163
Cancer Causes and Control 12: 163±172, 2001.
Ó 2001 Kluwer Academic Publishers. Printed in the Netherlands.
Dietary patterns, nutrient intake and gastric cancer in a high-risk area of Italy Domenico Palli1,*, Antonio Russo1 & Adriano Decarli2 Sezione Epidemiologia Analitica, U.O. Epidemiologia, Centro per lo Studio e la Prevenzione Oncologica (CSPO), Florence, Italy; 2Sezione di Statistica Medica e Biometria, UniversitaÁ degli Studi, Brescia, Italy
1
Received 8 February 2000; accepted in revised form 24 September 2000
Key words: attributable risk, diet, factor analysis (statistical), stomach neoplasms.
Abstract Objectives: To better understand the role of overall dietary patterns and major energy-providing components in gastric cancer etiology. Methods: In a population-based case±control study conducted in a high-risk area in central Italy, 382 gastric cancer cases and 561 controls were available for analysis. Multivariate models based on energy-adjusted residuals and completely partitioned logistic models were used; dietary patterns were evaluated by factor analysis and multiple correspondence analysis. Results: Gastric cancer risk was inversely related to high energy-adjusted intakes of vegetable fat, sugar, betacarotene, vitamin C, alpha-tocopherol, and nitrates. In contrast, signi®cant positive associations emerged with high intake of protein, nitrite, and sodium. According to energy decomposition models, gastric cancer risk increased with increasing intake of protein and decreased with increasing intake of sugar and total fat. The pattern analysis identi®ed four dietary pro®les, overall explaining 75% of total dietary variability. Two patterns, named ``traditional'' and ``vitamin-rich'', were strongly associated with gastric cancer risk and overall accounted for 44% of estimated gastric cancer attributable risk. The other two patterns, ``re®ned'' and ``fat-rich'', were not consistently associated with gastric cancer. Conclusion: Innovative methodological approaches may contribute to better evaluation of the complex relationship between diet and cancer risk and to planning dietary interventions.
Introduction Although gastric cancer incidence and mortality rates showed a consistent decline in recent decades worldwide, it is still one of the most common cancers, particularly in developing countries. The causes of declining rates are not clear but probably re¯ect widespread improvements in lifestyle and dietary habits. Among European countries, Italy has one of the highest death rates for stomach neoplasms, with marked internal variation. Data provided by several cancer registries located across the country have recently con®rmed that gastric cancer incidence and mortality vary widely, with
* Author for correspondence: Domenico Palli, MD, Epidemiology Unit, CSPO, Via di San Salvi 12, 50135 Florence , Italy; Ph.: +39 055 626 36 96; Fax: +39 055 67 99 54; e-mail:
[email protected]
risk highest in the central±northern regions and lowest in southern Italy [1]. Diet has been identi®ed as an important factor in the etiology of gastric cancer. In 1975 Correa et al. [2] postulated that a high consumption of foods rich in salt, nitrite, and preformed nitroso-compounds was associated with an increased risk of gastric cancer. Later, this hypothesis was developed in a more complex pathogenetic model, describing gastric carcinogenesis as a multistep process [3]. In an initial phase, increased pH resulting from widespread chronic atrophic gastritis (caused by high consumption of salt or by chronic infection with Helicobacter pylori) would lead to bacterial invasion, thus increasing the reduction of nitrate into nitrite and the formation of mutagenic and carcinogenic compounds. Protection against gastric cancer may be aorded by frequent consumption of foods rich in antioxidants,
