Portal de Periódicos da CAPES

Cohort analysis of fruit and vegetable consumption and lung cancer mortality in European men

2001; Wiley; Volume: 92; Issue: 6 Linguagem: Inglês

10.1002/ijc.1278

1097-0215

Margje C.J.F. Jansen, H. Bas Bueno-de-Mesquita, Leena R�s�nen, Flaminio Fidanza, Aulikki Nissinen, Alessandro Menotti, Frans J. Kok, D Kromhout,

Cancer Risks and Factors

International Journal of CancerVolume 92, Issue 6 p. 913-918 Research ArticleFree Access Cohort analysis of fruit and vegetable consumption and lung cancer mortality in European men Margje C.J.F. Jansen, Corresponding Author Margje C.J.F. Jansen [email protected] Department of Chronic Diseases Epidemiology, National Institute of Public Health and the Environment, Bilthoven, The Netherlands Division of Human Nutrition and Epidemiology, Wageningen University, Wageningen, The Netherlands Fax: +31-30-274-4407National Institute of Public Health and the Environment, Department of Chronic Diseases Epidemiology, PO Box 1, 3720 BA Bilthoven, The NetherlandsSearch for more papers by this authorH. Bas Bueno-de-Mesquita, H. Bas Bueno-de-Mesquita Department of Chronic Diseases Epidemiology, National Institute of Public Health and the Environment, Bilthoven, The NetherlandsSearch for more papers by this authorLeena Räsänen, Leena Räsänen Division of Nutrition, University of Helsinki, Helsinki, FinlandSearch for more papers by this authorFlaminio Fidanza, Flaminio Fidanza Nutrition Section, Department of Internal Medicine, University of Perugia, Perugia, ItalySearch for more papers by this authorAulikki M. Nissinen, Aulikki M. Nissinen Department of Public Health and General Practice, University of Kuopio, Kuopio, FinlandSearch for more papers by this authorAlessandro Menotti, Alessandro Menotti Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, MN, USASearch for more papers by this authorFrans J. Kok, Frans J. Kok Division of Human Nutrition and Epidemiology, Wageningen University, Wageningen, The NetherlandsSearch for more papers by this authorDaan Kromhout, Daan Kromhout Division of Public Health Research, National Institute of Public Health and the Environment, Bilthoven, The NetherlandsSearch for more papers by this author Margje C.J.F. Jansen, Corresponding Author Margje C.J.F. Jansen [email protected] Department of Chronic Diseases Epidemiology, National Institute of Public Health and the Environment, Bilthoven, The Netherlands Division of Human Nutrition and Epidemiology, Wageningen University, Wageningen, The Netherlands Fax: +31-30-274-4407National Institute of Public Health and the Environment, Department of Chronic Diseases Epidemiology, PO Box 1, 3720 BA Bilthoven, The NetherlandsSearch for more papers by this authorH. Bas Bueno-de-Mesquita, H. Bas Bueno-de-Mesquita Department of Chronic Diseases Epidemiology, National Institute of Public Health and the Environment, Bilthoven, The NetherlandsSearch for more papers by this authorLeena Räsänen, Leena Räsänen Division of Nutrition, University of Helsinki, Helsinki, FinlandSearch for more papers by this authorFlaminio Fidanza, Flaminio Fidanza Nutrition Section, Department of Internal Medicine, University of Perugia, Perugia, ItalySearch for more papers by this authorAulikki M. Nissinen, Aulikki M. Nissinen Department of Public Health and General Practice, University of Kuopio, Kuopio, FinlandSearch for more papers by this authorAlessandro Menotti, Alessandro Menotti Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, MN, USASearch for more papers by this authorFrans J. Kok, Frans J. Kok Division of Human Nutrition and Epidemiology, Wageningen University, Wageningen, The NetherlandsSearch for more papers by this authorDaan Kromhout, Daan Kromhout Division of Public Health Research, National Institute of Public Health and the Environment, Bilthoven, The NetherlandsSearch for more papers by this author First published: 03 May 2001 https://doi.org/10.1002/ijc.1278Citations: 27 AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Abstract Our aim was to examine the relationship between fruit and vegetable consumption and lung cancer mortality in a cohort of European males. Around 1970, dietary intake of Finnish, Italian and Dutch