day) Highly salted food intake (g/day) Vegetable juice consumption (g/day) Fruit juice consumption (g/day) Use of vitamin A supplement (%) Use of vitamin C supplement (%) Use of vitamin E supplement (%) Data were obtained from food frequency questionnaire. 2 Sum of salted fish roe and salted fish gut. TABLE II RELATIVE RISK (95% CONFIDENCE INTERVALS) OF GASTRIC CANCER BY FREQUENCY OF VEGETABLE AND FRUIT CONSUMPTION: JPHC STUDY Frequency/week 1 day (1) days (2) 3 4 days (3) Almost daily (4) Test for trend p value 1 day/week (2 4) Green vegetables Cases Person-years 8,019 90, , , ,019 RR ( ) 0.77 ( ) 0.73 ( ) ( ) RR ( ) 0.80 ( ) 0.77 ( ) ( ) RR ( ) 0.76 ( ) 0.75 ( ) ( ) Yellow vegetables Cases Person-years 21, , ,092 73, ,745 RR ( ) 0.71 ( ) 0.66 ( ) ( ) RR ( ) 0.71 ( ) 0.66 ( ) ( ) RR ( ) 0.63 ( ) 0.56 ( ) ( ) White vegetables Cases Person-years 6,351 79, , , ,686 RR ( ) 0.50 ( ) 0.55 ( ) ( ) RR ( ) 0.54 ( ) 0.59 ( ) ( ) RR ( ) 0.44 ( ) 0.49 ( ) ( ) Pickled vegetables Cases Person-years 61,449 68,368 64, , ,588 RR ( ) 1.20 ( ) 0.99 ( ) ( ) RR ( ) 1.05 ( ) 0.86 ( ) ( ) RR ( ) 1.12 ( ) 0.97 ( ) ( ) Fruit Cases Person-years 29, , , , ,614 RR ( ) 0.65 ( ) 0.65 ( ) ( ) RR ( ) 0.67 ( ) 0.70 ( ) ( ) RR ( ) 0.69 ( ) 0.70 ( ) ( ) RR, rate ratio. 1 Reference category. 2 Adjusted for age, gender and area. 3 Adjusted for age, gender, area, educational level, smoking status, BMI, alcohol intake, use of vitamin A, C, E supplement, total energy intake, highly salted food intake, history of peptic ulcer and family history of gastric cancer. 4 Analyses with first-year and second-year cases excluded. years, the risk reductions in yellow and white vegetables were clearer. Table III presents RRs of total gastric cancer according to quintiles of total vegetable consumption. Pickled vegetables were excluded from total vegetables. Compared to subjects with the lowest quintiles of vegetable consumption, those with higher quintiles tended to have a lower risk of gastric cancer. The results were similar after the exclusion of gastric cancer cases diagnosed in first and second follow-up years. However, the risk did not decline in a stepwise manner as consumption increased from the second through the highest quintiles. Consequently, tests for a linear trend were not statistically significant. When the second to highest quintiles of consumption were combined and compared to the lowest quintile of total vegetable consumption, a nonsignificantly lower risk was observed. We did not take vegetable juice consumption into account in our estimation of total vegetable consumption because of many miss-

4 42 KOBAYASHI ET AL. TABLE III RELATIVE RISK (95% CONFIDENCE INTERVALS) OF GASTRIC CANCER BY HISTOLOGIC CLASSIFICATION ACCORDING TO QUINTILES OF TOTAL VEGETABLE CONSUMPTION 5 : JPHC STUDY Intake in quintiles/category 1 (low) (high) Test for trend p value Quintile 2 5 Mean intake (g/day) Total gastric cancer Cases Person-years 75,406 75,908 75,541 76,427 75, ,632 RR ( ) 0.75 ( ) 0.92 ( ) 0.74 ( ) ( ) RR ( ) 0.75 ( ) 0.90 ( ) 0.75 ( ) ( ) RR ( ) 0.70 ( ) 0.85 ( ) 0.73 ( ) ( ) Cardia cancer Cases RR ( ) 0.61 ( ) 0.98 ( ) 0.53 ( ) ( ) RR ( ) 0.60 ( ) 1.10 ( ) 0.65 ( ) ( ) RR ( ) 0.65 ( ) 1.17 ( ) 0.52 ( ) ( ) All noncardia Cases RR ( ) 0.85 ( ) 1.05 ( ) 0.88 ( ) ( ) RR ( ) 0.82 ( ) 1.04 ( ) 0.84 ( ) ( ) RR ( ) 0.76 ( ) 1.00 ( ) 0.90 ( ) ( ) Undifferentiated type Cases RR ( ) 0.84 ( ) 1.04 ( ) 1.11 ( ) ( ) RR ( ) 0.87 ( ) 1.06 ( ) 1.22 ( ) ( ) RR ( ) 0.81 ( ) 0.93 ( ) 1.39 ( ) ( ) Differentiated type Cases RR ( ) 0.84 ( ) 0.91 ( ) 0.65 ( ) ( ) RR ( ) 0.78 ( ) 0.88 ( ) 0.53 ( ) ( ) RR ( ) 0.72 ( ) 0.88 ( ) 0.54 ( ) ( ) RR, rate ratio. 1 Reference category. 2 Adjusted for age, gender and area. 3 Adjusted for age, gender, area, educational level, smoking status, BMI, alcohol intake, use of vitamin A, C, E supplement, total energy intake, highly salted food intake, stomach disorders and family history of stomach carcinoma. 