164 including vitamins C and E, and carotenoids and polyphenols as well [4, 5]; these compounds have freeradical scavenging properties and have been shown to inhibit the production of carcinogenic N-nitroso compounds in the gastric microenvironment in humans. Diets high in starch have long been suspected to play an etiologic role. The multistep process of transition from chronic atrophic gastritis to gastric cancer is also in¯uenced by nutritional factors [3]. We present here the updated results from a section of a large multicenter case±control study on diet and gastric cancer, using an expanded series of controls identi®ed in a high-risk area in central Italy at the coordinating center of the original study. Relevant associations between the intake of speci®c nutrients and gastric cancer emerged during the analysis of the large multicenter case±control study and have previously been published [6]. Recently, however, alternative methodological approaches have been proposed to disentangle the eect of each macronutrient from one another and from that of total energy [7±13]. In addition, newly updated Italian food composition tables have been provided [14]. A more detailed analysis of the currently available Florence data was therefore carried out with the aim of better understanding the role of overall dietary patterns and major energy-providing components in gastric cancer etiology. Materials and methods Study subjects The current series of gastric cancer cases were identi®ed in 1985±1987 in a high-risk area in central Italy from one of the participating centers (the coordinating center of the study) in a population-based case±control study carried out in several areas in Italy [6, 15]. All gastric cancer diagnoses were histologically con®rmed [16]. Computerized lists of residents were used to identify a random sample of eligible controls [15]. Overall, 382 gastric cancer cases and 561 controls with complete data were available for the present analysis, including an additional series of 142 controls enrolled at the end of the study period in order to obtain a more representative sample of the general population of the area in the younger age groups and not included in the original paper. Dietary and other interview data A detailed description of the questionnaire has been published elsewhere [6, 15]. Brie¯y, the questionnaire
D. Palli et al. recorded demographic, anthropometric, socioeconomic, residential, occupational, smoking, medical, family, and dietary information. Diet was assessed by asking the usual frequency of consumption of 181 food items and beverages. With the aid of an instruction manual and an atlas with pictures of the most frequently consumed food items, the usual portion size (small, medium, and large) in a 12-month period before the interview was assessed for 146 food and beverage items. A standard portion size was assumed for the remaining items. Amounts of nutrients and energy provided by each food were estimated using newly updated Italian food composition tables [14]. When not available in tables, the amount of dierent food components in complex dishes was de®ned on the basis of traditional Italian recipes. Thermolability was taken into account by reducing estimated contents of ascorbic acid by 50% and beta-carotene by 15% in cooked foods. For all subjects, a cumulative daily average intake for each nutrient was computed by summing values for each food. Use of vitamin supplements was shown to be uncommon during a pilot phase carried out in 1985 and was not considered in the original study questionnaire. The body mass index (BMI), calculated as weight in kilograms divided by height in squared meters, was used as a measure of obesity, and gender-speci®c tertiles were de®ned on the control distribution. Smokers were de®ned as individuals who had smoked at least one cigarette, pipe, or cigar per day for at least one year. Ex-smokers were de®ned as those who had stopped smoking at least one year before study entry. Average daily ethanol consumption was computed by multiplying the amount of ethanol present in each alcoholic beverage reported at interview by the reported frequency of consumption; drinkers were grouped into four categories according to daily consumption (60 114 a b
(13.6) (13.6) (22.8) (29.8) (20.3)
(10.7) (7.3) (29.6) (23.8) (28.5)
Adjusted for age and sex; ** p < 0.01, * p < 0.05. At least a ®rst-degree relative aected with gastric cancer.
was inversely associated with gastric cancer; a modest positive association with alcohol consumption also emerged. No signi®cant association was observed with BMI and cigarette smoking. The mean values for each tertile of daily intake of nutrients, alcohol, minerals, and other compounds among controls, and tertiles cut-o point, are shown in the Appendix. For most nutrients, the ratio between the mean of highest and lowest tertiles was approximately 2, although in a few cases it exceeded 4 (sugar, retinol), reaching a maximum value of an eight-fold variation for alcohol intake.
