Cancer Causes & Control

, Volume 20, Issue 6, pp 847–853

Dietary fiber and stomach cancer risk: a case–control study from Italy

Authors

    • Dipartimento di EpidemiologiaIstituto di Ricerche Farmacologiche “Mario Negri”
    • Istituto di Statistica Medica e Biometria “G.A. Maccacaro”, Università degli Studi di Milano
  • Lorenza Scotti
    • Unità di Biostatistica e Epidemiologia, Dipartimento di StatisticaUniversità degli Studi di Milano-Bicocca
  • Cristina Bosetti
    • Dipartimento di EpidemiologiaIstituto di Ricerche Farmacologiche “Mario Negri”
  • Paola Bertuccio
    • Dipartimento di EpidemiologiaIstituto di Ricerche Farmacologiche “Mario Negri”
  • Eva Negri
    • Dipartimento di EpidemiologiaIstituto di Ricerche Farmacologiche “Mario Negri”
  • Carlo La Vecchia
    • Dipartimento di EpidemiologiaIstituto di Ricerche Farmacologiche “Mario Negri”
    • Istituto di Statistica Medica e Biometria “G.A. Maccacaro”, Università degli Studi di Milano
Original Paper

DOI: 10.1007/s10552-009-9309-z

Cite this article as:
Bravi, F., Scotti, L., Bosetti, C. et al. Cancer Causes Control (2009) 20: 847. doi:10.1007/s10552-009-9309-z

Abstract

Objective

Fiber intake has been inversely related to stomach cancer risk, although this issue is still controversial.

Methods

A case–control study was conducted in Italy between 1997 and 2007, including 230 cases with incident, histologically confirmed stomach cancer, and 547 controls with acute, non-neoplastic diseases. Dietary habits were investigated through a validated food frequency questionnaire.

Results

Compared with the lowest quintile of intake, the multivariate odds ratios (ORs, including terms for major recognised confounding factors and total energy intake) for the highest quintile were 0.47 (95% confidence interval (CI): 0.28–0.79) for total fiber, 0.50 (95% CI: 0.30–0.85) for soluble non-cellulose polysaccharides (NCP), 0.39 (95% CI: 0.23–0.66) for total insoluble fiber, 0.54 (95% CI: 0.32–0.91) for insoluble NCP, 0.37 (95% CI: 0.22–0.64) for cellulose, and 0.59 (95% CI: 0.36–0.98) for lignin. With reference to the sources of fiber, an inverse association was found for fiber from vegetable (OR = 0.42, 95% CI: 0.24–0.72), and to a lesser extent from fruit (OR = 0.65, 95% CI: 0.38–1.10), but not for fiber from grain (OR = 1.25, 95% CI: 0.77–2.03).

Conclusions

This study found an inverse relationship between stomach cancer risk and various types of fiber, derived, in particular, from vegetables and fruit.

Keywords

Case–control studyDietDietary fiberRisk factorsStomach cancer

Introduction

Gastric cancer risk has been related to various aspects of diet and nutrition [14], and a favorable role of dietary fiber on this neoplasm has been reported in several case–control studies [518]. However, a few other case–control studies [1923] did not find any association. Similarly, a cohort study conducted in the Netherlands [24] and the European Prospective Investigation into Cancer and Nutrition (EPIC) study reported no meaningful relation between total fiber intake and stomach cancer risk [25].

Three studies have further analysed the role of specific types of fiber; of these, two reported inverse relations with gastric cancer risk for cellulose [14], soluble, and insoluble fiber [15], while a prospective study did not find any association with any specific type of fiber, including total, soluble, and insoluble non-starch polysaccharides [24].

Concerning the role of fiber from different sources, two case–control studies found inverse relations, respectively, with fiber from vegetables and fruit [8], and fiber from fruit [14]. A case–control study from Sweden and the EPIC cohort found a protective role of fiber from cereals, though no relation was evident for fiber from vegetables and from fruit [16, 25]. A case–control study from the USA reported inverse relations between fiber from vegetables, fruit and cereals and gastric cardia adenocarcinomas, while observed an inverse relation for fibre from cereals only and distal gastric cancer [18].