4 Analyses with first-year and second-year cases excluded. 5 Pickled vegetables excluded. ing values and the possibility of misclassification from the different number of frequency categories compared to vegetable and fruit. We did not observe any association between vegetable or fruit juice intake and the risk of gastric cancer (data not shown). Nor did we find any substantial modification of the association between vegetables, fruit and the risk of gastric cancer in analyses stratified by alcohol consumption, smoking status, high and low consumption of nitrate- and nitrite-containing foods including processed meat and dried fish (data not shown). These findings did not differ when the analyses were limited to men (data not shown). Table III also presents RRs of subsites and histologic typespecific gastric cancer according to the quintiles of total vegetable consumption. No striking differences in associations were seen according to subsite-specific gastric cancer, although apparent protective effects appeared in the differentiated type of gastric cancer. DISCUSSION In this population-based prospective study in Japan, both vegetables and fruits were associated with a lower risk of gastric cancer. This risk was fairly constant over the consumption of 1 2 or more days of vegetables and fruits per week and also across the second through the fifth quintiles of total vegetable intake. Furthermore, a comparison of less than 1 day per week to at least 1 day per week of vegetable and fruit consumption and a comparison of the lowest quintile to the combination of the other four quintiles of total vegetable consumption showed inverse associations with the risk of gastric cancer. Our results suggest that consumption of vegetables and fruit as low as 1 day per week may serve to protect against gastric cancer. We also found that the association between vegetable and fruit consumption and the risk of gastric cancer did not vary substantially according to the anatomical subsites or the histologic subtypes. Our findings are in general agreement with those of 4 prospective studies, which show an inverse association between the risk of gastric cancer and the consumption of vegetables and fruits. In an early study of 122,261 Japanese men, even the rare or occasional consumption of green and yellow vegetables, compared to no consumption, was significantly associated with lower risk of gastric cancer. 5 In 2 studies among Japanese Hawaiians, 1 study found an inverse association with the consumption of vegetables and fruits, 11 while the other revealed an inverse association with the consumption of fruits and the combined intake of vegetables and fruits, but no association with regard to vegetable intake. 12 A study of Swedish twins showed a significant inverse association for the combined intake of vegetables and fruits. 13 In contrast, 7 prospective studies found no association of both vegetable and fruit consumption with the risk of gastric cancer. 6 10,14,15 Higher consumption of vegetables and fruits has been associated with a reduced risk of gastric cancer in numerous case-control studies. 3 However, in such studies, reports of vegetable and fruit consumption are susceptible to biased recall of past diet. 34 In our study, we assessed the consumption of vegetables and fruits before the gastric cancer and other diseases were diagnosed, thus avoiding recall bias. Furthermore, we observed similar results in the analysis with the exclusion of gastric cancer cases diagnosed in the first 2 years of follow-up. These findings suggest that the protective effect of vegetables and fruits demonstrated in case-control studies is not entirely attributable to recall bias. Because a large number of gastric cancer cases were identified in our cohort study, it was possible to analyze the association between vegetable and fruit consumption and the risk of gastric cancer by histologic subtypes and anatomic subsites. Some studies suggest that etiologic or environmental factors for cardia cancer were different from those for noncardia cancer. 35 However, in our study, no remarkable difference was observed in the association