middle-aged men was assessed using a cross-check dietary history. Complete baseline information was available for 3,108 men, of whom 1,578 were baseline smokers. We used Cox proportional hazard analyses to calculate risk estimates for the consumption in country-specific tertiles on lung cancer in smokers. During 25 years of follow-up, 149 lung cancer deaths occurred in the smokers. Fruit consumption was inversely associated with lung cancer mortality among smokers; compared with the lowest, adjusted RRs for the intermediate and highest tertiles were 0.56 (0.37–0.84) and 0.69 (0.46–1.02), p-trend 0.05. Only in the Dutch cohort was this association statistically significant [adjusted relative risks (RRs) 1.00, 0.33 (0.16–0.70) and 0.35 (0.16–0.74), p-trend 0.004]. In Finland lung cancer risk was lower with higher fruit intake but not significantly, whereas in Italy no association was observed. Stratifying on cigarette smoking intensity (non, light and heavy) revealed an inverse association in the heavy smokers only [adjusted RRs (95% confidence intervals [CI]) 1; 0.47 (0.26–0.84); 0.40 (0.20–0.78)). Vegetable consumption was not related to lung cancer risk in smokers. However, analyses stratified on cigarette smoking intensity gave some indication for a lower lung cancer risk with higher intake. In conclusion, in this prospective analysis among European smoking men, fruit intake was inversely related to lung cancer mortality. This association was confined to heavy cigarette smokers. © 2001 Wiley-Liss, Inc. Lung cancer is a major public health problem because of both high incidence1 and high fatality rates.2 Cigarette smoking is the dominant risk factor for this type of cancer; with an attributable risk of 90% in men.3 Additionally, diet may influence lung cancer risk. Many epidemiological studies have shown that fruit and vegetable intake is associated with a lower lung cancer risk.4-8 The estimated relative risks for high vs. low consumption vary considerably between studies, and the overall estimate ranges from 0.455 to 0.5–0.7;8 a formal meta-analysis has not been performed. Although the inverse association between fruit and vegetable intake and lung cancer seems convincing, some issues have been raised that may weaken this relationship. First, residual confounding by smoking: Smoking is such a dominant factor in lung cancer etiology that adjustment for this risk factor, especially when only smoking status is used, may not fully remove its effect. Smokers tend to eat less fruit and vegetables,9 and probably have a less healthy lifestyle in general, which may result in residual confounding. Second, it has been suggested that results from cohort studies, compared with case-control studies, are less conclusive.10 Case-control studies may overestimate the association due to recall and selection bias. Cigarette smokers have a high exposure to carcinogens. Therefore, potential anticarcinogens provided by fruit and vegetables are expected to exert their action especially in this group. However, results by smoking status are not consistent. Some studies found inverse associations especially in current smokers,11 in heavy smokers,12, 13 and others in light smokers,14, 15 in former smokers14,16 or in nonsmokers.17, 18 We examined the relationship between fruit and vegetable consumption and lung cancer mortality in a prospective study among males (n = 3,108) from Finland, Italy and The Netherlands, all participating in the Seven Countries Study. Baseline data from around 1970 were used and mortality was followed for 25 years. Because lung cancer mainly occurs in smokers, we restricted most analyses to baseline smokers (n = 1,578). MATERIAL AND METHODS Study population Between 1958 and 1964, 16 population samples of men aged 40–59 years from 7 countries were enrolled and examined for the Seven Countries Study, with a participation rate of more than 90%.19 In Finland this study started in 1959 and in Italy and The Netherlands in 1960. Dietary intake was only assessed in a small subgroup per cohort at that time. Around 1965, a second round of this study was conducted, however, no dietary information was gathered in Finland. As baseline for the present analysis, we