In order to add further information on this issue, we have investigated the role of dietary fiber on stomach cancer risk, taking into account different types and sources of fiber, using data from an Italian case–control study.

Materials and methods

A case–control study on stomach cancer was conducted in the greater Milan area between 1997 and 2007. Cases were 230 patients (143 men, 87 women) under age 80 years (median age 63, range 22–80 years) with incident, histologically confirmed stomach cancer (ICD-IX: 151.0–151.9) admitted to major teaching and general hospitals of the study area, with no previous diagnosis of cancer. Most cases were from fundus/corpus (about 50%) or pylorus (about 40%); only ten cases were from cardias. Controls were 547 subjects (286 men, 261 women) under age 80 years (median age 63, range 22–80 years) admitted to the same hospitals as cases for acute, non-neoplastic conditions unrelated to stomach cancer and to long-term dietary modifications. Cases and controls were matched for age and sex, with a case to control ratio of 1:2 for men and of 1:3 for women. Twenty percent of controls were admitted for traumas (mostly fractures and sprains), 23% for other orthopaedic disorders (such as low back pain and disc disorders), 22% for acute surgical conditions and 35% for other miscellaneous illnesses (including eye, nose, ear, skin, or dental disorders). Less than 5% of both cases and controls (4.3% of cases and 3.9% of controls) contacted refused to participate. The main reasons for refusal were poor health status or unavailability at the time of interview.

Trained interviewers interviewed cases and controls during their hospital stay using a structured questionnaire, including information on sociodemographic characteristics, anthropometric measures, lifestyle habits, such as tobacco smoking and alcohol consumption, personal medical history, and family history of cancer in first-degree relatives.

The subjects’ usual diet during the two years before cancer diagnosis or hospitalization (for controls) was assessed through a satisfactorily valid [26] and reproducible [27] food frequency questionnaire (FFQ) including 78 foods and beverages, as well as a range of the most common Italian recipes, grouped into six sections: milk and hot beverages; bread and cereal dishes (first course); meat and other main dishes (second course); vegetables (side dishes); fruit; sweets, desserts, and soft drinks; plus an additional section on alcoholic beverages [28]. Subjects were asked to indicate the average weekly frequency of consumption for each dietary item; intakes lower than once a week, but at least once a month, were coded as 0.5 per week. For a few vegetables and fruit, seasonal consumption, and the corresponding duration, were elicited.

In order to estimate total energy and nutrient intake, an Italian food composition database was used, integrated with other sources when needed [29, 30].

Dietary fiber intake for each food item included in the FFQ was derived using the Englyst procedure [31, 32], which measures fiber as non-starch polysaccharides. A value was obtained for total fiber, soluble, and insoluble fibre intake. A modification of this method was used to obtain values for cellulose separately from insoluble non-cellulose polysaccharides (NCP). Values for lignin, a minor component of human diet, were also provided separately. We did not include resistant starch in the computation of total fiber, because the amount depends on how each food is processed and consumed [33], and related food composition tables were not available. Total fiber intake was also divided according to food types from which it originated (i.e., vegetable, fruit, or grain).

In order to evaluate the effect of fiber independently from total caloric intake, calorie-adjusted fiber intakes were derived using the residual method [34]. The calorie-adjusted fiber was categorized into approximate quintiles on the basis of the control distribution, and the corresponding odds ratios (OR) and 95% confidence intervals (CI) were estimated using unconditional multiple logistic regression models adjusted for age, sex, year of interview, years of education, body mass index (BMI), tobacco smoking, family history of stomach cancer in first-degree relatives, and total energy intake. Continuous ORs were computed using as measurement unit the difference between the upper cutpoint of the fourth and that of the first quintile.

Results

Table 1 shows the distribution of 230 cases of stomach cancer and 547 controls according to sex, age, and selected covariates. Compared with controls, cases were more often smokers, somewhat less educated, and reported more often first-degree relatives with stomach cancer. BMI distribution was similar in cases and controls.
Table 1

Distribution of 230 cases of stomach cancer and 547 controls according to age, education, and other selected variables (Italy, 1997–2007)

 

Cases

Controls

No.

%

No.