5 VEGETABLES, FRUIT AND GASTRIC CANCER 43 between vegetable or fruit consumption and the risk of cardia or noncardia cancers. With regard to the relation between the histologic type of distal cancer and vegetables and fruits, 1 case-control study in the United States reported vegetables and fruits may play an important role in preventing the intestinal type in men and women combined, 24 but the other study in Japan reported the intake of vegetables and fruits was only associated with intestinal type in women and not in men. 36 In contrast, 2 other case-control studies reported the same pattern of association between vegetable and fruit consumption and both the intestinal and diffuse type. 23,25 Some studies reported that current smokers showed a strong increase in risk for gastric cancer 37 and that vegetable and fruit consumption is more protective among smokers than nonsmokers. 38 This is because of the relative deficit of vegetable and fruit consumption in smokers who then experience a marked riskreducing effect when their intake increases. In our study, the percentage of current smokers was relatively high even in the highest quintile of vegetable consumption. Hence, the difference in risk reduction between current smokers and nonsmokers was not readily apparent. Results were inconclusive for nitrate- and nitritecontaining foods and gastric cancer in an epidemiologic study, 3 although many N-nitroso compounds are known to be carcinogenic from animal experiments. 39 In our study, no association was observed between pickled vegetable consumption and the risk of gastric cancer. The intake frequency of pickled vegetables was relatively common, being categorized as almost daily in 48.6%. The consumption of pickled vegetables might have been more varied in subjects consuming almost daily. The consumption of processed meat and dried fish had no substantial effect on the association between vegetable and fruit consumption and the risk of gastric cancer. We did not measure H. pylori infection, a strong risk factor for gastric cancer. 40 Our previous ecologic study showed a prevalence of H. pylori seropositivity from 63 76% among subjects years of age in these 4 study areas. However, it was not associated with vegetable and fruit intake in a cross-sectional analysis. 41 But a synergistic interaction between H. pylori infection and dietary factors in gastric cancer has been suggested. 42 On the other hand, the study in Sweden found that ascorbic acid and beta carotene were protective in both cardia and noncardia cancer only in H. pylori-infected subjects. 25 In our study, the histologic classification of gastric cancer cases was based on the available data from pathology records, but the original specimens were neither collected nor reclassified. This limitation may have led to histologic misclassification to some extent and contributed to the absence of differences in the findings according to the histologic subtypes. We used a relatively crude food frequency questionnaire, which includes only 44 food items (only 5 questions for fruits and vegetables), uses rough frequency scales (4 or 6 categories), does not ask or specify portion size information and asks about the recent consumption (rather than the average consumption during a previous year as usual). This is because we had not originally intended, at the outset of our study, to estimate the absolute amount of consumption of foods and nutrients based on this questionnaire. Nevertheless, we observed moderate correlations between many of the food and nutrient intakes estimated from this questionnaire with the consumption measured by 28- or 14-day diet records collected from a subsample of participants in the main cohort study over 4 or 2 