Is Dietary Pattern Associated with Gastric Cancer Risk? A Case-control Study in Iran

Objective Diet is considered an important contributor to cancers and one of the best approaches for assessing the combined effect of nutrients and food is the dietary pattern approach. In the present study, the association of dietary patterns identied by factor analysis with gastric cancer risk was studied. Results Four major dietary patterns with a 55.48% prediction rate, namely “tubers and spices”, “cereals and dairies”, “healthy” and “Western-style”, were identied. Tubers and spices (males: 11.42 (4.17, 26.75); females: 6.94 (2.24, 21.56)) and “Western-style” dietary patterns (males: 1.16 (1.00-4.35); females: 2.25 (1.10, 6.49) signicantly increased the odds of gastric cancer risk in both sex. However, “healthy” dietary pattern and “cereals and dairies” dietary pattern were not associated with gastric cancer risk (P > 0.05).


Introduction
Gastric cancer (GC) is one of the most important leading causes of death worldwide and it affects the patient's quality of life and health 1 . Considering the high mortality rate of gastric cancer, primary prevention is an important approach for the improvement of gastric prognosis 2 . Many factors including genetic predisposition, lifestyle, and environmental factors are considered as risk factors of gastric cancer 3 . Diet is also considered as one of the important contributors to cancers 4 and many attempts have been made to identify the important food and nutrients associated with gastric cancer. However, nutrients or foods are not consumed isolated and the etiology of cancers could not be explained only by assessment of a single nutrient or food 5 . One of the approaches for assessment of the combined effect of nutrients and food is the dietary pattern approach 6,7 . Using this approach, the complete picture of food and nutrient interactions and their synergic effects could be achieved and the link between diet and chronic disease could be evaluated 8 .
Previously, the association between dietary pattern and gastric cancer risk has been investigated in some studies and provided mixed results in different populations. In Uruguay, De Stefani et al showed that starchy food pattern was signi cantly associated with gastric cancer risk and the healthy dietary pattern has a negative association with gastric cancer risk 9 . In another case-control study in Canada, it was indicated that Western dietary patterns were increased the odds of stomach cancer, however, healthy dietary patterns were associated with decreased risk 10 . In another study in Japan, the traditional Japanese dietary pattern was associated with an increased risk of gastric cancer, however, the western dietary pattern did not associate with gastric cancer risk 11 . Moreover, the results of a meta-analysis that included the results of studies conducted in European countries, the united states and japan showed about the 2-fold difference in gastric cancer risk between a 'healthy' dietary pattern and Western dietary pattern 12 . Although, previously the association of different food items and gastric cancer risks was studied in the Iranian population 13 , to the best of our knowledge, the relationship between dietary pattern and gastric cancer was not studied. The prevalence of gastric cancer in the northwest of Iran is high and Iran recently faced with nutrition transition namely the adoption of a Western diet, and its combination with traditional dietary patterns provides a particular occasion for further studies in the eld of dietary pattern 14 . So, in the current study, the association of dietary patterns identi ed by factor analysis with gastric cancer risk was studied.

Materials And Methods
In the present hospital-based case-control study, cases and controls were selected from the four major and reference hospitals located in Tabriz

Cases recruitment
The cases were newly diagnosed patients who had histologically con rmed cancer of stomach without other cancers, or any other diseases that affect the dietary pattern, lived in East Azerbaijan province for more than twenty years, and aged between 20-85 years. They were recruited from two major hospitals in Tabriz, Iran. These hospitals are the reference hospitals in cancer in the region. A total of 192 cases were identi ed. The participation fraction was 90.56% for cases. There were no signi cant differences in age, sex, marital status, educational level, and BMI between responders and nonresponses.

Controls recruitment
Two controls were randomly selected for each case from orthopedic and ophthalmic wards of two other hospitals in Tabriz which are the reference hospitals for orthopedic (Shohada hospital) and ophthalmic (Alavi hospitals) disorders in the region. The controls were excluded if they reported the history of cancer or any other diseases that affect dietary patterns such as diabetes or cardiovascular diseases, family history of common cancers; and gastrointestinal disorders. Totally, 365 individuals were selected as a control group. The participation fraction was 90.34% for controls. There was no signi cant differences in age, sex, marital status, educational level and BMI between responders and nonresponses.

Data collection
Covariates for multivariate regression analyses included age, body mass index (BMI), education, marital status, smoking status (over the past year), alcohol consumption (over the past year), self-reported history of H.Pylori infection, number of meals/day, the habit of drinking or eating hot tea and foods. The trained nurses in each hospital gathered information about the dietary and demographic and previous diseases. The demographic information included age, alcohol, and smoking status. For BMI calculation, the body-weight was measured to the nearest 0.1 kg on a Seca digital weighing scale (Dubai, United Arab Emirates), and height was measured to the nearest 0.1 cm, with bare feet using a stadiometer xed to the wall and BMI was calculated by dividing weight (kg) to height 2 (m 2 ). Waist circumference was measured at the minimum circumference between the iliac crest and the rib cage.
For food pattern determination, a 100-item qualitative food frequency questionnaire (FFQ), asking about the consumption of food and beverages over the past year in the control group and over the past year before diagnosis in the case group was used. This questionnaire was completed by face-to-face personal interviews. Each food item was assigned to one of the de ned food groups according to their nutrient content (Table S1).

Statistical analysis
For statistical analyses, SPSS V18 was used. Factor analysis with principal component extraction method and Varimax rotation was used to determine major dietary patterns. Food groups that had commonalities > 0.20 were considered. The derived dietary patterns were labeled considering the food group that had high positive loading and also considering the prior literature. The factor score was calculated and for each dietary pattern, participants received a factor score. Participants were categorized according to tertiles of dietary pattern scores. The between-groups comparisons for continuous and categorical variables were determined using independent sample t-test and chi-square test respectively. The stepwise forward Logistic regression was used to determine the association between dietary patterns and gastric cancer (criteria for entry and retain in the model: p ≤ 0.05 and p ≤ 0.20, respectively). The covariate candidates for inclusion were those statistically signi cant in univariate analyses (p < 0.15) (age, education, marital status, BMI, smoking, and history of H.Pylori infection and hot teat drinking and hot food eating habit).

Results And Discussion
Using the factor analysis method, four major dietary patterns were identi ed ( Table 1). The rst pattern accounted for 19.35% of the variance, loaded positively for tubers, spices, vegetables, salt, oil, eggs and legumes, labeled tubers, and spices. The second pattern (named cereals and dairies), accounted for 13.25% of the total variance, which was characterized by a high intake of cereals and dairy products. Considering the positive load of nut and dry fruits and fruits on the third dietary pattern, this pattern labeled healthy patterns. This pattern accounted for 12.53% of the total variance. Finally, the last extracted pattern accounted for 1.33% of the variance, was characterized by high consumption of high energy drinks, processed foods and snacks, and desserts, and labeled "Western-style" dietary pattern.
Totally, these patterns explained 55.48% of the total variance in dietary patterns. The demographic characteristics of the two series (cases and controls) were presented in Table 2. There were signi cant differences between the two groups in the case of age, sex, BMI, smoking status, and self-reported history of H.Pylori infection. Previous studies showed a positive association between tubers and spices consumption and the risk of gastric cancer 9 . Tubers are rich in starch and nitrite that had been shown as possible risks of gastric cancers. Although there is an inconsistency regarding the association of the nitrite and gastric cancer, a recent meta-analysis study showed that there is a signi cant positive association between nitrite and gastric cancer risk 15 . Spices were the next food group that has signi cantly loaded on this dietary pattern. A recent meta-analysis con rmed the signi cant unfavorable effect of high consumption of spices on gastric cancer 16 . Previous animal studies also showed the carcinogenetic effect of some spices such as chili extract. In addition, human studies showed that high-level consumption of capsaicin-containing foods was associated with an increased risk of cancer 16 . Another food that showed signi cant loading on this dietary pattern was salt. Salt has been considered as an important risk factor of gastric cancer. The effect of high consumption of salt and salty food on gastric cancer risk can be attributed to its direct effect on gastric mucus and its synergic effect with H.Pylori. Epidemiological studies showed a signi cant association between salt consumption and H.Pylori infection rate 17 .
Another food group that showed high loading on this pattern was vegetables. Although some previous studies showed the protective effect of vegetables in gastric cancer, a recent meta-analysis in East Asian countries could not show this protective effect 18 . Moreover, according to previous studies conducted in Iran, the nitrite content of the vegetables consumed in Iran is higher than the recommended amount of WHO 19 . So, this may additionally justify the observed positive association between this dietary pattern and gastric cancer risk.
The result of the present study showed that in the case of "Western-style" dietary pattern, the risk of gastric cancer is signi cantly higher in individuals in the highest tertile of this dietary pattern (males: 1.16 (1.00-4.35); females: 2.25 (1.10, 6.49)). In contrast, a "healthy" dietary pattern and "cereals and dairies" dietary pattern was not signi cantly associated with gastric cancer risk neither in males nor females (P > 0.05). This result is consistent with the results of a recent meta-analysis in this regard. In the present study, this pattern was characterized by high consumption of processed meat, high energy drinks, and desserts. It is postulated that these food groups had an unfavorable effect on gastric cancer through increasing overweight and obesity 20 .
There was not a signi cant association between "cereals and dairies" dietary patterns and gastric cancer risk. In the present study, this dietary pattern was characterized by high consumption of cereals and dairy products. Cereals are rich in starch. Previously it has been shown that starchy foods increase the risk of gastric cancer 21 . On the other hand, in earlier studies, the protective effect of dairy products on gastric cancer risk had been demonstrated 22 . So, the absence of an association between this dietary pattern and gastric cancer risk could be attributed to the counteraction effect of these food groups on gastric cancer.

Limitations
In the present study, we used a factor analysis approach. Different steps used in this method are subjective. Moreover, qualitative FFQ was used in the dietary pattern approach. The limitations of this questionnaire also apply to the dietary pattern approach. Moreover, due to using qualitative FFQ, it could not be possible to control the results for energy intake. Additionally, the hospital-based design of the study could be considered as another limitation of the study. Moreover, we did not match controls individually with cases but in the regression analysis, the results were controlled for a large number of potential confounders such as age, dietary behaviors (the number of meals and hot tea drinking, or hot food eating habit), and anthropometric measurements. Availability of data and materials The datasets supporting the conclusions of this research are included in the article.
Competing interests none