%

Sex

    Male

143

62.2

286

52.3

    Female

87

37.8

261

47.7

Age (years)

    <50

39

17.0

97

17.7

    50–59

58

25.2

137

25.1

    60–70

86

37.4

202

36.9

    ≥70

47

20.4

111

20.3

Education (years)a

    <7

95

41.8

237

43.6

    7–11

86

37.9

173

31.9

   ≥12

46

20.3

133

24.5

Body mass index (kg/m2)a

    <20

12

5.3

33

6.0

    20–24

106

46.9

215

39.4

    25–29

82

36.3

223

40.9

    ≥30

26

11.5

74

13.6

Smoking statusa

    Never smokers

96

41.9

261

47.8

    Ex smokers

75

32.8

167

30.6

    Current smokers

    

    <15 cigarettes/day

25

10.9

49

9.0

    ≥15 cigarettes/day

33

14.4

69

12.6

Family history of stomach cancer

    No

200

87.0

516

94.3

    Yes

30

13.0

31

5.7

aThe sum does not add up to the total because of some missing values

Table 2 shows the daily mean intake of total fiber, and various types and sources of fiber, and their Pearson correlation coefficients among controls. Mean intake was 15.8 g for total fiber, 7.7 g for soluble NCP, 7.3 g for total insoluble fiber, 4.0 g for insoluble NCP, 3.3 g for cellulose, 1.4 g for lignin, 5.9 g for vegetable fiber, 5.7 g for fruit fiber, and 4.5 g for grain fiber. Most correlation coefficients were positive and several of them were above 0.5, indicating a substantial collinearity between various types of fiber. Comparatively lower correlations were observed between fibers from various sources.
Table 2

Mean intake (g) and Pearson correlation coefficients between various types/sources of fiber among controls (Italy, 1997–2007)

 

Mean (SD)

Total (Englyst) fiber

Soluble NCP

Total insoluble fiber

Insoluble NCP

Cellulose

Lignin

Vegetable fiber

Fruit fiber

Grain fiber

Total (Englyst) fiber

15.78 (5.20)

1.00000

        

Soluble NCP

7.67 (2.48)

0.97729

1.00000

       

Total insoluble fiber

7.27 (2.88)

0.87674

0.80000

1.00000

      

Insoluble NCP

3.95 (1.33)

0.91761

0.83988

0.86466

1.00000

     

Cellulose

3.33 (1.74)

0.87622

0.79518

0.90592

0.81105

1.00000

    

Lignin

1.42 (0.48)

0.58408

0.58422

0.51805

0.58108

0.49331

1.00000

   

Vegetable fiber

5.89 (2.56)

0.61442

0.57281

0.54352

0.53051

0.55797

0.20502

1.00000

  

Fruit fiber

5.66 (3.45)

0.78409

0.80155

0.67797

0.62072

0.74064

0.44306

0.11987

1.00000

 

Grain fiber

4.51 (1.92)

0.12596

0.07910

0.13182

0.41837

−0.04749

0.40043

−0.05669

−0.20469

1.00000

SD standard deviation; NPC non-cellulose polysaccharides

Table 3 gives the multivariate ORs and the corresponding CIs for intakes of various types of fiber (in quintiles and continuous), after allowance for energy intake and other major confounding factors. The ORs for the highest versus the lowest quintiles of intake were 0.47 (95% CI: 0.28–0.79) for total fiber (p for trend = 0.0005), 0.50 (95% CI: 0.30–0.85) for soluble NCP (p for trend = 0.0057), 0.39 (95% CI: 0.23–0.66) for total insoluble fiber (p for trend < 0.0001), 0.54 (95% CI: 0.32–0.91) for insoluble NCP (p for trend = 0.01), 0.37 (95% CI: 0.22–0.64) for cellulose (p for trend < 0.0001), and 0.59 (95% CI: 0.36–0.98) for lignin (p for trend = 0.11). The continuous OR was 0.63 for total fiber, and varied between 0.58 of total insoluble fiber and 0.78 of lignin. With reference to the sources of fiber, an inverse association was found for fiber from vegetable (OR = 0.42, 95% CI: 0.24–0.72, p for trend = 0.004), and from fruit (OR = 0.65, 95% CI: 0.38–1.10, p for trend = 0.076), but not for fiber from grain (OR = 1.25, 95% CI: 0.77–2.03, p for trend = 0.1742). After mutual adjustment, the OR for fiber from vegetable, fruit, and cereals were 0.46 (95% CI: 0.26–0.81), 0.71 (95% CI: 0.41–1.24), and 1.11 (95% CI: 0.67–1.82), respectively.
Table 3