166
D. Palli et al.
Table 2 shows results of analyses of energy-adjusted nutrient intake. A positive association with increasing intake of total protein (mainly animal protein), sodium, and nitrite was evident. Subjects reporting medium and high intakes of starch also showed a modest increased risk, but no clear trend emerged. Signi®cant inverse associations with gastric cancer risk were shown for vegetable fat, sugar, vitamin C, beta-carotene, vitamin E, and nitrates. Intake of linoleic acid showed a marginally signi®cant negative association with gastric cancer risk. The positive association with alcohol
consumption shown by the preliminary age- and sexadjusted analysis disappeared when taking into account other potential confounders and total energy. A multivariate logistic stepwise procedure selected three nutrients strongly associated with gastric cancer: total protein (p-value for trend 0.0007), alpha-tocopherol (p-value for trend 0.01) and beta-carotene (p-value for trend 0.005). Mean daily caloric intakes provided by each energyproviding nutrient, except alcohol, are shown in Table 3 for control subjects. Mean daily intakes of speci®c nutrients were 80 g of protein, 102 g of sugars, 205 g of
Table 2. Odds ratios (ORa) for gastric cancer risk and corresponding 95% con®dence intervals (95% CI) according to tertile level of intake of nutrients, alcohol, minerals, and other compounds (Florence Gastric Cancer Study) Tertile of intake
p-value for linear trend
2 ORb (95% CI)
3 ORb (95% CI)
Total protein Animal protein Vegetable protein Total fat Animal fat Vegetable fat
1.4 1.0 1.3 1.1 0.9 0.8
(1.0±2.0) (0.7±1.4) (0.9±1.8) (0.8±1.6) (0.7±1.3) (0.6±1.1)
1.7 1.4 1.1 0.9 1.0 0.7
(1.2±2.5) (1.0±2.0) (0.8±1.5) (0.6±1.3) (0.7±1.4) (0.5±1.0)
0.002 0.04 0.7 0.6 0.8 0.04
Fatty acids Total saturated fat Oleic acid Total monounsaturated fat Linoleic acid Linolenic acid Total polyunsaturated fat Cholesterol Carbohydrates Starch Sugar Fiber Alcohol
1.0 0.8 0.9 0.9 1.0 1.0 1.0 1.1 1.6 0.8 1.1 1.1
(0.7±1.4) (0.6±1.2) (0.6±1.2) (0.7±1.3) (0.7±1.4) (0.7±1.4) (0.7±1.3) (0.8±1.5) (1.1±2.3) (0.6±1.2) (0.8±1.5) (0.8±1.6)
0.8 0.9 0.9 0.7 0.9 0.7 1.1 1.0 1.4 0.6 0.8 1.3
(0.6±1.1) (0.6±1.3) (0.6±1.2) (0.5±1.0) (0.6±1.3) (0.5±1.0) (0.8±1.6) (0.7±1.3) (1.0±2.0) (0.4±0.8) (0.5±1.1) (0.9±1.8)
0.2 0.5 0.5 0.04 0.6 0.08 0.6 0.8 0.08 0.002 0.2 0.2
Minerals Iron Calcium Sodium Potassium
1.1 1.0 1.2 0.8
(0.8±1.5) (0.7±1.4) (0.8±1.7) (0.5±1.1)
1.3 0.8 1.7 0.9
(0.9±1.8) (0.6±1.2) (1.2±2.4) (0.6±1.2)
0.2 0.3 0.002 0.4
Vitamins Vitamin C Retinol Beta-carotene Vitamin E
0.8 1.2 0.7 0.7
(0.6±1.1) (0.8±1.6) (0.5±1.0) (0.5±0.9)
0.6 1.3 0.6 0.4
(0.4±0.8) (1.0±1.9) (0.4±0.8) (0.3±0.7)
0.001 0.08 0.0009 0.0001
Nitroso-compounds and precursors Nitrates Nitrite DMA NDMA
0.7 1.4 1.3 1.1
(0.5±1.0) (1.0±2.0) (0.9±1.8) (0.8±1.6)
0.6 1.4 0.9 1.1
(0.4±0.9) (1.0±2.0) (0.7±1.3) (0.8±1.5)
0.01 0.04 0.7 0.7
a Estimates from multiple logistic regression equations including terms for nondietary variables (age, sex, social class, family history of gastric cancer, area of rural residence, BMI tertiles), total energy and tertiles of the residuals of each nutrient of interest. b ORs were estimated for intake at mean level of second and third tertile compared to mean level of ®rst tertile.
Dietary patterns and gastric cancer
167
Table 3. Odds Ratios (ORa) for gastric cancer risk and corresponding 95% con®dence intervals (95% CI) for a 100 kcal/day intake variation of all major macronutrients (except alcohol) (Florence Gastric Cancer Study) Macronutrient
Mean daily energy intakeb (kcal)
ORa (95% CI) for 100 kcal/day A
B
C
D
E 2.2 (1.4±3.5)
Total protein Animal protein Vegetable protein
318 180 138
2.6 (1.8±3.7) ± ±
2.4 (1.6±3.4) ± ±
2.4 (1.6±3.7) ± ±
± 2.7 (1.8±3.9) 0.6 (0.2±2.2)
Total carbohydrates Sugar Starch
1177 381 796
0.9 (0.9±1.0) ± ±
± 0.8 (0.7±0.9) 0.9 (0.8±1.0)
± 0.8 (0.7±0.9) 0.9 (0.8±1.0)
± 0.8 (0.7±0.9) 1.1 (0.9±1.4)
787 373 414 258 425 104
0.8 (0.7±0.9) ± ± ± ± ±
0.9 (0.8±0.98) ± ± ± ± ±
± ± ± 0.9 (0.6±1.4) 0.9 (0.7±1.2) 0.7 (0.3±1.6)
0.89 (0.80±0.98) ± 0.9 (0.8±1.1) ± 0.9 (0.8±0.98) ± ± ±
Total fat Animal fat Vegetable fat Saturated fat Monounsaturated fat Polyunsaturated fat
0.8 (0.7±0.9) 0.9 (0.8±1.0)
a Estimates from multiple logistic regression equations including terms for nondietary variables (age, sex, social class, family history of gastric cancer, area of rural residence, BMI tertiles) and nutrients. b Among 561 population controls. A From a complete partition model including as nutrients total protein, total carbohydrates, and total fat. B From a complete partition model including as nutrients total protein, sugar, starch, and total fat. C From a complete partition model including as nutrients total protein, sugar, starch, saturated, monounsaturated, and polyunsaturated fat. D From a complete partition model including as nutrients animal protein, vegetable protein, sugar, starch, and total fat. E From a complete partition model including as nutrients total protein, sugar, starch, animal and vegetable fat.
starch, and 87 g of total fat (28 g of saturated, 47 g of monounsaturated, and 12 g of polyunsaturated fat). Table 3 also shows OR estimates for a 100 kcal/day equivalent increased intake of each macronutrient derived from a series of complete energy decomposition models adjusted for nondietary variables. Model B shows that gastric cancer risk increased with increasing intake of protein (OR for 100 kcal/day = 2.4, 95% CI 1.6±3.4), and decreased with increasing intake of sugar (OR for 100 kcal/day = 0.8; 95% CI 0.7±0.9) and total fat (OR for 100 kcal/day = 0.9; 95% CI 0.8±0.98). No signi®cant association was found for starch. Models including a complete fat partitioning showed similar results, suggesting a negative association with fat from vegetable sources. Factor analysis identi®ed four nutrient-intake factors that, overall, explained more than 75% of total variability (Table 4-A), accounting for 21%, 19%, 18%, and 17%, respectively. A ®rst dietary pattern, named ``vitamin-rich'', was mainly correlated to all antioxidant vitamins but also to sugar and to a series of components derived from vegetables as nitrates and ®ber. The second pattern (``traditional'') loaded heavily on starch, protein, alcohol (which resulted negatively correlated with the previously described pattern), nitrite, and NDMA. The third pattern (``re®ned'') was associated with nutrients such as cholesterol, saturated fat, protein,
retinol, vitamin E, D, and NDMA. The last pattern, designated ``fat-rich'', loaded heavily on all types of fat and vitamin E. Evaluation of these speci®c nutrient intake patterns showed that a ``traditional'' diet was signi®cantly associated with an increased risk of gastric cancer. In contrast, the ``vitamin-rich'' pattern was associated with a decreased risk (Table 4-B). The two other nutrientintake patterns (``re®ned'' and ``fat-rich'') were not consistently associated with gastric cancer risk. An attributable risk analysis showed that 26% and 39% of cases were due to ``vitamin-rich'' and ``traditional'' patterns, respectively (Table 4-C). The estimated proportion of cases attributable to the combined eect of ``vitamin-rich'' and ``traditional'' nutrient-intake patterns was 44% (95% CI 18.7±69.5). A graphical representation, in a Euclidean space, of the multiple correspondence analysis is shown in Figure 1. Cases were joined with a positive family history for gastric cancer, low social class, rural residence, and low scores for the ``vitamin-rich'' and ``re®ned'' dietary patterns. In contrast, controls aggregated with a negative family history for gastric cancer, urban residence, medium and high social class, and high scores for the ``vitamin-rich'' and ``re®ned'' dietary patterns. The ``traditional'' pattern aggregated with male gender, high consumption of alcohol, and a high score
168
D. Palli et al.
Table 4. Four major dietary patterns identi®ed by factor analysis: (A) factor-loading matrix (factors are ordered according to variance explained); (B) odds ratios (OR) for gastric cancer risk, corresponding 95% con®dence intervals (95% CI) and p-value for linear trend for score tertiles of each factor; (C) Attributable risk (AR) and corresponding 95% con®dence intervals (95% CI) for gastric cancer for each factor (Florence Gastric Cancer Study) (A) Factors
Total protein Saturated fat Oleic acid Monounsaturated fat Linoleic acid Linolenic acid Other polyunsaturated Cholesterol Starch Sugar Fiber Alcohol Vitamin C Retinol Beta-carotene Vitamin E Vitamin D Nitrite Nitrates NDMA Variance explained (%)
Vitamin-rich
Traditional
Re®ned
Fat-rich
0.33 0.18 0.14 0.15 0.18 0.36 0.11 0.12 0.15 0.48 0.76 )0.20 0.82 0.03 0.83 0.53 0.20 0.06 0.80 0.10
0.66 0.16 0.10 0.10 0.27 0.31 0.26 0.26 0.79 )0.12 0.38 0.63 )0.10 0.01 )0.05 )0.02 0.06 0.67 0.13 0.43
0.49 0.56 0.22 0.25 0.30 0.35 0.75 0.72 0.10 0.55 0.18 )0.15 0.18 0.50 0.11 0.40 0.80 0.32 0.03 0.49
21.3
18.6
18.3
17.1
0.32 0.70 0.93 0.93 0.76 0.70 0.30 0.42 0.03 0.26 0.17 0.12 0.09 0.08 0.18 0.64 0.27 0.30 0.16 0.24
(B) Tertile
OR (95% CI)
II III p-value for linear trend
0.6 (0.4±0.9) 0.5 (0.4±0.7) 0.0003
2.2 (1.5±3.4) 3.0 (1.8±4.8) 0.0001
1.1 (0.8±1.5) 1.2 (0.8±1.7) 0.4
0.9 (0.6±1.2) 0.8 (0.5±1.1) 0.2
25.9 (8.8±43.0)
38.9 (21.7±56.1)
10.2 ()7.7±27.9)
8.5 ()10.7±27.8)
(C) AR (95% CI)
Values with a loading of 0.40 or greater are shown in bold typeface.
for the ``fat-rich'' pattern. Females, on the other hand, tended to correlate negatively with alcohol intake, ``traditional'', and ``fat-rich'' patterns. The four dietary patterns are clearly orthogonal to each other, although two of them (``re®ned'' and ``vitamin-rich'') actually appear to be quite close in the Euclidean representation; this is probably due to the fact that variables identi®ed by this analysis have to be imagined in a multidimensional space and may be compressed by the bi-dimensional representation. Figure 1 suggests that ``vitamin-rich'' and ``traditional'' patterns, as identi®ed from factor analysis, in this speci®c population were dierently related to social class. The ``vitamin-rich'' pattern tended to be strongly and positively associated with social class: study subjects classi®ed in the low tertile of ``vitamin-rich'' are very
close to the point representing the low category of ``social class.'' In contrast, the line joining the two points representing the low and high scores for the ``traditional'' dietary pattern is roughly orthogonal to the lines joining low and medium social class, and rural and urban residence, respectively. This might suggest that the ``traditional'' dietary pattern is somehow unrelated to both current social class and residence, and could be more indicative of living conditions and dietary habits in the past. Discussion Our results from energy-adjusted residual models are in agreement with most previous results based on a
Dietary patterns and gastric cancer
169
Fig. 1. Bi-dimensional graphical representation of correspondence analysis. Case/control status was included in the analysis as supplementary variable (shaded label) (Florence Gastric Cancer Study). Legend (variables are in italic, with categories reported in Figure 1 in bold). Factors (Vitamin-rich, Traditional, Re®ned, Fat-rich): 1° = 1st (lowest) tertile, 3° = 3rd (highest) tertile. Points corresponding to the 2nd tertile of the four dietary pattern scores, identi®ed from factor analysis, were excluded from this graphical representation. Age groups: 60 g/die = more than 60 g/day of alcohol intake. Socioeconomic status (SES): Low SES, Medium SES, High SES. Family history for gastric cancer (FH): Neg FH, Pos FH. Residence in the study area: Urban, Rural. Migration from southern Italy: Local, Migrated. Gender: Males, Females. Case/control status: Cases, Controls.
standard approach, in which total caloric intake, on a log scale, was simply included in each model to adjust for total energy intake; positive associations were found for nutrients that are considered to be involved in endogenous nitrosation (nitrite and protein), while decreased risks were found for high intakes of nutrients reported to play an inhibitory role in this process (ascorbic acid and alpha-tocopherol). Our multivariate analyses selected a strong protective nutrient intake pattern rich in antioxidant vitamins (including beta-carotene) and low in protein. In
addition, we identi®ed a strong positive association of gastric cancer risk with sodium intake, and a clear protective eect of sugar emerged. The present analyses included a large series of gastric cancer cases and an expanded series of population controls identi®ed in this high-risk area in central Italy, at the co-ordinating centre of the original multicenter study [6, 15, 16]. Recently, new methodological approaches have been introduced in the analysis of case±control studies on diet and cancer in order to disentangle the eect of each
170
D. Palli et al.
macronutrient from one another and from that of total energy [7±13]. Only Hansson et al. [23] reported results according to various methods of energy adjustment, particularly from a complete partitioning model in a study on gastric cancer. The other papers on gastric cancer reported analyses according to standard [6, 24, 25] or residual models [26±28], or did not apply any adjustment for caloric intake [29±32]. Our results based on complete energy partition models provide additional information and con®rm a direct and strong association with high protein intake and an inverse association with sugar and, possibly, vegetable fat intake. At the same time, the modest positive association with starch intake, observed in a multivariate residual model, disappeared after taking into account other speci®c sources of energy. These results underline the importance, when studying the in¯uence of energy-providing nutrients on cancer risk, of using a model in which energy intake is decomposed into its several sources, with each source simultaneously included in the regression model. Our results providing evidence of a negative association between gastric cancer risk and sugar intake are in contrast with other studies [30, 33, 34], but all our models consistently found this inverse association (OR from a residual model 0.5, 95% CI 0.4±0.8; OR from a complete partition model 0.8, 95% CI 0.7±0.9). A recent study, conducted in Uruguay, reported strong protective associations with glucose and fructose intake, even after controlling for fruit consumption [33]. Factor analysis provided additional details supporting these ®ndings in this high-risk Italian population. Sugar intake was correlated with all protective dietary patterns and, on the other hand, was negatively associated with the ``traditional'' high-risk pattern.
The inverse association with nitrates is probably due to the fact that in western populations nitrates are, overall, a good indicator of total vegetable consumption (as con®rmed by the high correlation with the ``vitaminrich'' dietary pattern score shown in Table 4-A). Dietary studies are hampered by the large number of highly correlated variables, and traditional classi®cation methods may lead to unstable estimates. The multivariate classi®cation method we have used to de®ne dietary patterns may represent an alternative approach to the evaluation of individual nutrients. In our pattern analysis, we identi®ed two quite opposite dietary pro®les, de®ned as ``traditional'' and ``vitamin-rich'', strongly associated with gastric cancer risk, and overall accounting for over 40% of the attributable risk for gastric cancer in this population. Such an approach may be potentially useful for planning dietary interventions or preparing recommendations for the general population. Acknowledgements The authors thank the study participants and collaborators for their cooperation in the original study, and Chiara Zappitello for editorial assistance. We are indebted to Claudio Coppi (ARPAT, Pistoia, Italy) for providing DMA and NDMA concentrations in local foods, and to Giovanna Cordopatri and Marco Ceroti (CSPO, Florence, Italy) for the coordination of dietary interviews, and help with the compilation of the data set for this study. CSPO is supported by Regione Toscana, Florence, Italy. Financial support was provided by a generous grant from the Associazione Italiana per la Ricerca sul Cancro (AIRC, Milan, Italy).
Appendix: Tertile levels of estimated daily intake of nutrients, alcohol, minerals and other compounds and mean values within tertiles among 561 population controls (Florence Gastric Cancer Study) Nutrient
Tertile of intakea I
II
III Q3/Q1
Mean
33%b
Mean
67%c
Mean
Total protein (g) Animal protein (g) Vegetable protein (g) Total fat (g) Animal fat (g) Vegetable fat (g)
60.7 32.2 25.1 67.7 28.2 33.6
70.6 38.9 30.0 79.9 34.9 41.4
77.5 43.6 34.2 90.5 41.4 48.2
85.7 48.3 38.4 100.6 47.5 54.6
100.5 59.1 44.4 177.7 61.2 62.9
1.7 1.8 1.8 2.6 2.2 1.9
Fatty acids Total saturated fat (g) Oleic acid (g)
20.1 32.5
24.3 39.5
27.7 44.4
31.2 49.4
38.2 57.3
1.9 1.8
Dietary patterns and gastric cancer
171
Appendix: (Continued ) Nutrient
Tertile of intakea I
II
III Q3/Q1
Mean
33%b
Mean
67%c
Mean
Total monounsaturated fat (g) Linoleic acid (g) Linolenic acid (g) Total polyunsaturated fat (g) Cholesterol (mg) Carbohydrates (g) Starch (g) Sugar (g) Fiber (g) Alcohol (g) Total calories (kcal) Total calories (kJ)
34.5 6.8 1.0 8.2 209.2 229.9 147.0 63.1 16.7 8.0 2007 8396
41.9 8.1 1.1 9.9 260.0 277.2 180.9 80.5 19.7 36.7 2334 9763
46.9 9.4 1.3 11.1 303.3 310.3 209.1 94.7 22.0 44.2 2582 10803
51.9 10.6 1.4 12.6 353.8 345.1 236.8 113.2 24.4 53.7 2819 11792
60.4 13.0 1.7 15.2 448.8 402.1 280.4 266.1 29.4 69.2 3236 13540
1.8 1.9 1.7 1.9 2.1 1.7 1.9 4.2 1.8 8.7 1.6 1.6
Minerals Iron (mg) Calcium (mg) Sodium (mg) Potassium (mg)
12.6 540.1 2106.7 2518.1
15.0 693.6 2508.8 2997.4
16.8 806.0 2841.5 3242.7
18.6 903.5 3186.5 3554.3
21.6 1090.3 3945.5 4119.4
1.7 2.0 1.9 1.6
Vitamins Vitamin C (mg) Retinol (lg) Beta-carotene (lg) Vitamin E (mg)
56.2 262.2 1804.5 5.5
76.4 364.0 2275.3 6.7
91.0 490.8 2650.1 7.5
109.2 636.9 3036.0 8.3
142.7 1244.8 3910.3 10.0
2.5 4.7 2.2 1.8
Nitroso-compounds and precursors Nitrates (mg) Nitrite (mg) DMA (lg) NDMA (lg) a b c
62.6 2.5 0.22 0.12
79.7 3.1 0.31 0.17
93.2 3.5 0.39 0.20
107.4 4.0 0.47 0.24
132.9 5.4 0.73 0.33
2.1 2.6 3.3 2.8
Absolute estimates of daily intake from FFQ. Cut point between ®rst and second tertile. Cut point between second and third tertile.
References 1. Parkin DM, Whelan SL, Ferlay J, Raymond L, Young J, eds (1997) Cancer Incidence In Five Continents, Vol. 1. IARC Scienti®c Publications No. 143. Lyon: International Agency for Research on Cancer. 2. Correa P, Haenszel W, Cuello C, Tannenbaum S, Archer M (1975) A model for gastric cancer epidemiology. Lancet 2: 58±60. 3. Correa P (1992) Human gastric carcinogenesis: a multistep and multifactorial process ± First American Cancer Society Award Lecture on Cancer Epidemiology and Prevention. Cancer Res 52: 6735±6740. 4. Kono S, Hirohata T (1996) Nutrition and stomach cancer. Cancer Causes Control 7: 41±56. 5. Nomura A (1996) Stomach cancer. In: Schottenfeld D, Fraumeni JF Jr, eds. Cancer Epidemiology and Prevention, 2nd edn. New York: Oxford University Press, pp. 707±724. 6. Buiatti E, Palli D, Decarli A, et al. (1990) A case±control study of gastric cancer and diet in Italy: II. Association with nutrients. Int J Cancer 45: 896±901.
7. Willett WC, Stampfer MJ (1986) Total energy intake implications for epidemiologic analyses. Am J Epidemiol 124: 17± 27. 8. Palmgren J, Kushi LH (1991) Re: Total energy intake: implications for epidemiologic analyses. Am J Epidemiol 133: 1291±1293. 9. Pike MC, Peters RK, Bernstein L (1993) Re: Total energy intake: implications for epidemiologic analyses (letter). Am J Epidemiol 137: 811±813. 10. Wacholder S, Schatzkin A, Freedman LS, Kipnis V, Hartman A, Brown CC (1994) Can energy adjustment separate the eects of energy from those of speci®c macronutrients? Am J Epidemiol 140: 848±855. 11. Witte JS, Pike MC (1995) Re: Can energy adjustment separate the eects of energy from those of speci®c macronutrients? Am J Epidemiol 142: 222. 12. Brown CC, Kipnis V, Freedman LS, Hartman AM, Schatzkin A, Wacholder S (1994) Energy adjustment methods for nutritional epidemiology: the eect of categorization. Am J Epidemiol 139: 323±328.
172 13. Decarli A, Favero A, La Vecchia C, et al. (1997) Macronutrients and breast cancer risk: implications from dierent models. Epidemiology 8: 425±428. 14. Salvini S, Parpinel M, Gnagnarella P, Maisonneuve P, Turrini A, eds (1998) Banca dati di composizione degli alimenti per studi epidemiologici in Italia. Milan: Istituto Europeo di Oncologia. 15. Buiatti E, Palli D, Decarli A, et al. (1989) A case±control study of gastric cancer and diet in Italy. Int J Cancer 44: 611±616. 16. Palli D, Russo A, Saieva C, Salvini S, Amorosi A, Decarli A (2000) Dietary and familial determinants of 10-year survival among patients with gastric carcinoma. Cancer 89: 1205±1213. 17. Breslow NE, Day NE (1980) Statistical Methods in Cancer Research, Vol. 1: The Analysis of Case±Control Studies. IARC Scienti®c Publications No. 32. Lyon: International Agency for Research on Cancer. 18. Willett W (1998) Nutritional Epidemiology. New York: Oxford University Press. 19. Bruzzi P, Green SB, Byar DP, Brinton LA, Schairer C (1985) Estimating the population attributable risk for multiple risk factors using case±control data. Am J Epidemiol 122: 904±914. 20. Benichou J, Gail MH (1990) Variance calculations and con®dence intervals for estimates of the attributable risk based on logistic models. Biometrics 46: 991±1003. 21. Mezzetti M, Ferraroni M, Decarli A, La Vecchia C, Benichou J (1996) Software for attributable risk and con®dence interval estimation in case±control studies. Comput Biomed Res 29: 63±75. 22. Greenacre MJ (1984) Theory and Application of Correspondence Analysis. London: Academic Press. 23. Hansson LE, Nyren O, Bergstrom R, et al. (1994) Nutrients and gastric cancer risk. A population-based case±control study in Sweden. Int J Cancer 57: 638±644.
D. Palli et al. 24. Graham S, Haughey B, Marshall J, et al. (1990) Diet in the epidemiology of gastric cancer. Nutr Cancer 13: 19±34. 25. Risch HA, Jain M, Choi NW, et al. (1985) Dietary factors and the incidence of cancer of the stomach. Am J Epidemiol 122: 947±959. 26. Gonzalez CA, Riboli E, Badosa J, et al. (1994) Nutritional factors and gastric cancer in Spain. Am J Epidemiol 139: 466±473. 27. Ramon JM, Serra-Majem L, Cerdo C, Oromi J (1993) Nutrient intake and gastric cancer risk: a case±control study in Spain. Int J Epidemiol 22: 983±988. 28. CorneÂe J, Pobel D, Riboli E, Guyader M, Hemon B (1995) A case± control study of gastric cancer and nutritional factors in Marseille, France. Eur J Epidemiol 11: 55±65. 29. Correa P, Fontham E, Pickle LW, Chen V, Lin Y, Haenszel W (1985) Dietary determinants of gastric cancer in south Louisiana inhabitants. J Natl Cancer Inst 75: 645±654. 30. La Vecchia C, Negri E, Decarli A, D'Avanzo B, Franceschi S (1987) A case±control study of diet and gastric cancer in northern Italy. Int J Cancer 40: 484±489. 31. You WC, Blot WJ, Chang YS, et al. (1988) Diet and high risk of stomach cancer in Shandong, China. Cancer Res 48: 3518±3523. 32. Boeing H, Frentzel-Beyme R, Berger M, et al. (1991) Case±control study on stomach cancer in Germany. Int J Cancer 47: 858±864. 33. Kaaks R, Tuyns AJ, Haelterman M, Riboli E (1998) Nutrient intake patterns and gastric cancer risk: a case±control study in Belgium. Int J Cancer 78: 415±420. 34. Tuyns AJ, Kaaks R, Haelterman M, Riboli E (1992) Diet and gastric cancer. A case±control study in Belgium. Int J Cancer 51: 1±6. 35. De Stefani E, Boetta P, Deneo-Pellegrini H, Mendilaharsu M, Carzoglio JC, Ronco A (1999) Carbohydrates and risk of stomach cancer in Uruguay (letter). Int J Cancer 82: 618±621.