seasons of the year. 29 The results of this validation study indicate that our questionnaire has practical utility in assessing the relative consumption of various nutrients and foods including fruits and vegetables. The nondifferential misclassification of fruit and vegetable consumption would have led to the underestimation of the inverse association between these foods and the risk of gastric cancer. Since vegetable intake information was gathered using only 3 questions, it was difficult to classify the details beyond the quintile level, suggesting that total vegetable intake was underestimated from the FFQ. In our validation study, the mean daily total vegetable intake of g was estimated from diet records, against only g from the FFQ in the same subjects. This underestimation makes it difficult to determine the exact weight to assign vegetable intake for reducing the risk of gastric cancer. In conclusion, our investigation supported that vegetable and fruit intake, even in low amounts, is associated with a lower risk of gastric cancer. Although no striking differences in the association were seen between cardia and noncardia cancer, an inverse association was higher in differentiated rather than in undifferentiated types of gastric cancer. ACKNOWLEDGEMENTS M.K. was awarded a Research Resident Fellowship from the Foundation for Promotion of Cancer Research in Japan. REFFERENCES 1. Parkin DM, Pisani P, Ferlay J. Estimates of the worldwide incidence of eighteen major cancers in Int J Cancer 1993;54: Steinmetz KA, Potter JDN. Vegetables, fruit, and cancer. I. Epidemiology. Cancer Causes Control 1991;2: Kono S, Hirohata TN. Nutrition and stomach cancer. Cancer Causes Control 1996;7: Willett WC. Diet and cancer. Oncologist 2000;5: Hirayama T. Nutrition and cancer a large scale cohort study. Prog Clin Biol Res 1986;206: Nomura A, Grove JS, Stemmermann GN, Severson RKN. A prospective study of stomach cancer and its relation to diet, cigarettes, and alcohol consumption. Cancer Res 1990;50: Kneller RW, McLaughlin JK, Bjelke E, Schuman LM, Blot WJ, Wacholder S, Gridley G, CoChien HT, Fraumeni JF Jr. A cohort study of stomach cancer in a high-risk American population. Cancer 1991; 68: Chyou PH, Nomura AM, Hankin JH, Stemmermann GN. A casecohort study of diet and stomach cancer. Cancer Res 1990;50: Kato I, Tominaga S, Ito Y, Kobayashi S, Yoshii Y, Matsuura A, Kameya A, Kano T, Ikari A. A prospective study of atrophic gastritis and stomach cancer risk. Jpn J Cancer Res 1992;83: Guo W, Blot WJ, Li JY, Taylor PR, Liu BQ, Wang W, Wu YP, Zheng W, Dawsey SM, Li B, Fraumeni JF Jr. A nested case-control study of oesophageal and stomach cancers in the Linxian nutrition intervention trial. Int J Epidemiol 1994;23: Nomura AM, Stemmermann GN, Chyou PH. Gastric cancer among the Japanese in Hawaii. Jpn J Cancer Res 1995;86: Galanis DJ, Kolonel LN, Lee J, Nomura A. Intakes of selected foods and beverages and the incidence of gastric cancer among the Japanese residents of Hawaii: a prospective study. Int J Epidemiol 1998;27: Terry P, Nyren O, Yuen J. Protective effect of fruits and vegetables on stomach cancer in a cohort of Swedish twins. Int J Cancer 1998;76: Botterweck AA, van den Brandt PA, Goldbohm RA. A prospective cohort study on vegetable and fruit consumption and stomach cancer risk in The Netherlands. Am J Epidemiol 1998;148: McCullough ML, Robertson AS, Jacobs EJ, Chao A, Calle EE, Thun MJ. A prospective study of diet and stomach cancer mortality in United States men and women. Cancer Epidemiol Biomarkers Prev 2001;10: Yang PC, Davis S. Epidemiological characteristics of adenocarcinoma of the gastric cardia and distal stomach in the United States, Int J Epidemiol 1988;17: Craanen ME, Dekker W, Blok P, Ferwerda J, Tytgat GN. Time trends in gastric carcinoma: changing patterns of type and location. Am J Gastroenterol 1992;87: Inoue M, Tajima K, Hirose K, Kuroishi T, Gao CM, Kitoh T. Life-style and subsite of gastric cancer joint effect of smoking and drinking habits. Int J Cancer 1994;56: Kaneko S, Yoshimura T. Time trend analysis of gastric cancer incidence in Japan by histological types, Br J Cancer 2001; 84: Uemura N, Okamoto S, Yamamoto S, Matsumura N, Yamaguchi S, Yamakido M, Taniyama K, Sasaki N, Schlemper RJ. Helicobacter