Odds ratioa (OR) of stomach cancer and corresponding 95% confidence intervals (CI) according to the intake of various types of fiber (Italy, 1997–2007)

Type of fiber (g/day)

Quintile of intakeb

χ2trend (p-value)

Continuous ORd

Q1

Q2

Q3

Q4

Q5

Total (Englyst) fiber

Upper limit

11.37

14.13

16.65

20.10

  

OR (95% CI)

1

0.81 (0.50–1.29)

0.48 (0.29–0.79)

0.48 (0.29–0.80)

0.47 (0.28–0.79)

12.30 (0.0005)

0.63 (0.48–0.84)

Soluble NCP

Upper limit

5.64

6.81

8.10

9.66

  

OR (95% CI)

1

0.72 (0.44–1.16)

0.58 (0.35–0.96)

0.56 (0.34–0.93)

0.50 (0.30–0.85)

7.66 (0.0057)

0.67 (0.51–0.89)

Total insoluble fiber

Upper limit

4.81

6.18

7.65

9.42

  

OR (95% CI)

1

0.60 (0.37–0.97)

0.61 (0.38–0.98)

0.38 (0.22–0.64)

0.39 (0.23–0.66)

15.78 (<0.0001)

0.58 (0.42–0.80)

Insoluble NCP

Upper limit

2.80

3.49

4.19

4.97

  

OR (95% CI)

1

0.70 (0.43–1.15)

0.63 (0.39–1.03)

0.55 (0.33–0.92)

0.54 (0.32–0.91)

6.57 (0.0104)

0.71 (0.55–0.91)

Cellulose

Upper limit

1.89

2.74

3.38

4.66

-

  

OR (95% CI)

1

0.67 (0.42–1.08)

0.44 (0.26–0.72)

0.48 (0.29–0.80)

0.37 (0.22–0.64)

15.44 (<0.0001)

0.62 (0.47–0.82)

Lignin

Upper limit

0.99

1.26

1.52

1.78

  

OR (95% CI)

1

0.49 (0.29–0.82)

0.76 (0.48–1.23)

0.63 (0.39–1.04)

0.59 (0.36–0.98)

2.455 (0.1102)

0.78 (0.61–1.01)

Vegetable fiber

Upper limit

3.82

4.95

6.02

7.91

  

OR (95% CI)

1

0.82 (0.50–1.33)

0.79 (0.49–1.28)

0.73 (0.44–1.19)

0.42 (0.24–0.72)

8.38 (0.0038)

0.72 (0.55–0.95)

Fruit fiber

Upper limit

2.82

4.49

5.92

8.05

  

OR (95% CI)

 

0.90 (0.55–1.47)

0.81 (0.49–1.33)

0.73 (0.43–1.23)

0.65 (0.38–1.10)

3.15 (0.0759)

0.72 (0.55–0.95)

Grain fiber

Upper limit

2.91

3.89

4.66

5.91

  

OR (95% CI)

1

0.65 (0.37–1.12)

1.04 (0.63–1.72)

0.95 (0.56–1.59)

1.25 (0.77–2.03)

1.84 (0.1742)

1.03 (0.84–1.26)

aCalorie-adjusted estimates (according to the residual method) from multiple logistic regression models including terms for age, sex, year of interview, years of education, body mass index, tobacco smoking, family history of stomach cancer, and total energy intake

bQuintiles are computed on the distribution of controls

cReference category

dContinuous OR for an increment equal to the difference between the 80th and 20th percentile

NPC non-cellulose polysaccharides

Table 4 shows the continuous ORs for different fiber according to strata of sex, age (<65 vs. ≥65 years), education (<7 vs. ≥7 years), BMI (<25 vs. ≥25 Kg/m2), and smoking status (never vs. ever smoker). The inverse associations for total fiber, soluble NCP, total insoluble fiber, insoluble NCP, and cellulose were stronger in the younger age group. Risk estimates were consistent across strata of the other covariates.
Table 4

Odds ratio (OR) of stomach cancer and corresponding 95% confidence intervals (CI) according to the intake of various types of fiber, in strata of selected covariates (Italy, 1997–2007)

Type of fiber (g/day)

Continuous ORa (95% CI)

Sex

Age

Education (years)

Body mass index (Kg/m2)

Smoking status

Men

Women

<65

≥65

<7

≥7

<25

≥25

Never

Ever

Total (Englyst) fiber

0.56 (0.38–0.84)

0.73 (0.48–1.12)

0.43b (0.28–0.66)

0.93b (0.62–1.38)

0.52 (0.32–0.87)

0.68 (0.48–0.99)

0.60 (0.40–0.92)

0.68 (0.45–1.04)

0.72 (0.47–1.11)

0.57 (0.39–0.85)

Soluble NCP

0.61 (0.42–0.91)

0.76 (0.49–1.16)

0.45b (0.30–0.69)

1.01b (0.68–1.52)

0.57 (0.35–0.93)

0.72 (0.50–1.04)

0.62 (0.40–0.95)

0.74 (0.49–1.12)

0.73 (0.47–1.13)

0.63 (0.43–0.92)

Total insoluble fiber

0.65 (0.46–0.93)

0.77 (0.52–1.14)

0.45b (0.30–0.67)

1.07b (0.75–1.53)

0.71 (0.45–1.11)

0.69 (0.49–0.97)

0.62 (0.41–0.92)

0.80 (0.56–1.14)

0.76 (0.51–1.12)

0.66 (0.46–0.94)

Insoluble NCP

0.50 (0.32–0.78)

0.71 (0.44–1.13)

0.38b (0.23–0.62)

0.83b (0.54–1.28)

0.49 (0.28–0.86)

0.61 (0.41–0.92)

0.57 (0.36–0.91)

0.60 (0.38–0.95)

0.72 (0.45–1.15)

0.48(0.31–0.76)

Cellulose

0.53 (0.36–0.77)

0.76 (0.51–1.14)

0.44b (0.28–0.67)

0.80b (0.55–1.17)

0.51 (0.31–0.83)

0.66 (0.47–0.94)

0.62 (0.42–0.93)

0.61 (0.41–0.91)

0.74 (0.50–1.10)

0.51 (0.35–0.77)

Lignin

0.80 (0.57–1.11)

0.73 (0.49–1.09)

0.67 (0.46–0.97)

0.90 (0.63–1.27)

0.76 (0.52–1.12)

0.81 (0.56–1.16)

0.61 (0.41–0.89)

0.93 (0.65–1.33)

0.68 (0.46–0.99)

0.86 (0.60–1.22)

Vegetable fiber

0.54 (0.37–0.80)

0.99 (0.69–1.41)

0.62 (0.41–0.94)

0.81 (0.56–1.19)

0.56 (0.34–0.93)

0.79 (0.56–1.12)

0.86 (0.59–1.24)

0.63 (0.41–0.97)

0.97 (0.66–1.43)

0.58 (0.39–0.86)

Fruit fiber

0.78 (0.55–1.10)

0.63 (0.40–1.02)

0.56 (0.38–0.83)

0.97 (0.64–1.47)

0.61 (0.38–0.97)

0.80 (0.56–1.14)

0.63 (0.40–0.97)

0.78 (0.54–1.14)

0.65 (0.42–1.00)

0.74 (0.51–1.07)

Grain fiber

1.13 (0.87–1.46)

0.89 (0.63–1.26)

0.94 (0.70–1.24)

1.16 (0.87–1.54)

1.17 (0.85–1.61)

0.94 (0.71–1.24)

0.86 (0.63–1.18)

1.19 (0.91–1.57)

1.06 (0.76–1.48)

1.04 (0.81–N1.34)

aCalorie-adjusted estimate (according to the residual method) from multiple logistic regression models including terms for age, sex, year of interview, years of education, body mass index, tobacco smoking, family history of stomach cancer, and total energy intake

bp for heterogeneity < 0.05

NPC non-cellulose polysaccharides

Discussion

The results of our study, based on a comprehensive FFQ providing estimate of various types of fiber, further supported an inverse relationship between intake of dietary fiber and stomach cancer risk. The inverse relations were consistent across strata of sex, education, BMI, and smoking status, although they were stronger for subjects younger than 65 years of age. Moreover, a similar inverse relation was found for soluble, insoluble fiber, lignin, and for fiber derived from vegetables and fruit, while no association was found for grain fiber. This may well reflect the positive correlation between various types of fiber, as well as the inverse relation between fruit and vegetables, and stomach cancer risk [4].

Several other case–control studies found an inverse relationship between stomach cancer and total fiber intake [518]. Only a few reported inverse associations for some specific subtypes [14, 15, 24], and sources of fiber [8, 14, 16, 18, 25] as well. Our study is in any case of particular interest, since it was conducted in a southern European population for which little is known on the issue. Moreover, it adds additional information concerning the different types and sources of fiber.

Although the mechanism is unclear, fiber may act by a mechanical cleansing action, removing, or diluting the carcinogens from the epithelial surface. Furthermore, fiber, and particularly water-soluble fibre (deriving mainly from fruit and vegetables), can also delay the absorption of starch, thus reducing the glycemic load [35] (and the consequent hyperinsulinemia), which is related to the risk of stomach cancer in our data (OR = 2.5 for high glycemic load) as well as in other studies [3638]. The lack of protection of grain fiber, found to be protective in other populations [16, 18, 25], may be explained by the fact that it comes mainly from refined cereals, highly consumed in this Italian population. Refined cereals are low in dietary fiber and have been related to an increased risk of stomach cancer [39]. Consequently, the unfavorable action of starch may overwhelm any possible protective effect of fiber. High dietary fiber may simply be an indicator of a diet rich in fruit and vegetables, and of various antioxidant nutrients which have been shown to protect against stomach cancer in our dataset [39, 40], as in other studies [24, 4144]. However, given the high correlation between vegetable, fruit, and fibre, it is difficult to disentangle their separate effect.

A possible limitation of our study is the lack of information on Helicobacterpylori (H. pylori) infection. Though the prevalence of this condition is declining [45], it was relatively high (about 45%) in Italy in the mid-1990s, it increased with age (was over 65% above age 50), and was more frequent in men than in women [46]. Thus, a large proportion of the population studied is likely to be H. Pylori positive. However, case–control studies, unless nested in prospective cohort studies, have limited ability to measure H. pylori infection, because H. pylori infection and the level of circulating anti-H. pylori antibodies may fall after the development of cancer [2, 47].

Among other limitations of the study is the relatively small number of subjects, which did not allow separate analysis by tumor site or histological type. Moreover, dietary habits of hospital controls may differ from those of the general population. However, we carefully excluded all diagnoses that might have involved long-term dietary modifications from the control group. In order to reduce the possibility of bias due to dietary modifications after cancer onset, we elicited patients’ dietary habits in the 2 years before diagnosis. Moreover, in this population there was little or no awareness on a potential relationship between fiber and stomach cancer. The observation that most associations were stronger in subjects under 60 years weighs in favor of the existence of real associations, since information is more accurate and valid in middle age than in the elderly [48].

The almost complete participation of both cases and controls, and the similar interview setting and catchment areas weighs against any relevant role of selection bias in our study. Other strengths of this study are the use of a reproducible and valid FFQ, with an adjusted correlation coefficient of 0.58 for total fiber [26, 27], and the possibility of allowance for several relevant covariates, including total energy intake, in the analysis.

In conclusion, our study adds original information on the role of various subtypes and sources of fiber, and supports the hypothesis that both soluble and insoluble fiber (deriving, in particular, from vegetables and fruit) exerts a significant protection against stomach cancer risk.

Acknowledgment

The authors thank Mrs I Garimoldi for editorial assistance. This work was conducted with the contribution of the Italian Association for Cancer Research, and the Italian League against Cancer. The work of this article was undertaken while Carlo La Vecchia was a Senior Fellow at the International Agency for Research on Cancer. Paola Bertuccio was supported by a fellowship from the Italian Foundation for Cancer Research (FIRC).

Copyright information

© Springer Science+Business Media B.V. 2009