used the third round of the study, which was around 1970. We used data of 5 of the original 16 population samples: 2 in Finland (East Finland and West Finland), 2 in Italy (Crevalcore and Montegiorgio) and 1 in The Netherlands (Zutphen) because dietary data were available. The Dutch cohort consisted of inhabitants of a small commercial town, Zutphen, whereas the other cohorts were situated in rural areas. Information was gathered in Finland in 1969 (n = 612 in East Finland and n = 694 in West Finland), in Zutphen (n = 615) and in Crevalcore (n = 592) in 1970. In the Montegiorgio cohort, dietary information was gathered in 1970 only from a subset. Therefore, the dietary data collected in 1965 from the men still alive in 1970 were used as an approximation of dietary intake in 1970 (n = 662 men). Although fruit and vegetable consumption was higher in the subset of men in 1970, differences were not statistically significant.20 For analyses, complete information was available for 3,108 men, of whom 1,578 were baseline cigarette smokers. Examinations Food intake around 1970 was estimated by using the cross-check dietary history method. This method provides information about the usual food consumption pattern 6 to 12 months preceding the interview.21 First, the usual food consumption pattern of a person during weekdays and weekends was assessed. This part concerned questions about the foods used at breakfast, lunch, dinner and between the meals. Thereafter, a checklist with an extensive number of foods was used to calculate and to verify the participant's food consumption pattern. Experienced dietitians and nutritionists carried out the interviews. In Finland, the dietary surveys were held in autumn and in Italy and The Netherlands in spring. Although this method was adjusted to the local situation, the methodology was comparable. In all countries food items were categorized into food groups, such as fruit and vegetables, in the same way. However, the types of fruits and vegetables within the food groups could differ per country. The nutrient intake was assessed using computerized versions of the local food tables for the countries.22-24 In all cohorts, information on age and smoking of the participants was collected in a standardized way.25 Men were asked to report their current smoking status (never, former or current smoker). Current and former smokers reported the number of cigarettes they smoked or used to smoke per day (1–4, 5–9, 10–19, 20–29 or more than 30 cigarettes). The midpoints of these categories (2, 7, 15, 25 and 35) were used as number of cigarettes smoked per day. For 70 men, of which 63 were from Montegiorgio, baseline information on smoking was missing. For 63 of the 70 men, smoking data of 1965 were appropriate to use as a proxy for the 1970 data. No ethical guidelines existed at the time of the first surveys, however, the study was retrospectively approved by the Medical Ethical Committee of the University of Leiden, The Netherlands in 1985. Follow-up The participating men were followed for mortality during 25 years, i.e., from around 1970 to 1995. None of them was lost to follow-up. The underlying causes of death were coded in a standardized way by one reviewer, using the 8th revision of the WHO International Classification of Diseases (ICD). The cause of death was based on information from the official death certificate, in combination with information from medical and hospital records. In case of multiple causes of death, priority was given to accidents, followed by cancer in advanced stages, coronary heart disease and stroke. For the present analyses, lung cancer mortality was defined as ICD8 code 162 as the primary (n = 183) or secondary (n = 4) cause of death. Statistical methods Cox proportional hazard analysis was used to associate intake of fruit and vegetables with 25-year lung cancer mortality. Fruit intake was analyzed as total fruit consumption, that is the sum of fresh fruit, dried fruit multiplied by 2 (to account for water loss), canned fruit and fruit juices. In Italy and The Netherlands, almost all fruit was eaten as fresh fruit. In Finland, however, other forms of fruit, especially fruit juices (mainly berry juice), were substantially consumed (Table I). Therefore, analyses for fresh fruit were performed as well. Vegetable consumption did not include potatoes. Table I. Baseline Characteristics and Daily Intake (Mean ± SD) of Male Nonsmokers (Never + Former Smokers) and of Male Current Smokers in the Finnish, Italian and Dutch Cohorts Studied and Mortality During 25 Years of Follow-Up Finland Italy The Netherlands Nonsmokers (n = 651) Current smokers (n = 637) Nonsmokers (n = 591) Current smokers (n = 616) Nonsmokers (n = 288) Current smokers (n = 325) Age (years) 59.4 ± 5.5 58.7 ± 5.5 59.8 ± 5.0 59.0 ± 4.9 60.4 ± 5.4 58.7 ± 5.2 Former smokers (% of total) 31.2 20.0 39.5 Never smokers (% of total) 19.3 29.0 7.5 Number of cigarettes daily 17.5 ± 9.0 11.7 ± 7.7 15.0 ± 8.5 Body mass index (kg/m2) 25.8 ± 3.7 23.7 ± 3.7 27.1 ± 3.8 25.0 ± 3.8 25.4 ± 2.7 24.8 ± 2.8 Energy (MJ) 15.3 ± 4.6 15.7 ± 4.8 12.1 ± 3.5 12.3 ± 3.1 10.8 ± 2.2 11.0 ± 2.3 Fat (En%) 36.7 ± 6.7 38.0 ± 6.9 28.1 ± 8.0 28.0 ± 7.6 40.5 ± 5.6 41.4 ± 5.1 Saturated fat (En%) 21.3 ± 4.6 22.2 ± 4.7 9.3 ± 3.7 9.6 ± 3.6 16.5 ± 3.0 16.8 ± 2.8 Fruit (g) 186 ± 195 172 ± 173 162 ± 181 137 ± 146 192 ± 146 147 ± 110 Fresh fruit (g) 87 ± 84 72 ± 74 161 ± 181 137 ± 145 190 ± 145 144 ± 109 Vegetables (g) 80 ± 57 79 ± 63 71 ± 51 64 ± 55 184 ± 60 177 ± 57 Potatoes (g) 283 ± 142 288 ± 148 24 ± 23 22 ± 21 186 ± 102 191 ± 97 Total mortality (n) 457 532 378 446 195 233 Person-years 11,141 9,076 10,146 10,105 4,960 5,346 Total cancer mortality (n) 94 142 100 152 61 88 Lung cancer mortality (n) 17 68 5 36 16 45 Former smokers 12 1 15 Never smokers 5 4 1 Lung cancer mortality rate (per 10,000 person-years) 74.9 35.6 84.2 Consumption was divided in country-specific tertiles. Relative risks (RRs) and 95% confidence intervals (95% CI) were calculated, with the lowest tertile as the reference category. Statistical significance was determined by 2-sided tests with a critical value of 0.05 or less. Countries were analyzed separately, combining the 2 cohorts in Finland and the 2 cohorts in Italy. Because no effect modification by country was present, analyses were also performed for all countries pooled, using the STRATA option of the PHREG procedure.26 Survival analyses were done crudely, adjusted for age and number of cigarettes per day (model 1) and for model 1 plus energy intake, fruit intake (in vegetable intake analyses) or vegetable intake (in fruit intake analyses) and for country (in pooled analyses) or cohort (in analyses for Finland and Italy) (model 2). Tests for trend were performed by assigning the integers 0, 1 and 2 to the tertiles of the intake variables. Pearson correlations between fruit and vegetable intake ranged from −0.01 to 0.24; i.e., low enough to combine these variables in one model. Because some studies indicate that (saturated) fat may be positively associated with lung cancer risk,6 energy intake in model 2 was replaced by intakes of fat or saturated fat to evaluate their confounding effect on the relation between fruit and vegetable intake and lung cancer mortality. No such effect was observed (data not shown). Information on prevalence of total cancer at baseline was present but not for lung cancer specifically. Therefore, to examine the potential confounding effect of (subclinical) lung cancer cases at baseline, analyses were repeated after excluding the lung cancer deaths within the first 2 (n = 11) and 5 years (n = 28) of follow-up. Since the exclusion of these cases did not affect the associations for fruit and vegetables in any material way, results are presented for all cases. Also restricting the follow-up time to 20 years did not substantially change the results for fruit and for vegetable intake (data not shown). To explore the shape of the relationship between fruit and vegetable intake and lung cancer mortality, rates expressed as number of lung cancer deaths per 10,000 person-years per tertile of intake were plotted. These rates were directly standardized for age and number of cigarettes smoked at baseline using the total population of current cigarette smokers as reference group. To examine whether the associations with fruit and vegetable intake differed by smoking intensity, baseline smokers were categorized as light smokers, i.e., 1–19 cigarettes per day, and heavy smokers, i.e., 20 or more cigarettes per day. Moreover, never and former smokers were combined as nonsmokers. Limited cases in the latter 2 groups did not allow us to separate them in the analyses. Joint effects of smoking intensity (non, light and heavy smokers) and country-specific tertiles of intake were calculated but only for the countries pooled because of the small number of cases per country. All analyses were performed using the Statistical Analysis Systems (SAS) software package (version 6.12). RESULTS Roughly half of the men in each country smoked cigarettes around 1970. Prevalences of never and former smokers differed per country. The mean number of cigarettes smoked was highest in Finland and lowest in Italy. Age at baseline was around 59 years on average, with baseline smokers being slightly younger and leaner (Table I). Dietary intake varied across the countries, although within a country, smokers and nonsmokers did not differ substantially on intake, except for intake of fruit, with nonsmokers consuming more fruits (Table I). During 25 years of follow-up, 187 men died from lung cancer, of which 149 were baseline smokers, 28 former smokers and 10 never smokers. The lung cancer mortality rate among smokers was highest in The Netherlands (84.2 per 10,000 person-years), followed by Finland (74.9 per 10,000 person-years) and lowest in Italy (35.6 per 10,000 person-years) (Table I). When pooling the populations, fruit consumption was inversely associated with lung cancer mortality in smokers (p-trend = 0.05) (Table II). Adjustment for potential confounders did not materially change the risk estimates, although statistical significance was no longer reached. Analyses for smokers per country showed a statistically significant inverse association for the Dutch cohort only (adjusted p-trend = 0.004). In Finland, relative risks were below unity but not statistically significant, and in Italy no association was found. In contrast to fruit, the fresh fruit consumption in Finland showed only a nonsignificant lower risk in the highest tertile: RRs (adjusted model 2, 95% CI) 1.00; 1.18 (0.66–2.11); 0.79 (0.41–1.51). Pooling the effect of fresh fruit across the 3 countries resulted in the following RRs (95% CI), adjusted according to model 2: 1.00; 0.83 (0.56–1.22); 0.70 (0.47–1.06), with a p-trend of 0.09. Vegetable consumption was not related to lung cancer mortality in smokers (Table II). Table II. Relative Risks and 95% CI of 25-Year Lung Cancer Mortality According to Fruit and Vegetable Consumption in Country-Specific Tertiles in Smoking Men Aged 50–69 Years, Pooled and Per Country Fruit consumption Vegetable consumption Low Intermediate High p-trend Low Intermediate High p-trend Populations pooled (country-specific tertiles) Cases (n)/total (n) 62/523 40/529 47/526 56/528 43/522 50.528 Person-years 7,770 8,368 8,389 7,770 8,281 8,476 Mortality rate (n/10,000 py) 79.8 47.8 56.0 72.1 51.9 59.0 RR crude 1 0.59 (0.40–0.88) 0.69 (0.47–1.00) 0.05 1 0.71 (0.48–1.05) 0.80 (0.54–1.17) 0.24 RR adjusted model 1 1 0.58 (0.39–0.86) 0.70 (0.48–1.02) 0.05 1 0.76 (0.51–1.13) 0.92 (0.63–1.36) 0.66 RR adjusted model 2 1 0.56 (0.37–0.84) 0.69 (0.46–1.02) 0.05 1 0.73 (0.49–1.09) 0.90 (0.61–1.33) 0.59 Finland Cases (n)/total (n) 27/212 19/213 22/212 26/213 18/212 24/212 Person-years 2,857 3,113 3,105 2,946 3,037 3,092 Mortality rate (n/10,000 py) 94.5 61.0 70.8 88.2 59.3 77.6 Median intake (g/d) (range) 44.4 (0–79.8) 118.1 (80.2–177.6) 289.1 (177.8–1221) 25.7 (0–44.7) 62.4 (45.2–91.2) 131 (91.6–530.8) RR crude 1 0.63 (0.35–1.13) 0.74 (0.42–1.30) 0.29 1 0.66 (0.36–1.21) 0.87 (0.50–1.52) 0.63 RR adjusted model 1 1 0.61 (0.34–1.09) 0.71 (0.40–1.25) 0.23 1 0.71 (0.39–1.30) 0.98 (0.56–1.72) 0.93 RR adjusted model 2 1 0.63 (0.34–1.14) 0.82 (0.45–1.50) 0.50 1 0.69 (0.38–1.27) 0.99 (0.56–1.78) 0.96 Italy Cases (n)/total (n) 12/203 10/207 14/206 14/205 10/202 12/209 Person-years 3,294 3,412 3,400 3,040 3,341 3,724 Mortality rate (n/100,000 py) 36.4 29.3 41.2 46.1 29.9 32.3 Median intake (g/d) (range) 5 (0–48) 100 (50–158) 247.5 (159–1000) 18 (0–37) 55 (38–74) 98 (75–392) RR crude 1 0.81 (0.35–1.87) 1.14 (0.53–247) 0.72 1 0.64 (0.28–1.43) 0.65 (0.30–1.40) 0.28 RR adjusted model 1 1 0.79 (0.34–1.84) 1.12 (0.52–2.44) 0.75 1 0.81 (0.36–1.84) 1.02 (0.46–2.26) 0.98 RR adjusted model 2 1 0.79 (0.33–1.89) 1.08 (0.45–2.62) 0.85 1 0.83 (0.36–1.90) 1.05 (0.47–2.35) 0.92 The Netherlands Cases (n)/total (n) 23/108 11/109 11/108 16/110 15/108 14/107 Person-years 1,619 1,844 1,883 1,784 1,903 1,659 Mortality rate (n/10,000 py) 142.0 59.7 58.5 89.7 78.8 84.6 Median intake (g/d) (range) 43 (0–93) 130 (94–175) 241.5 (179–850) 126.5 (0–154) 175 (155–194) 223 (195–530) RR crude 1 0.39 (0.19–0.80) 0.39 (0.19–0.79) 0.006 1 0.84 (0.42–1.71) 0.95 (0.46–1.95) 0.87 RR adjusted model 1 1 0.40 (0.19–0.82) 0.42 (0.20–0.88) 0.01 1 0.82 (0.40–1.66) 0.92 (0.45–1.88) 0.80 RR adjusted model 2 1 0.33 (0.16–0.70) 0.35 (0.16–0.74) 0.004 1 0.82 (0.40–1.66) 0.88 (0.43–1.82) 0.73 Model 1: adjusted for age, number of cigarettes smoked at baseline. Model 2: adjusted for age, number of cigarettes smoked at baseline, country (for pooled analysis)/cohort (for Finland/Italy), energy intake and vegetable intake (for fruit)/fruit intake (for vegetable). We plotted country-specific lung cancer mortality rates in smokers against tertiles of fruit intake (median), standardized for age and numbers of cigarettes per day (Fig. 1). This figure indicates that among smokers, fruit intake may be only inversely associated with lung cancer mortality in those with low intake and at high absolute risk. The absolute risk was low in Italy and did not change with fruit intake, as we observed using relative risks. For the Zutphen cohort and to a lesser extent for the Finnish cohorts, the absolute risks decreased with intake in a way suggestive for a log-linear relationship. Figure 1Open in figure viewerPowerPoint Lung cancer mortality rates standardized for age and number of cigarettes according to median of fruit intake in country-specific tertiles in smoking men aged 50–69 years followed for 25 years. F1, lowest tertile Finland; F2, intermediate tertile Finland; F3, highest tertile Finland; I1, lowest tertile Italy; I2, intermediate tertile Italy; I3, highest tertile Italy; N1, lowest tertile The Netherlands; N2, intermediate tertile The Netherlands; N3, highest tertile The Netherlands. In Figure 2a, the joint effects of cigarette smoking intensity (non, light and heavy) and fruit intake adjusted for age, energy intake, vegetable intake and country are shown. An inverse association was observed in the heavy smokers [RR (95% CI) 1; 0.47 (0.26–0.84); 0.40 (0.20–0.78)]; in the light smokers no relation was seen. Figure 2b consists of the same figure for vegetable intake. An inverse association in the heavy smokers was indicated, however, risk estimates did not reach statistical significance [RR (95% CI) 1; 0.92 (0.52–1.65); 0.64 (0.34–1.22)]. The lung cancer risk among nonsmokers was about 10% of the risk in the reference group of heavy smokers with a low fruit/vegetable intake. Figure 2Open in figure viewerPowerPoint (a) Relative risks of 25-year lung cancer mortality according to country-specific tertiles of fruit intake and smoking status, adjusted for age, country, energy intake and vegetable intake. (b) Relative risks of 25-year lung cancer mortality according to country-specific tertiles of vegetable intake and smoking status, adjusted for age, country, energy intake and fruit intake. DISCUSSION In this prospective study among European smoking men, fruit intake was inversely related to lung cancer mortality. This association was particularly present in the Dutch cohort and in the heavy smokers. Vegetable intake was not related to lung cancer in our study. However, in heavy smokers, there was an indication for a decreased risk with higher intakes. Although our results may not be fully comparable because we only examined smokers, they are in line with some but not all cohort studies. Fraser et al.27 and Knekt et al.28 observed an inverse association for fruit but not for vegetables, although the latter study did find an inverse relation for a vegetable subgroup, i.e., root vegetables. Moreover, the nonsignificant associations observed by Chow et al.29 were suggestive for a protective effect of fruit but not of vegetables. In contrast, Steinmetz et al.16 observed a statistically significant inverse relation for vegetable but not for fruit consumption, but she studied women only. Kvåle et al.30 observed associations for both fruit and vegetables. Others only reported associations for fruit,31 for fruit and vegetables combined15 or for specific fruits and vegetables.32 Most case-control studies conducted6-8 did find an inverse association between lung cancer and fruit and/or vegetable intake, among them also an Italian study.33 As discussed in the introduction, results by smoking status are not consistent. Other cohort studies15,16,18 did not report stronger associations in heavier smokers, as we did. But some case-control studies did observe (stronger) associations in heavy smokers.12, 13 Because heavy smokers have presumably the highest exposure to carcinogens and thus are at the highest risk, it may be plausible that the greatest risk reduction can be achieved in this group. On the other hand, it has been shown that in the Finnish cohorts the lighter smokers in 1959 quit smoking during follow-up more often than heavier smokers.34 It is unclear to what extent this phenomenon played a role in the time period we studied and in the other cohorts included. However, we did check whether light smokers consumed more fruits and vegetables at baseline because this may have led to bias. This was not the case (data not shown). The cross-check dietary history we used is assumed to be valid for measuring the habitual intake.35 The reproducibility was investigated for the Zutphen cohort and was satisfactory,36 with Spearman correlation coefficients of 0.65 for fruit and 0.50 for vegetables. However, assessment of fruit and even more of vegetable consumption is generally considered to be difficult.37 A priori, fruit consumption was expected to be highest in the Italian cohorts. Total fruit consumption was, however, in the same range in the 3 countries, although intake of fresh fruits was lower in Finland. We think that the cohorts studied are not representative for the whole country, particularly due to the specific region and type of the cohorts. In the rural cohorts, availability of fruit mainly depended on the production in the own region. While in Zutphen, fruits were mainly available by trade and thus less dependent of season. Because season of data collection differed, this may have also led to less accurate measurements of the actual consumption level. However, we have no reasons to believe that the ranking of subjects, i.e., used in our analyses, is affected. It is possible that we observed an association for fruit and not for vegetable consumption because of a larger variation in fruit intake. Another reason could be the fairly crude way, i.e., in food groups, in which we examined fruit and vegetable intake. The contents of these food groups and preparation methods could have differed between countries. We saw, for instance, that Knekt et al.28 observed an association for a vegetable subgroup but not for all vegetables combined. On the other hand, Voorrips et al.11 found no specific vegetables or fruits responsible for the association observed for total vegetable or fruit intake. The way our dietary information was computerized stopped us from constructing equal subgroups of fruits and vegetables for the 3 countries. Further, by using country-specific tertiles for analyzing the total population, we compared men with low intakes with those with high intakes in the specific countries, ignoring the exact level of consumption. We did this to prevent overrepresentation of countries in 1 of the tertiles. For the country-specific analyses, the number of lung cancer deaths were small, giving uncertain risk estimates with broad confidence intervals. This complicates interpreting differences in results between countries as real differences, for instance, between Finland and The Netherlands. The results from Italy, however, seem quite different. In this country, the absolute lung cancer mortality was much lower compared with the other 2 countries. Perhaps such findings indicate that only higher absolute risks may be lowered by fruit consumption. We coul