6 44 KOBAYASHI ET AL. pylori infection and the development of gastric cancer. N Engl J Med 2001;345: Hccg. H. Gastric cancer and Helicobacter pylori: a combined analysis of 12 case control studies nested within prospective cohorts. Gut 2001;49: Buiatti E, Palli D, Bianchi S, Decarli A, Amadori D, Avellini C, Cipriani F, Cocco P, Giacosa A, Lorenzini L, Marubini E, Puntoni R, et al. A case-control study of gastric cancer and diet in Italy. III. Risk patterns by histologic type. Int J Cancer 1991;48: Boeing H, Jedrychowski W, Wahrendorf J, Popiela T, Tobiasz-Adamczyk B, Kulig A. Dietary risk factors in intestinal and diffuse types of stomach cancer: a multicenter case-control study in Poland. Cancer Causes Control 1991;2: Harrison LE, Zhang ZF, Karpeh MS, Sun M, Kurtz RC. The role of dietary factors in the intestinal and diffuse histologic subtypes of gastric adenocarcinoma: a case-control study in the U.S. Cancer 1997;80: Ekstrom AM, Serafini M, Nyren O, Hansson LE, Ye W, Wolk A. Dietary antioxidant intake and the risk of cardia cancer and noncardia cancer of the intestinal and diffuse types: a population-based casecontrol study in Sweden. Int J Cancer 2000;87: Tsugane S, Fahey MT, Sasaki S, Baba S. Alcohol consumption and all-cause and cancer mortality among middle-aged Japanese men: seven-year follow-up of the JPHC study Cohort I. Japan Public Health Center. Am J Epidemiol 1999;150: Tsugane S, Tsuda M, Gey F, Watanabe S. Cross-sectional study with multiple measurements of biological markers for assessing stomach cancer risks at the population level. Environ Health Perspect 1992; 98: Tsubono Y, Kobayashi M, Tsugane SN. Food consumption and gastric cancer mortality in five regions of Japan. Nutr Cancer 1997;27: Tsubono Y, Sasaki S, Kobayashi M, Akabane M, Tsugane S. Food composition and empirical weight methods in predicting nutrient intakes from food frequency questionnaire. Ann Epidemiol 2001;11: Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, Hennekens CH, Speizer FE. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985;122: WHO. International classification of disease for cncology, 2nd ed. Geneva: World Health Organization, Japanese Research Society for Gastric Cancer. The general rules for the gastric cancer study, 12th ed. Tokyo: Kanehara, SAS/STAT User s Guide, version 8, vol 2. Cary, NC: SAS, Samet JM. Surrogate sources of dietary information. Nutr Epidemiol 1990: Inoue M, Tajima K, Yamamura Y, Hamajima N, Hirose K, Kodera Y, Kito T, Tominaga S. Family history and subsite of gastric cancer: data from a case-referent study in Japan. Int J Cancer 1998;76: Kato I, Tominaga S, Ito Y, Kobayashi S, Yoshii Y, Matsumura A, Kameya A, Kano T. A comparative case-control analysis of stomach cancer and atrophic gastritis. Cancer Res 1990;50: Tredaniel J, Boffetta P, Buiatti E, Saracci R, Hirsch A. Tobacco smoking and gastric cancer: review and meta-analysis. Int J Cancer 1997;72: Hansson LE, Baron J, Nyren O, Bergstrom R, Wolk A, Adami HO. Tobacco, alcohol and the risk of gastric cancer. A population-based case-control study in Sweden. Int J Cancer 1994;57: Mirvish SS. The etiology of gastric cancer. Intragastric nitrosamide formation and other theories. J Natl Cancer Inst 1983;71: Graham DY. Helicobacter pylori infection is the primary cause of gastric cancer. J Gastroenterol 2000;35: Tsugane S, Tei Y, Takahashi T, Watanabe S, Sugano K. Salty food intake and risk of Helicobacter pylori infection. Jpn J Cancer Res 1994;85: Yamaguchi N, Kakizoe T. Synergistic interaction between Helicobacter pylori gastritis and diet in gastric cancer. Lancet Oncol 2001;2: