- Email: [email protected]
Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based case–control study in a rural population
Clinical Nutrition xxx (2015) 1e7
Contents lists available at ScienceDirect
Clinical Nutrition journal homepage: http://www.elsevier.com/locate/clnu
Original article
Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population Xudong Liu a, Xiaorong Wang a, b, *, Sihao Lin a, b, Xiangqian Lao a, Jin Zhao c, Qingkun Song d, Xuefen Su a, Ignatius Tak-Sun Yu a, b a
JC School of Public Health and Primary Care, The Chinese University of Hong Kong, Hong Kong, China Hong Kong Occupational and Environmental Health Academy, Hong Kong, China Shenzhen Center for Disease Control and Prevention, Shenzhen, China d Beijing Key Laboratory of Cancer Therapeutic Vaccine, Capital Medical University Cancer Center, Beijing Shijitan Hospital, Beijing, China b c
a r t i c l e i n f o
s u m m a r y
Article history: Received 6 August 2015 Accepted 16 November 2015
Background & aims: Few studies were available in exploring the roles of dietary patterns in the development of esophageal cancer, especially in China. This study aimed to investigate the roles of dietary patterns in the risk of esophageal squamous cell carcinoma (ESCC) in a Chinese rural population. Methods: A population-based casesecontrol study was designed and conducted in Yanting County, Sichuan Province of China during two years (between June 2011 and May 2013). A total of 942 pairs of ESCC cases and controls were recruited. A food frequency questionnaire was adopted to collect information of dietary consumption. Dietary patterns were extracted by using principle component and factor analysis based on 24 dietary groups. Odds ratios (ORs) with 95% confidence intervals (95% CI) were calculated by using logistic regression model, with adjustment for possible confounding variables. Results: Four major dietary patterns were identified, which were labeled as “prudent”, “vegetable and fruits”, “processed food” and “alcohol drinking”. In comparison of the highest with the lowest quartiles of pattern scores, the processed food pattern (OR: 2.84, 95% CI: 2.13e3.80) and alcohol drinking pattern (OR: 2.69, 95% CI: 1.95e3.71) were significantly associated with an increased risk of ESCC, while the vegetable and fruit pattern (OR: 0.70, 95% CI: 0.53e0.92) was associated with reduced risk by 30%. The prudent pattern was associated with a reduced risk by 33% (OR: 0.67, 95% CI: 0.50e0.88) in a multivariate logistic regression model, but no statistical significance was reached in a composite model. Conclusions: The results suggest an important role of dietary patterns in ESCC. Diets rich in vegetables and fruits may decrease the risk of ESCC, whereas diets rich in processed food and drinking alcohol may increase the risk. © 2015 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved.
Keywords: Dietary pattern Esophageal squamous cell carcinoma Caseecontrol design Principle component and factor analysis Chinese rural population
1. Introduction Esophageal cancer (EC) is the one of the most common cancers and the most common causes of cancer death worldwide [1]. The overwhelming majority of EC cases and deaths occurred in the developing countries, including China [1]. There were 288,000 cases diagnosed annually in China [2], contributing to about 50% of
* Corresponding author. JC School of Public Health and Primary Care, The Chinese University of Hong Kong, 4/F School of Public Health, Prince of Wales Hospital, Shatin, N.T., Hong Kong, China. Tel.: þ852 22528756; fax: þ852 2606 3500. E-mail addresses: [email protected] (X. Liu), [email protected] (X. Wang), [email protected] (S. Lin), [email protected] (X. Lao), szhaojin@ gmail.com (J. Zhao), [email protected] (Q. Song), [email protected] (X. Su), [email protected] (I. Tak-Sun Yu).
the world's total EC cases [1]. In China, EC was the third most common cancer in rural areas and the sixth in urban areas [3]. It was estimated that more than 90% of diagnosed EC cases in the country were esophageal squamous cell carcinoma (ESCC) [3]. Existing evidence has suggested some dietary components have pivotal roles in carcinogenic process of EC [4,5]. In practice, different foods and nutrients are often mixed together and consumed foods contain literally thousands of nutrients and chemicals. These foods and nutrients likely have interactive effects. Studies on individual nutrients or foods hardly assess synergistic or inhibitory effects. Therefore, observed associations between individual nutrient/food and health outcomes may be attenuated [6]. Dietary pattern, as a holistic and comprehensive approach, may better capture the effects of diets. Dietary patterns can more accurately assess actual dietary exposure, better control dietary
http://dx.doi.org/10.1016/j.clnu.2015.11.009 0261-5614/© 2015 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved.
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
2
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7
confounders, deal with problems of conflicting results, and identify nutrients or foods that may have a relatively small effect. A recent literature review provided some suggestive evidence on the association between dietary patterns and ESCC risk [7]. Furthermore, dietary pattern might be a better method for appropriate intervention [8]. However, no study was available on the association between dietary pattern and ESCC risk in China, where there were high incidence and mortality rates of ESCC. Therefore, we designed and conducted the current study in Yanting County located in Sichuan Province, which is one of the areas with the highest incidence and mortality of ESCC in China [2]. The incidence of ESCC in Yanting was 100.11/100,000 for males and 57.10/100,000 for females according to 2009 data [3], much higher than the rates of Sichuan Province (10.38/100,000 for combined genders) [9], and rates of the country (24.42/100,000 for males, 9.60/100,000 for females) [2], and rural area as a whole (32.96/100,000 for males, 14.21/100,000 for females) [2]. Most of people in Yanting were farmers with low socioeconomic status. The people had a special preference for cereals and tuber-crops, pickled and preserved vegetables, and salted meat [10e15]. Although there have been a great change in eating habits and lifestyle in Chinese population in recent decades, such changes have been relatively small and slow in Yanting area because of historically low economic status and geographically remote [16]. Some lab studies indicated that that Yanting diets could induce esophageal cancer in human cell lines [17,18]. Epidemiological studies conducted in Yanting also suggested possible roles of several individual foods or dietary habits in the development of ESCC [11e15]. The objective of the current study was to explore the associations of dietary patterns with the risk of ESCC. 2. Subjects and methods The caseecontrol study was carried out during two years, from June 2011 to May 2013. A detailed recruitment of cases and controls was described elsewhere [13]. Briefly, ESCC cases were selected from Yanting Tumor Hospital, a main institution to diagnose and treat ESCC in the area [14]. Inclusion criteria for cases were: 1) men or women who were aged 40e70 years, 2) newly diagnosed primary incident ESCC confirmed pathologically (ICD-10, code-C15), and 3) having lived in the Yanting area for at least 15 years. Those who had lived outside of Yanting area for six months or longer in the last 15 years were excluded. Among eligible ESCC cases, there were 942 (96%) cases who accepted an interview, accounting for more than 70% of incident cases during the period of the study. By comparing the recruited cases with non-recruited eligible cases, we did not find significant differences in age and gender. The controls were selected by using a multi-stage sampling method from local residents. Inclusion criteria were: 1) no history of neoplasm at any site; 2) no digestive tract disease; and 3) having lived in the Yanting area for not less than 15 years. Those who had lived outside of the Yanting area for six months or longer in the past 15 years were also excluded. The procedures of sampling were described in detail in previous report [13]. In the end, 942 residents were recruited as individual matched controls. Thus, 942 matched cases and controls (672 males and 270 females) were included in the study, with median age of 60 years. All participating subjects provided written consent following the Declaration of Helsinki. The study was approved by the Joint Chinese University of Hong Kong e New Territories East Cluster Clinical Research Ethics Committee. 2.1. Data collection Questionnaires were administered with a face-to-face interview. Interviewers and study subjects were all unaware of the
hypothesis and objectives of the study. Cases were interviewed at the hospital within a week once diagnosed with ESCC, and controls were interviewed at local health clinics. All questionnaires were checked once more each day after information was collected and any unclear responses were clarified by a contact with related subjects to reduce invalid information/data. A structured questionnaire was used to collect information, including individual demographics (age, gender, education, marital status, occupation, monthly household income and history of residence), family cancer history, tobacco smoking and alcohol drinking history, and self-reported height and weight five years before ESCC diagnosed for the cases and prior to the interviews for the controls. An ever drinker was defined as consuming any alcoholic beverage, which included beer, wine or distilled spirits containing at least 20 g of ethanol, per week for minimum 6 months [19], while a never as one who had never been a regular or social drinker. An ever smoker was defined as smoking at least 10 cigarettes or equal amount of tobacco per week for minimal six months [19], while never smoker as individuals who had never smoked as many as one cigarette a day or equivalent for the duration of one year. The group of ever users could be further categorized into current users and former users. Current users were defined as one who had any of these habits within one year before the interview, while former users as had stopped any of these habits for at least one year before diagnoses or interviews. Body mass index (BMI) was obtained according to the formula: weight (kg)/[height (m)]2. A food frequency questionnaire (FFQ) with 76 items, which was proved to have moderate validity and reproducibility [20], was used to investigate usual dietary intake during the past five years before ESCC diagnosis for the cases or before the interviews for the controls. Three basic questions were asked to collect information on the consumption of each of the items. First, “Did you have ever consumed one certain dietary item” was asked. If the answer was “yes”, “how often the consumption was and how much consumed each time” were further asked. Average daily intake of each food or drink was estimated according to the formula: Average daily intake ¼ Frequency of intake per day Amount of intake each time. The detailed calculation of average daily intake was described in previous reports [13,20].
2.2. Statistical analysis t-Test or chi-square test was used to describe the difference of demographic variables and other factors between the case and controls. The association between possible risk factors and ESCC risk was examined by using logistic regression model (forward stepwise), expressed as odds ratios (OR) with 95% Confidence interval (95% CI). The individual and joint effects of smoking and alcohol drinking were deeply analyzed and the synergy index [21] was adopted to display the interaction of alcohol drinking and tobacco smoking on ESCC risk. However, in this study there were only 11 former drinkers and 14 former smokers in the cases and 0 former drinkers and 0 former smokers in the controls. Hence, it is not meaningful to divided ever smokers or drinkers into current and former subgroups. The 76 dietary items were further grouped into 24 dietary groups, according to local eating habits, dietary guideline and balance diet pagoda for Chinese population [22] and whether or not the food was processed, so to derive dietary patterns. The 24 groups were used in all dietary assessment methods after similarities were tested, which ensured the consistency of groups entering into the factor analysis. The original absolute food intake was standardized before factor analysis was performed to avoid problems arisen from different measurement units and different intake amount.
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7
Principle component and factor analysis (PCFA) with correlation matrix of 24 dietary groups were used to derive dietary patterns. Orthogonal (varimax) transformation was adopted to simplify structure and make results interpretable. Eigenvalues (>1.2), the scree plot construction and the total variance and interpretability of the factors were considered for determining the number of factors [23]. The Bartlett's test of sphericity and the KaisereMeyereOlkin measure of sampling (KMO) were used to determine the relationships between variables of interest in the factor analysis [24]. The extracted factors were labeled according to the interpretation of the data [25]. Foods with absolute rotated factor loadings of more than 0.4 were referred to as ‘dominant foods’ hereafter. The dietary pattern scores emerging from data source were calculated by using the regression method. Furthermore, complementary analysis was performed to assess the robustness of the dietary patterns identified. First, the factor scores were calculated with multiple regression method that may standardize results [23]. Then, PCFA was carried out in men and women, respectively. In order to evaluate the internal reproducibility of identified dietary patterns, subjects were randomly divided into two equally-sized subgroups. Then, performance of PCAF was repeated in each subgroup. In light of the identified patterns being robust, factor scores were generated from all subjects and been used in all subsequent analyses. In each of the patterns, subjects were grouped into four categories based on the quartiles of pattern scores among the controls. Crude and adjusted OR and 95%(CI) were estimated from univariate and multivariate logistic regression (forward stepwise) models, respectively [26]. Then, a stratified analysis by gender was performed. In addition, a stratified analysis by smoking status among males was carried out, as few women smoked or drank in the subjects and the study area. The Wald statistics was calculated to determine an exposureeresponse trend, by putting the median scores of each quartile as a continuous variable into univariate and multivariate logistic regression. Potential confounding variables were adjusted in the multivariate model, including age, gender, educational status, tobacco smoking, BMI, family cancer history, and total energy (calories, kJ) intake. Total energy intake for each subject was calculated based on the Chinese food composition tables [27]. All statistical analyses were performed with IBM SPSS (version 19.0, 2010, IBM SPSS Inc.). The tests were two-tailed with p < 0.05 being considered to be statistically significant. 3. Results Table 1 displays basic information of cases and controls. Mean calorie intake was 2210.2 kcal in the cases and 2445.1 kcal in the controls; BMI was 21.6 kg/m2 in the cases and 22.7 kg/m2 in the controls. Smoking and alcohol drinking, as well as family history of cancer were more common in the cases than in the controls. In the controls, a mean value of average daily consumption of each dietary group increased from the lowest to highest quartiles of pattern scores when the positive factor loading increased (Table A in S1 File). On the contrary, a mean value of average daily consumption of each dietary group decreased with a decrease of negative factor loadings. No significant difference in the mean value of average daily consumption of each dietary group was observed across quartiles of dietary pattern scores when absolute factor loading was small. To give more information on alcohol drinking and smoking related risk in the subjects, Table B and Table C in S1 File show the association between alcohol drinking/smoking and the risk of ESCC. Drinking frequency, drinking amount of liquor each time and average ethanol consumption per week were all significantly associated with a higher risk, showing an exposureeresponse trend (p for trend < 0.001) (Table B in S1 File). Compared to never drinkers, liquor drinkers had a double risk (OR: 2.07, 95% CI:
3
Table 1 Characteristics of study subjects. Cases (N ¼ 942) Age, mean (SD), years Gender, males, n (%) Education status, m (%) Lower than primary school Primary school Middle school and above Family cancer history, n (%) No Yes BMI Mean (SD), kg/m2 N (%) <18.5 kg/m2 18.5e<23 kg/m2 23e<25 kg/m2 25 kg/m2 Total calorie intake, kcal/day, mean (SD) Smoking status Mean (SD), pack-years Male, n (%) Never smokersc Ever smokers Female, n (%) Never smokers Ever smokers Drinking status Ethanol intake, mean (SD), g/week Male, n (%) Never drinkersd Ever drinkers Female, n (%) Never drinkers Ever drinkers
Controls (N ¼ 942)
p Value
60.1 (6.8) 672 (71.3)
60.2 (6.8) 672 (71.3)
0.592a /
198 (21.1) 542 (57.5) 202 (21.4)
209 (22.2) 511 (54.2) 222 (23.6)
0.341b
699 (74.2) 243 (25.8)
755 (80.1) 187 (19.9)
21.6 (2.9)
22.7 (2.5)
0.002b
118 540 174 110 2212.2
(12.5) (57.3) (18.5) (11.7) (771.0)
46 470 240 186 2453.1
(4.9) (49.9) (25.5) (19.7) (812.2)
41.5 (20.6)
36.8 (18.3)
102 (15.2) 570 (84.8)
306 (45.5) 366 (54.5)
258 (95.6) 12 (4.4)
263 (97.4) 7 (2.6)
304.9 (471.6)
98.5 (16.8)
100 (14.9) 572 (85.1)
271 (40.3) 401 (59.7)
233 (86.3) 37 (13.7)
239 (88.5) 31 (11.5)
<0.001a <0.001b
<0.001a
<0.001a <0.001b
0.243b
0.001a <0.001b
0.436b
a
p Value for paired t-test. p Value for Chi-square test. c Ever smokers included current smoker and former smoker. There were 14 former smokers in the cases and 0 former smokers in the controls. d Ever drinkers included current drinkers and ever drinkers. There were only 11 former drinkers in the cases and 0 former drinkers in the controls. b
1.57e2.73); consumption of ethanol more than 443.8 g per week was associated with over 7-fold risk (OR: 7.16, 95% CI: 4.61e11.13). However, drinking beer was not significantly associated with risk. Smoking, no matter expressed as the frequency, duration or cumulative smoking amount, was significantly associated with an increased risk of ESCC, with clear exposureeresponse trends (p for trend < 0.001) (Table C in S1 File). Compared to never smokers, smokers had nearly 2.8-fold higher risk (OR: 2.81, 95% CI: 2.12e3.73) after potential confounding factors were adjusted. In the analysis of their joint effect (Table D in S1 File), the greatest risk was observed in those who both smoked and drank (OR: 5.54, 95% CI: 4.04e7.67). The synergy index for smoking and alcohol drinking was 4.10 (95% CI: 1.80e9.35). Overall value of KMO was 0.738 (>0.5 is regarded as acceptable [24]), and p for Bartlett's test of sphericity was less than 0.001 (<0.001 as acceptable [24]), suggesting a proper sample size to perform PCFA. The values of KMO for 24 dietary groups were all greater than 0.55 (Table E in S1 File). Four factors were identified through the PCFA (Table 2), which explained 41% of the total variance. The first defined factor, which was characterized by a higher factor loading of rice, wheat, tofu and dry bean-curd, dry-beans, nuts and seeds, red meat, and fresh eggs, was labeled as ‘prudent pattern’. It mainly represented the diet in the rural area and explained 13.7% of the total variance. The second factor was labeled as ‘vegetable and fruit pattern’, which had a higher factor loading of
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
4
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7
Table 2 Factor loadings for the relationship between food groups and factors representing dietary patterns.a Diet items
Rice Wheat Corn Tuber-crops Cruciferous vegetables Deep-green vegetables Deep-yellow vegetables Citrus fruits Non-citrus fruits Tofu and dry bean-curd Dry-beans Nuts and seeds Red meat White meat Fresh eggs Preserved vegetables Pickled vegetables Salted meat Salted eggs Animal fat Plant oil Tea Liquor Beer Variance explained VAR (%) Cumulative explained VAR (%)
Prudent pattern 0.42 0.45
0.25 0.24 0.54 0.62 0.52 0.41 0.32 0.47 0.14 0.23 0.15 0.26 0.33 0.13
Vegetable and fruits pattern
0.12 0.12 0.13 0.80 0.79 0.51 0.64 0.68 0.11 0.13 0.13
0.14 0.21 0.08 0.28
Processed food pattern
Alcohol drinking pattern
0.12
0.18
0.10
0.22 0.21 0.29 0.24 0.26
0.12 0.11
0.20 0.68 0.46 0.64 0.41
0.11
0.12 0.11 0.25 0.15 0.11
0.18 0.11 0.22 0.36 0.72 0.58
13.77
9.56
9.09
8.64
13.77
32.42
32.42
41.06
a Principle component and factor analysis was performed based on 24 dietary items. With orthogonal rotation, the factor loading scores are identical to the correlation coefficients. The magnitude of each loading indicates the importance of the corresponding items to the factor. Loadings 0.40 were shown in bold typeface and loadings < 0.10 were suppressed for clarity.
cruciferous vegetables, dark-green vegetables, dark-yellow vegetables, citrus fruits and non-citrus fruits explaining 9.6% of the total variance. The third factor was labeled as ‘processed food pattern’, which had high loadings of pickled vegetables, preserved vegetables, salted meat and salted eggs, explaining 9.1% of the total variance. The fourth factor, as ‘alcohol drinking pattern’, had a high loading of beer and liquor, which explained 8.6% of the total variance. In the multivariate analysis, the processed food pattern and alcohol drinking pattern were associated with an increased risk of ESCC (Table 3). The highest quartile of pattern scores was related to a nearly 3-fold risk for the processed food pattern (OR: 2.84, 95% CI: 2.13e3.80, p for trend < 0.001) and the alcohol drinking pattern (OR: 2.69, 95% CI: 1.95e3.71, p for trend < 0.001), in comparison with the lowest quartiles of the scores. The highest quartile of the prudent pattern scores was related to a lower risk by 33% (OR: 0.67, 95% CI: 0.53e0.92). Similarly, the highest quartile of the vegetable and fruits pattern scores was related to 30% lower risk (OR: 0.70, 95% CI: 0.53e0.92, p for trend < 0.005). When all the four patterns and major confounding variables were entered simultaneously into a same model (Table F in S1 File), the positive or inverse associations remained, with the exception of the prudent pattern, of which association with the risk did not reach a statistical significance in the composite model. In stratified analysis by gender (Table 4), both processed food pattern and alcohol drinking pattern were associated with an increased risk in males. In the comparison of the highest with the lowest quartile of scores, OR was 2.58 (95% CI: 1.83e3.64) for the
Table 3 Associations between dietary patterns and ESCC risk. Dietary pattern
Case/control
Prudent Quartile 1 294/235 Quartile 2 242/236 Quartile 3 203/235 Quartile 4 203/236 p for trend Vegetable and fruits Quartile 1 317/235 Quartile 2 221/236 Quartile 3 205/235 Quartile 4 199/236 p for trend Processed food Quartile 1 139/235 Quartile 2 187/235 Quartile 3 200/236 Quartile 4 415/236 p for trend Alcohol drinking Quartile 1 167/235 Quartile 2 166/236 Quartile 3 162/235 Quartile 4 446/236 p for trend
Crude OR (95% CI)
Adjusted ORa (95% CI)
1.00 0.82 (0.64, 1.05) 0.69 (0.54, 0.89) 0.68 (0.53, 0.89) 0.002
1.00 0.80 (0.61, 1.05) 0.70 (0.53, 0.93) 0.67 (0.50, 0.88) 0.040
1.00 0.69 (0.54, 0.89) 0.64 (0.50, 0.83) 0.63 (0.49, 0.81) <0.001
1.00 0.80 (0.61, 1.05) 0.74 (0.56, 0.98) 0.70 (0.53, 0.92) 0.005
1.00 1.35 (1.01, 1.79) 1.43 (1.08, 1.89) 2.94 (2.26, 3.83) <0.001
1.00 1.25 (0.92, 1.71) 1.38 (1.01, 1.88) 2.84 (2.13, 3.80) <0.001
1.00 0.99 (0.75, 1.32) 0.97 (0.73, 1.28) 2.67 (2.07, 3.44) <0.001
1.00 1.06 (0.78, 1.45) 1.05 (0.76, 1.46) 2.69 (1.95, 3.71) <0.001
a OR was adjusted for age, gender, educational status (lower than primary school, primary school, higher than primary school), BMI (continuous, kg/m2), smoking status (no vs. yes), family history of cancer (no vs. yes), and total calorie intake (kcal, log 10-transformation).
processed food pattern, and 2.99 (95% CI: 2.07e4.31) for the alcohol drinking pattern. The highest scores of vegetable and fruits pattern were associated with a reduced risk by 46% (OR: 0.54, 95% CI: 0.38e0.78). Similar results for the processed food pattern and vegetable fruit pattern were obtained in females. No significant association with the prudent pattern was observed in either gender. In the stratified analysis by smoking status in males (Table 5), both processed food pattern and the alcohol drinking pattern were significantly associated with an increased risk, and the vegetable and fruit pattern was associated with a reduced risk in either smokers or non-smokers, in the comparison of the highest with the lowest quartiles of pattern scores. Likewise, the prudent pattern was not associated with ESCC risk in either subgroup. 4. Discussion In this study, four main dietary patterns were identified, i.e. the prudent, vegetable and fruit, processed food and alcohol drinking patterns. The processed food pattern and the alcohol drinking pattern were found to be associated with an increased risk of ESCC, whereas vegetable and fruit pattern was associated with a decreased risk. There was no significant association with the prudent pattern observed in the study subjects, however. Some components in the dietary patterns of this study were similar with those of the patterns defined in some other studies, although not all components were identical. For instance, the processed food pattern had higher factor loadings in four types of the processed foods, such as processed meat, pickled vegetables, preserved vegetables and salted eggs, which were similar to the western pattern labeled in studies from Uruguay (containing barbecued meat) [28], Iran (containing pickle and processed meat) [29], and Sweden (containing processed meat) [30]. These types of processed food were thought to be risk factors for digestive cancers [14,31e33]. Most of the components of the prudent pattern in the present study were similar to those of the prudent pattern reported in Uruguay [28] and the healthy pattern in Iran and Sweden [29,30].
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7 Table 4 Associations between dietary patterns and ESCC risk by gender. Dietary patterns
Males Cases/ controls
Prudent Quartile 1 201/168 Quartile 2 167/168 Quartile 3 154/168 Quartile 4 153/168 p for trend Vegetable and fruits Quartile 1 175/168 Quartile 2 218/168 Quartile 3 174/168 Quartile 4 105/168 p for trend Processed food Quartile 1 103/168 Quartile 2 137/168 Quartile 3 147/168 Quartile 4 285/168 p for trend Alcohol drinking Quartile 1 85/168 Quartile 2 81/168 Quartile 3 128/168 Quartile 4 378/168 p for trend
Females ORa (95% CI)
Cases/ controls
ORa (95% CI)
1.00 0.83 (0.60, 1.16) 0.82 (0.58, 1.15) 0.78 (0.56, 1.08) 0.196
87/68 73/67 57/67 53/68
1.00 0.73 (0.45, 1.19) 0.63 (0.38, 1.04) 0.60 (0.36, 1.01) 0.041
1.00 1.28 (0.92, 1.77) 0.91 (0.65, 1.27) 0.54 (0.38, 0.78) <0.001
92/68 62/67 49/67 62/68
1.00 0.79 (0.48, 1.29) 0.93 (0.56, 1.53) 0.61 (0.36, 0.84) 0.639
1.00 1.35 (0.94, 1.96) 1.33 (0.92, 1.91) 2.58 (1.83, 3.64) <0.001
34/68 54/67 54/67 128/68
1.00 1.41 (0.80, 2.49) 1.49 (0.85, 2.63) 3.16 (1.87, 5.35) <0.001
1.00 0.77 (0.51, 1.16) 1.10 (0.75, 1.62) 2.99 (2.07, 4.31) <0.001
68/68 51/67 70/67 81/68
1.00 0.65 (0.38, 1.10) 0.99 (0.59, 1.64) 1.27 (0.77, 2.09) 0.189
a OR was adjusted for age, educational status (lower than primary school, primary school, higher than primary school), BMI (continuous, kg/m2), family history of cancer (yes vs. no), and total calorie intake (kcal, log 10-transformation).
Salted meat, which was found to increase the risk of ESCC [34], was dominant components of the processed food pattern. On the other hand, red meat and fresh eggs, which were two main components of the prudent in this study, were labeled as the western pattern in other studies [28e30]. It was reported that red meat and processed
Table 5 Associations between dietary patterns and ESCC risk by smoking status in males. Dietary patterns
Non-smoker Cases/ controls
Prudent Quartile 1 24/77 Quartile 2 22/76 Quartile 3 22/76 Quartile 4 34/77 p for trend Vegetable and fruits Quartile 1 31/77 Quartile 2 37/76 Quartile 3 19/76 Quartile 4 15/77 p for trend Processed food Quartile 1 19/77 Quartile 2 25/76 Quartile 3 20/76 Quartile 4 38/77 p for trend Alcohol drinking Quartile 1 18/77 Quartile 2 23/76 Quartile 3 17/76 Quartile 4 44/77 p for trend
Smoker a
OR (95% CI)
Cases/ controls
ORa (95% CI)
1.00 0.71 (0.35, 1.42) 0.73 (0.37, 1.46) 1.18 (0.62, 2.23) 0.537
182/92 143/91 126/81 119/92
1.00 0.88 (0.60, 1.30) 0.78 (0.53, 1.14) 0.68 (0.46, 1.01) 0.041
1.00 1.20 (0.66, 2.17) 0.66 (0.33, 1.29) 0.49 (0.24, 0.97) 0.019
153/92 173/91 156/91 88/92
1.00 1.16 (0.79, 1.69) 1.03 (0.70, 1.52) 0.53 (0.35, 0.80) 0.005
1.00 1.15 (0.56, 2.33) 1.08 (0.52, 2.23) 1.87 (1.05, 3.63) 0.065
92/92 106/91 130/91 242/92
1.00 1.21 (0.79, 1.86) 1.41 (0.93, 2.15) 2.73 (1.83, 4.05) <0.001
1.00 1.36 (0.66, 2.8) 1.09 (0.51, 2.33) 2.65 (1.38, 5.12) 0.004
86/92 89/91 103/91 292/92
1.00 1.10 (0.71, 1.70) 1.32 (0.86, 2.02) 3.97 (2.66, 5.92) <0.001
a OR was adjusted for age, gender, educational status (lower than primary school, primary school, higher than primary school), BMI (continuous, kg/m2), family cancer history (yes vs. no), and total calorie intake (kcal, log 10-transformation).
5
meat increased the risk of ESCC, especially when meat is barbecued, fried and rotated at higher temperature [35e37]. These components were classified into the prudent pattern, because many traditional foods were prepared with red meat and eggs in Chinese population [38,39]. Moreover, average intake of red meat (32 g/day) and salted meat (30 g/day) were much higher than intake of poultry (3 g/day) and seafood (1 g/day) in the study area [10]. Red meat and fresh eggs were two major food sources that provided protein, energy, some minerals and other nutrients in the rural area with a lower economic status in China. The components of the vegetable and fruit pattern were similar to same patterns reported from Australia [40] and the U.S. [41], and the healthy pattern from Uruguay [42]. The alcohol pattern in the study was similar to the drinker pattern reported from Uruguay men [28,42], the smoking/ alcohol pattern from the U.S. [41] and the alcohol drinker pattern from Sweden [30]. The components of the dietary patterns, however, were somewhat different in studies conducted in different countries, possibly due to differences in the dietary assessment methods, food availability, and ethnical and cultural differences [43]. We observed that the vegetable and fruit pattern was associated with a reduced risk by nearly 30%. Similar results were also obtained in the stratified analyses. The result was consistent with some other studies [40e42]. The World Cancer Research Fund report in 2007 [5] and two updated systematic review and metaanalysis [44,45] state that there is strong evidence to support that fruits and vegetables have a protecting effect against ESCC. This study provided further supportive data for that. Smoking and alcohol drinking were significantly associated with an increased risk of ESCC in this study, which was consistent with existing evidence [5]. However, we did not find an evident association of drinking beer with ESCC, possibly because of a very small number of the subjects who drank beer. Alcohol drinking is a well-established risk factor for ESCC [5,46,47]. This study observed an increased risk associated with the alcohol drinking pattern, which was consistent with the results from systematic review and meta-analysis [7]. In the stratified analysis by gender, a significant association with the alcohol drinking pattern was only observed in males, but not in females. An explanation was a small number of alcohol drinkers in the females (13.7% in the cases and 11.5% in the controls). Some studies showed that processed food, including preserved vegetables, pickled vegetables, and salted meat, was associated with an increased risk [13,14,48]. Pickled vegetables, preserved vegetables and salted meat were commonly consumed by Yanting residents. This study integrated the different kinds of processed food into the processed food pattern. The pattern was associated with 2.6-fold risk in males (OR: 2.58, 95% CI: 1.83e3.64) and 3.2fold risk in females (OR: 3.16, 95% CI: 1.87e5.35). Similar result related to a labeled western pattern was reported from an Iran study [29]. Processed foods are sources of nitrites and N-nitroso compounds [49,50], which could exert an effect on carcinogenesis [51] and contribute to the development of EC [52]. In this study, the prudent pattern was observed to be related to a 33% decreased risk of ESCC, when the highest and the lowest quartiles of pattern scores was compared. This was consistent with a study conducted in Uruguay, in which the prudent pattern was associated with a decreased risk by 50% [28]. However, there was no significant association with the prudent pattern in a composite model that included all interested patterns and confounders. Therefore, the role of prudent pattern in ESCC needs to further study. Factor analysis is a major posteriori method in identifying dietary patterns in studying the dietedisease relationship [53]. Factor analysis is a data-driven method and its reliability is largely
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
6
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7
dependent on sample size [24]. This study included 942 pairs of the cases and controls, which met the suggested criteria (>300) [24]. Although sample sizes might have reduced in stratified analysis, overall and individual values of KMO were all higher than 0.55, suggesting that factor analysis was acceptable in this study [24]. A limitation of the factor analysis may be a subjective decision in defining dietary patterns, including food/food groups selected, number of dominant factors to be retained and the interpretation of factors. However, the reproducibility and validity study of the FFQ [20] suggested PCFA was stable and robust in identifying dietary patterns. There was a significant difference in BMI and calorie intake between cases and controls. Due to the retrospective nature of this study, recall bias cannot be avoided. However, the food frequency questionnaire used in this study was proved to be reproducible and valid in exploring the dietary intake [20]. BMI, which was suggested to have critical effect on the occurrence of ESCC, was related to the total caloric intake [54]. High BMI might be due to the high level of calorie intake. Patients who suffered from tumors, especially the digestive tumors, might have an abnormal diet intake and then lose their weight during the progress of the disease [55]. However, in this study, the subjects were required to report the anthropological and diet information five years before being diagnosed as ESCC for the cases and prior to the interviews for the controls, which made sure that the influence on BMI by the disease progression was small. Moreover, these two variables as well as other confounders were adjusted via regression model to reduce the possible confounding effects. This study is among the few, if any, to assess the associations of dietary patterns and the risk of ESCC in China, where dietary patterns were largely different from those of the western countries. The strengths of the study included a matched caseecontrol design, high response rates and relatively large numbers of cases and controls. There were no differences in major socialdemographic characteristics between the recruited and nonrecruited cases and controls. All cases were histological confirmed incidence cases, thus largely reduced disease misclassification. In addition, detailed information on potential confounding factors was collected and adjusted in the analyses. However, because of a retrospective nature of the caseecontrol design, information bias, in particular recall bias, could be a major concern. However, we made an effort to minimize recall bias by blinding the cases, controls, and interviewers about the study hypothesis and objectives. In addition, the cases were interviewed within one week once diagnosed as ESCC, so to minimize the likelihood of changing a dietary habit following the diagnosis and treatment of the disease. In addition, Standard training was provided to the interviewers to handle cases and controls in the same way. The FFQ used in our study is reasonably reproducible and valid to collect information on the consumption of common foods and drinks [20]. Another limitation is that there was a small number of females and in some subgroups (e.g. male nonsmokers), which reduced the statistical power to detect a significant association in specific subgroups. In conclusion, this study identified four major dietary patterns in the subjects of the study. The results suggested that the processed food pattern and alcohol drinking pattern were associated with an increased risk of ESCC, whereas the vegetable and fruit pattern reduced the risk. The study provided further support to the current recommendation on decreasing the consumption of processed food and alcohol drinking and increasing the vegetable and fruit intake for EC prevention. Further prospective studies are needed to verify the roles of the dietary patterns in the development of ESCC in Chinese population.
Conflict of interest The authors state that they have no conflicts of interest. Funding source and its roles The study was supported by the World Cancer Research Fund, International (No. 2010/240). The founder had no role in the design, analysis or writing of this manuscript. Statement of authorship XW conceived and designed the study; XDL, SL and QS collected the data, XDL analyzed the data and drafted the manuscript under the supervision of XW; XW, SL, JZ, QS, XQL, XS, and IY reviewed and edited the manuscript. All co-authors provided comments and approved the final version. Acknowledgments The authors would like to thank nurses and doctors in the Yanting Tumor Hospital for their cooperation in data collection; thank all study subjects for their participation. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.clnu.2015.11.009. References [1] Stewart BW, Wild C. World cancer report 2014. Lyon, France: International Agency for Research on Cancer; 2014. [2] Chen WQ, Zheng RS, Zhang SW, Zeng HM, Fan YG, Qiao YL, et al. Esophageal cancer incidence and mortality in China. Thorac Cancer 2010;2014(5):343e8. [3] Chen WQ, Hao J. Chinese cancer registry annual report, 2012 (in Chinese). Beijing: Military Medical Science Press; 2012. [4] Reszka E, Wasowicz W, Gromadzinska J. Genetic polymorphism of xenobiotic metabolising enzymes, diet and cancer susceptibility. Br J Nutr 2006;96: 609e19. [5] American Institute for Cancer Research, World Cancer Research Fund. Food, nutrition, physical activity and the prevention of cancer: a global perspective: a project of World Cancer Research Fund International. Washington, D.C.: American Institute for Cancer Research; 2007. [6] Newby PK, Tucker KL. Empirically derived eating patterns using factor or cluster analysis: a review. Nutr Rev 2004;62:177e203. [7] Liu X, Wang X, Lin S, Yuan J, Yu IT. Dietary patterns and oesophageal squamous cell carcinoma: a systematic review and meta-analysis. Br J Cancer 2014;110:2785e95. [8] Prentice RL, Caan B, Chlebowski RT, Patterson R, Kuller LH, Ockene JK, et al. Low-fat dietary pattern and risk of invasive breast cancer: the women's health initiative randomized controlled dietary modification trial. JAMA 2006;295: 629e42. [9] Feng X, Liao J, He Y, Li L, Sun D. Epidemiological characteristics and trends of malignant tumors in Chengdu (in Chinese) Chin J Evid-Based Med 2012;12: 1291e5. [10] Xiao P, Tao D, Huang C, Zheng S, Wang H, Du H. Analysis on dietary structure for residents in high-incidence area of esophageal cancer (in Chinese) Modern Preventive Medicine 2006;33:393e402. [11] Zhao L, Liu CL, Song QK, Deng YM, Qu CX, Li J. Association between dietary behavior and esophageal squamous cell carcinoma in Yanting. Asian Pac J Cancer Prev 2014;15:8657e60. [12] Song Q, Zhao L, Li J, Ren J. Fruit consumption reduces the risk of esophageal cancer in Yanting, People's Republic of China. Asia-Pac J Public Health 2015;27:469e75. [13] Lin S, Wang X, Huang C, Liu X, Zhao J, Yu IT, et al. Consumption of salted meat and its interactions with alcohol drinking and tobacco smoking on esophageal squamous-cell carcinoma. Int J Cancer 2014;137:582e9. [14] Song Q, Wang X, Yu IT, Huang C, Zhou X, Li J, et al. Processed food consumption and risk of esophageal squamous cell carcinoma: a caseecontrol study in a high risk area. Cancer Sci 2012;103:2007e11. [15] Yang CX, Wang HY, Wang ZM, Du HZ, Tao DM, Mu XY, et al. Risk factors for esophageal cancer: a caseecontrol study in South-western China. Asian Pac J Cancer Prev 2005;6:48e53.
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7 [16] Tian D, He Q, Wang Y, Wang Y, Wang J, Wu Y, et al. Survey of eating habits of rural residents in high-incidence area of esophageal cancer and their cognition of esophageal cancer (in Chinese) J North Sichuan Med Coll 2010;26: 470e3. [17] Jiang Y, Du HZ, Zhu WY, Xiao HJ, Huang CY. Effects of a regional Chinese diet and its vitamin supplementation on proliferation of human esophageal cancer cell lines. Biomed Environ Sci 2008;21:442e8. [18] Jiang Y, Huang C, Lei J, Zhu W, Xiao H, Du H. Effects of a regional Chinese diet on proliferation of human esophageal cancer cell line Eca-109 by a serophysiology method. Nutr Cancer 2010;62:543e53. [19] Lee CH, Lee JM, Wu DC, Hsu HK, Kao EL, Huang HL, et al. Independent and combined effects of alcohol intake, tobacco smoking and betel quid chewing on the risk of esophageal cancer in Taiwan. Int J Cancer 2005;113:475e82. [20] Liu X, Wang X, Lin S, Song Q, Lao X, Yu IT. Reproducibility and validity of a food frequency questionnaire for assessing dietary consumption via the dietary pattern method in a Chinese rural population. PLoS One 2015;10: e0134627. [21] Knol MJ, VanderWeele TJ, Groenwold RH, Klungel OH, Rovers MM, Grobbee DE. Estimating measures of interaction on an additive scale for preventive exposures. Eur J Epidemiol 2011;26:433e8. [22] Chinese Society of Nutrition. Dietary guidelines for Chinese residents. Beijing: Tibet People's Publishing House; 2011. [23] Johnson RA, Wichern DW. Applied multivariate statistical analysis. Upper Saddler River, NJ: Prentice Hall; 2002. [24] Field AP. Discovering statistics using SPSS. 3rd ed. Los Angeles; London: SAGE; 2009. [25] J-o Kim, Mueller CW. Factor analysis: statistical methods and practical issues. Beverly Hills, Calif.: Sage Publications; 1978. [26] International Agency for Research on Cancer. Statistical methods in cancer research. Lyon: International Agency for Research on Cancer; 1980. [27] Yang YX, Wang GY, Pan XC. China food composition. 1.2nd ed. Beijing: Peking University Medical Press; 2009. [28] De Stefani E, Deneo-Pellegrini H, Boffetta P, Ronco AL, Aune D, Acosta G, et al. Dietary patterns and risk of cancer: a factor analysis in Uruguay. Int J Cancer 2009;124:1391e7. [29] Hajizadeh B, Rashidkhani B, Rad AH, Moasheri SM, Saboori H. Dietary patterns and risk of oesophageal squamous cell carcinoma: a caseecontrol study. Public Health Nutr 2010;13:1107e12. [30] Bahmanyar S, Ye W. Dietary patterns and risk of squamous-cell carcinoma and adenocarcinoma of the esophagus and adenocarcinoma of the gastric cardia: a population-based caseecontrol study in Sweden. Nutr Cancer 2006;54: 171e8. [31] Huang W, Han Y, Xu J, Zhu W, Li Z. Red and processed meat intake and risk of esophageal adenocarcinoma: a meta-analysis of observational studies. Cancer Causes Control 2013;24:193e201. [32] Islami F, Ren JS, Taylor PR, Kamangar F. Pickled vegetables and the risk of oesophageal cancer: a meta-analysis. Br J Cancer 2009;101:1641e7. [33] Ward MH, Pan WH, Cheng YJ, Li FH, Brinton LA, Chen CJ, et al. Dietary exposure to nitrite and nitrosamines and risk of nasopharyngeal carcinoma in Taiwan. Int J Cancer 2000;86:603e9. [34] Lin SH, Wang XR, Huang CY, Liu XD, Zhao J, Yu ITS, et al. Consumption of salted meat and its interactions with alcohol drinking and tobacco smoking on esophageal squamous-cell carcinoma. Int J Cancer 2015;137:582e9. [35] Salehi M, Moradi-Lakeh M, Salehi MH, Nojomi M, Kolahdooz F. Meat, fish, and esophageal cancer risk: a systematic review and doseeresponse meta-analysis. Nutr Rev 2013;71:257e67. [36] Qu X, Ben Q, Jiang Y. Consumption of red and processed meat and risk for esophageal squamous cell carcinoma based on a meta-analysis. Ann Epidemiol 2013;23:762e70.
7
[37] Choi Y, Song S, Song Y, Lee JE. Consumption of red and processed meat and esophageal cancer risk: meta-analysis. World J Gastroenterol 2013;19: 1020e9. [38] Liu R, Dang S, Yan H, Li Q, Zhao Y, Liu X. Dietary patterns and nutrients intakes in rural residents in Hanzhong of Shaanxi Province. Wei Sheng Yan Jiu 2012;41:997e1003. [39] Gao J, Fei J, Jiang L, Yao W, Lin B, Guo H. Assessment of the reproducibility and validity of a simple food frequency questionnaire used in dietary patters study (in Chinese) Acta Nutr Sin 2011;33:452e6. [40] Ibiebele TI, Hughes MC, Whiteman DC, Webb PM. Dietary patterns and risk of oesophageal cancers: a population-based caseecontrol study. Br J Nutr 2012;107:1207e16. [41] Navarro Silvera SA, Mayne ST, Risch HA, Gammon MD, Vaughan T, Chow WH, et al. Principal component analysis of dietary and lifestyle patterns in relation to risk of subtypes of esophageal and gastric cancer. Ann Epidemiol 2011;21: 543e50. [42] De Stefani E, Boffetta P, Ronco AL, Deneo-Pellegrini H, Correa P, Acosta G, et al. Exploratory factor analysis of squamous cell carcinoma of the esophagus in Uruguay. Nutr Cancer 2008;60:188e95. [43] Hu FB, Rimm E, Smith-Warner SA, Feskanich D, Stampfer MJ, Ascherio A, et al. Reproducibility and validity of dietary patterns assessed with a foodfrequency questionnaire. Am J Clin Nutr 1999;69:243e9. [44] Norat T, Riboli E. Fruit and vegetable consumption and risk of cancer of the digestive tract: meta-analysis of published caseecontrol and cohort studies. IARC Sci Publ 2002;156:123e5. [45] Liu J, Wang J, Leng Y, Lv C. Intake of fruit and vegetables and risk of esophageal squamous cell carcinoma: a meta-analysis of observational studies. Int J Cancer 2013;133:473e85. [46] Islami F, Fedirko V, Tramacere I, Bagnardi V, Jenab M, Scotti L, et al. Alcohol drinking and esophageal squamous cell carcinoma with focus on lightdrinkers and never-smokers: a systematic review and meta-analysis. Int J Cancer 2011;129:2473e84. [47] Bagnardi V, Rota M, Botteri E, Tramacere I, Islami F, Fedirko V, et al. Light alcohol drinking and cancer: a meta-analysis. Ann Oncol 2013;24:301e8. [48] Song QK, Zhao L, Li J, He YM, Jiang CP, Jiang HD, et al. Adverse effects of preserved vegetables on squamous cell carcinoma of esophagus and precancer lesions in a high risk area. Asian Pac J Cancer Prev 2013;14: 659e63. [49] Kamangar F, Chow WH, Abnet CC, Dawsey SM. Environmental causes of esophageal cancer. Gastroenterol Clin North Am 2009;38:27e57. [50] Siddiqi MA, Tricker AR, Kumar R, Fazili Z, Preussmann R. Dietary sources of Nnitrosamines in a high-risk area for esophageal cancer e Kashmir, India. Relevance to human cancer of N-nitroso compounds. Tob Mycotoxins 1991;105:210e3. [51] Mayne ST, Risch HA, Dubrow R, Chow WH, Gammon MD, Vaughan TL, et al. Nutrient intake and risk of subtypes of esophageal and gastric cancer. Cancer Epidemiol Biomark Prev 2001;10:1055e62. [52] Jakszyn P, Gonzalez CA. Nitrosamine and related food intake and gastric and oesophageal cancer risk: a systematic review of the epidemiological evidence. World J Gastroenterol 2006;12:4296e303. [53] Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol 2002;13:3e9. [54] Renehan AG, Tyson M, Egger M, Heller RF, Zwahlen M. Body-mass index and incidence of cancer: a systematic review and meta-analysis of prospective observational studies. Lancet 2008;371:569e78. [55] Wu J, Huang C, Xiao H, Tang Q, Cai W. Weight loss and resting energy expenditure in male patients with newly diagnosed esophageal cancer. Nutrition 2013;29:1310e4.
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
Contents lists available at ScienceDirect
Clinical Nutrition journal homepage: http://www.elsevier.com/locate/clnu
Original article
Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population Xudong Liu a, Xiaorong Wang a, b, *, Sihao Lin a, b, Xiangqian Lao a, Jin Zhao c, Qingkun Song d, Xuefen Su a, Ignatius Tak-Sun Yu a, b a
JC School of Public Health and Primary Care, The Chinese University of Hong Kong, Hong Kong, China Hong Kong Occupational and Environmental Health Academy, Hong Kong, China Shenzhen Center for Disease Control and Prevention, Shenzhen, China d Beijing Key Laboratory of Cancer Therapeutic Vaccine, Capital Medical University Cancer Center, Beijing Shijitan Hospital, Beijing, China b c
a r t i c l e i n f o
s u m m a r y
Article history: Received 6 August 2015 Accepted 16 November 2015
Background & aims: Few studies were available in exploring the roles of dietary patterns in the development of esophageal cancer, especially in China. This study aimed to investigate the roles of dietary patterns in the risk of esophageal squamous cell carcinoma (ESCC) in a Chinese rural population. Methods: A population-based casesecontrol study was designed and conducted in Yanting County, Sichuan Province of China during two years (between June 2011 and May 2013). A total of 942 pairs of ESCC cases and controls were recruited. A food frequency questionnaire was adopted to collect information of dietary consumption. Dietary patterns were extracted by using principle component and factor analysis based on 24 dietary groups. Odds ratios (ORs) with 95% confidence intervals (95% CI) were calculated by using logistic regression model, with adjustment for possible confounding variables. Results: Four major dietary patterns were identified, which were labeled as “prudent”, “vegetable and fruits”, “processed food” and “alcohol drinking”. In comparison of the highest with the lowest quartiles of pattern scores, the processed food pattern (OR: 2.84, 95% CI: 2.13e3.80) and alcohol drinking pattern (OR: 2.69, 95% CI: 1.95e3.71) were significantly associated with an increased risk of ESCC, while the vegetable and fruit pattern (OR: 0.70, 95% CI: 0.53e0.92) was associated with reduced risk by 30%. The prudent pattern was associated with a reduced risk by 33% (OR: 0.67, 95% CI: 0.50e0.88) in a multivariate logistic regression model, but no statistical significance was reached in a composite model. Conclusions: The results suggest an important role of dietary patterns in ESCC. Diets rich in vegetables and fruits may decrease the risk of ESCC, whereas diets rich in processed food and drinking alcohol may increase the risk. © 2015 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved.
Keywords: Dietary pattern Esophageal squamous cell carcinoma Caseecontrol design Principle component and factor analysis Chinese rural population
1. Introduction Esophageal cancer (EC) is the one of the most common cancers and the most common causes of cancer death worldwide [1]. The overwhelming majority of EC cases and deaths occurred in the developing countries, including China [1]. There were 288,000 cases diagnosed annually in China [2], contributing to about 50% of
* Corresponding author. JC School of Public Health and Primary Care, The Chinese University of Hong Kong, 4/F School of Public Health, Prince of Wales Hospital, Shatin, N.T., Hong Kong, China. Tel.: þ852 22528756; fax: þ852 2606 3500. E-mail addresses: [email protected] (X. Liu), [email protected] (X. Wang), [email protected] (S. Lin), [email protected] (X. Lao), szhaojin@ gmail.com (J. Zhao), [email protected] (Q. Song), [email protected] (X. Su), [email protected] (I. Tak-Sun Yu).
the world's total EC cases [1]. In China, EC was the third most common cancer in rural areas and the sixth in urban areas [3]. It was estimated that more than 90% of diagnosed EC cases in the country were esophageal squamous cell carcinoma (ESCC) [3]. Existing evidence has suggested some dietary components have pivotal roles in carcinogenic process of EC [4,5]. In practice, different foods and nutrients are often mixed together and consumed foods contain literally thousands of nutrients and chemicals. These foods and nutrients likely have interactive effects. Studies on individual nutrients or foods hardly assess synergistic or inhibitory effects. Therefore, observed associations between individual nutrient/food and health outcomes may be attenuated [6]. Dietary pattern, as a holistic and comprehensive approach, may better capture the effects of diets. Dietary patterns can more accurately assess actual dietary exposure, better control dietary
http://dx.doi.org/10.1016/j.clnu.2015.11.009 0261-5614/© 2015 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved.
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
2
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7
confounders, deal with problems of conflicting results, and identify nutrients or foods that may have a relatively small effect. A recent literature review provided some suggestive evidence on the association between dietary patterns and ESCC risk [7]. Furthermore, dietary pattern might be a better method for appropriate intervention [8]. However, no study was available on the association between dietary pattern and ESCC risk in China, where there were high incidence and mortality rates of ESCC. Therefore, we designed and conducted the current study in Yanting County located in Sichuan Province, which is one of the areas with the highest incidence and mortality of ESCC in China [2]. The incidence of ESCC in Yanting was 100.11/100,000 for males and 57.10/100,000 for females according to 2009 data [3], much higher than the rates of Sichuan Province (10.38/100,000 for combined genders) [9], and rates of the country (24.42/100,000 for males, 9.60/100,000 for females) [2], and rural area as a whole (32.96/100,000 for males, 14.21/100,000 for females) [2]. Most of people in Yanting were farmers with low socioeconomic status. The people had a special preference for cereals and tuber-crops, pickled and preserved vegetables, and salted meat [10e15]. Although there have been a great change in eating habits and lifestyle in Chinese population in recent decades, such changes have been relatively small and slow in Yanting area because of historically low economic status and geographically remote [16]. Some lab studies indicated that that Yanting diets could induce esophageal cancer in human cell lines [17,18]. Epidemiological studies conducted in Yanting also suggested possible roles of several individual foods or dietary habits in the development of ESCC [11e15]. The objective of the current study was to explore the associations of dietary patterns with the risk of ESCC. 2. Subjects and methods The caseecontrol study was carried out during two years, from June 2011 to May 2013. A detailed recruitment of cases and controls was described elsewhere [13]. Briefly, ESCC cases were selected from Yanting Tumor Hospital, a main institution to diagnose and treat ESCC in the area [14]. Inclusion criteria for cases were: 1) men or women who were aged 40e70 years, 2) newly diagnosed primary incident ESCC confirmed pathologically (ICD-10, code-C15), and 3) having lived in the Yanting area for at least 15 years. Those who had lived outside of Yanting area for six months or longer in the last 15 years were excluded. Among eligible ESCC cases, there were 942 (96%) cases who accepted an interview, accounting for more than 70% of incident cases during the period of the study. By comparing the recruited cases with non-recruited eligible cases, we did not find significant differences in age and gender. The controls were selected by using a multi-stage sampling method from local residents. Inclusion criteria were: 1) no history of neoplasm at any site; 2) no digestive tract disease; and 3) having lived in the Yanting area for not less than 15 years. Those who had lived outside of the Yanting area for six months or longer in the past 15 years were also excluded. The procedures of sampling were described in detail in previous report [13]. In the end, 942 residents were recruited as individual matched controls. Thus, 942 matched cases and controls (672 males and 270 females) were included in the study, with median age of 60 years. All participating subjects provided written consent following the Declaration of Helsinki. The study was approved by the Joint Chinese University of Hong Kong e New Territories East Cluster Clinical Research Ethics Committee. 2.1. Data collection Questionnaires were administered with a face-to-face interview. Interviewers and study subjects were all unaware of the
hypothesis and objectives of the study. Cases were interviewed at the hospital within a week once diagnosed with ESCC, and controls were interviewed at local health clinics. All questionnaires were checked once more each day after information was collected and any unclear responses were clarified by a contact with related subjects to reduce invalid information/data. A structured questionnaire was used to collect information, including individual demographics (age, gender, education, marital status, occupation, monthly household income and history of residence), family cancer history, tobacco smoking and alcohol drinking history, and self-reported height and weight five years before ESCC diagnosed for the cases and prior to the interviews for the controls. An ever drinker was defined as consuming any alcoholic beverage, which included beer, wine or distilled spirits containing at least 20 g of ethanol, per week for minimum 6 months [19], while a never as one who had never been a regular or social drinker. An ever smoker was defined as smoking at least 10 cigarettes or equal amount of tobacco per week for minimal six months [19], while never smoker as individuals who had never smoked as many as one cigarette a day or equivalent for the duration of one year. The group of ever users could be further categorized into current users and former users. Current users were defined as one who had any of these habits within one year before the interview, while former users as had stopped any of these habits for at least one year before diagnoses or interviews. Body mass index (BMI) was obtained according to the formula: weight (kg)/[height (m)]2. A food frequency questionnaire (FFQ) with 76 items, which was proved to have moderate validity and reproducibility [20], was used to investigate usual dietary intake during the past five years before ESCC diagnosis for the cases or before the interviews for the controls. Three basic questions were asked to collect information on the consumption of each of the items. First, “Did you have ever consumed one certain dietary item” was asked. If the answer was “yes”, “how often the consumption was and how much consumed each time” were further asked. Average daily intake of each food or drink was estimated according to the formula: Average daily intake ¼ Frequency of intake per day Amount of intake each time. The detailed calculation of average daily intake was described in previous reports [13,20].
2.2. Statistical analysis t-Test or chi-square test was used to describe the difference of demographic variables and other factors between the case and controls. The association between possible risk factors and ESCC risk was examined by using logistic regression model (forward stepwise), expressed as odds ratios (OR) with 95% Confidence interval (95% CI). The individual and joint effects of smoking and alcohol drinking were deeply analyzed and the synergy index [21] was adopted to display the interaction of alcohol drinking and tobacco smoking on ESCC risk. However, in this study there were only 11 former drinkers and 14 former smokers in the cases and 0 former drinkers and 0 former smokers in the controls. Hence, it is not meaningful to divided ever smokers or drinkers into current and former subgroups. The 76 dietary items were further grouped into 24 dietary groups, according to local eating habits, dietary guideline and balance diet pagoda for Chinese population [22] and whether or not the food was processed, so to derive dietary patterns. The 24 groups were used in all dietary assessment methods after similarities were tested, which ensured the consistency of groups entering into the factor analysis. The original absolute food intake was standardized before factor analysis was performed to avoid problems arisen from different measurement units and different intake amount.
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7
Principle component and factor analysis (PCFA) with correlation matrix of 24 dietary groups were used to derive dietary patterns. Orthogonal (varimax) transformation was adopted to simplify structure and make results interpretable. Eigenvalues (>1.2), the scree plot construction and the total variance and interpretability of the factors were considered for determining the number of factors [23]. The Bartlett's test of sphericity and the KaisereMeyereOlkin measure of sampling (KMO) were used to determine the relationships between variables of interest in the factor analysis [24]. The extracted factors were labeled according to the interpretation of the data [25]. Foods with absolute rotated factor loadings of more than 0.4 were referred to as ‘dominant foods’ hereafter. The dietary pattern scores emerging from data source were calculated by using the regression method. Furthermore, complementary analysis was performed to assess the robustness of the dietary patterns identified. First, the factor scores were calculated with multiple regression method that may standardize results [23]. Then, PCFA was carried out in men and women, respectively. In order to evaluate the internal reproducibility of identified dietary patterns, subjects were randomly divided into two equally-sized subgroups. Then, performance of PCAF was repeated in each subgroup. In light of the identified patterns being robust, factor scores were generated from all subjects and been used in all subsequent analyses. In each of the patterns, subjects were grouped into four categories based on the quartiles of pattern scores among the controls. Crude and adjusted OR and 95%(CI) were estimated from univariate and multivariate logistic regression (forward stepwise) models, respectively [26]. Then, a stratified analysis by gender was performed. In addition, a stratified analysis by smoking status among males was carried out, as few women smoked or drank in the subjects and the study area. The Wald statistics was calculated to determine an exposureeresponse trend, by putting the median scores of each quartile as a continuous variable into univariate and multivariate logistic regression. Potential confounding variables were adjusted in the multivariate model, including age, gender, educational status, tobacco smoking, BMI, family cancer history, and total energy (calories, kJ) intake. Total energy intake for each subject was calculated based on the Chinese food composition tables [27]. All statistical analyses were performed with IBM SPSS (version 19.0, 2010, IBM SPSS Inc.). The tests were two-tailed with p < 0.05 being considered to be statistically significant. 3. Results Table 1 displays basic information of cases and controls. Mean calorie intake was 2210.2 kcal in the cases and 2445.1 kcal in the controls; BMI was 21.6 kg/m2 in the cases and 22.7 kg/m2 in the controls. Smoking and alcohol drinking, as well as family history of cancer were more common in the cases than in the controls. In the controls, a mean value of average daily consumption of each dietary group increased from the lowest to highest quartiles of pattern scores when the positive factor loading increased (Table A in S1 File). On the contrary, a mean value of average daily consumption of each dietary group decreased with a decrease of negative factor loadings. No significant difference in the mean value of average daily consumption of each dietary group was observed across quartiles of dietary pattern scores when absolute factor loading was small. To give more information on alcohol drinking and smoking related risk in the subjects, Table B and Table C in S1 File show the association between alcohol drinking/smoking and the risk of ESCC. Drinking frequency, drinking amount of liquor each time and average ethanol consumption per week were all significantly associated with a higher risk, showing an exposureeresponse trend (p for trend < 0.001) (Table B in S1 File). Compared to never drinkers, liquor drinkers had a double risk (OR: 2.07, 95% CI:
3
Table 1 Characteristics of study subjects. Cases (N ¼ 942) Age, mean (SD), years Gender, males, n (%) Education status, m (%) Lower than primary school Primary school Middle school and above Family cancer history, n (%) No Yes BMI Mean (SD), kg/m2 N (%) <18.5 kg/m2 18.5e<23 kg/m2 23e<25 kg/m2 25 kg/m2 Total calorie intake, kcal/day, mean (SD) Smoking status Mean (SD), pack-years Male, n (%) Never smokersc Ever smokers Female, n (%) Never smokers Ever smokers Drinking status Ethanol intake, mean (SD), g/week Male, n (%) Never drinkersd Ever drinkers Female, n (%) Never drinkers Ever drinkers
Controls (N ¼ 942)
p Value
60.1 (6.8) 672 (71.3)
60.2 (6.8) 672 (71.3)
0.592a /
198 (21.1) 542 (57.5) 202 (21.4)
209 (22.2) 511 (54.2) 222 (23.6)
0.341b
699 (74.2) 243 (25.8)
755 (80.1) 187 (19.9)
21.6 (2.9)
22.7 (2.5)
0.002b
118 540 174 110 2212.2
(12.5) (57.3) (18.5) (11.7) (771.0)
46 470 240 186 2453.1
(4.9) (49.9) (25.5) (19.7) (812.2)
41.5 (20.6)
36.8 (18.3)
102 (15.2) 570 (84.8)
306 (45.5) 366 (54.5)
258 (95.6) 12 (4.4)
263 (97.4) 7 (2.6)
304.9 (471.6)
98.5 (16.8)
100 (14.9) 572 (85.1)
271 (40.3) 401 (59.7)
233 (86.3) 37 (13.7)
239 (88.5) 31 (11.5)
<0.001a <0.001b
<0.001a
<0.001a <0.001b
0.243b
0.001a <0.001b
0.436b
a
p Value for paired t-test. p Value for Chi-square test. c Ever smokers included current smoker and former smoker. There were 14 former smokers in the cases and 0 former smokers in the controls. d Ever drinkers included current drinkers and ever drinkers. There were only 11 former drinkers in the cases and 0 former drinkers in the controls. b
1.57e2.73); consumption of ethanol more than 443.8 g per week was associated with over 7-fold risk (OR: 7.16, 95% CI: 4.61e11.13). However, drinking beer was not significantly associated with risk. Smoking, no matter expressed as the frequency, duration or cumulative smoking amount, was significantly associated with an increased risk of ESCC, with clear exposureeresponse trends (p for trend < 0.001) (Table C in S1 File). Compared to never smokers, smokers had nearly 2.8-fold higher risk (OR: 2.81, 95% CI: 2.12e3.73) after potential confounding factors were adjusted. In the analysis of their joint effect (Table D in S1 File), the greatest risk was observed in those who both smoked and drank (OR: 5.54, 95% CI: 4.04e7.67). The synergy index for smoking and alcohol drinking was 4.10 (95% CI: 1.80e9.35). Overall value of KMO was 0.738 (>0.5 is regarded as acceptable [24]), and p for Bartlett's test of sphericity was less than 0.001 (<0.001 as acceptable [24]), suggesting a proper sample size to perform PCFA. The values of KMO for 24 dietary groups were all greater than 0.55 (Table E in S1 File). Four factors were identified through the PCFA (Table 2), which explained 41% of the total variance. The first defined factor, which was characterized by a higher factor loading of rice, wheat, tofu and dry bean-curd, dry-beans, nuts and seeds, red meat, and fresh eggs, was labeled as ‘prudent pattern’. It mainly represented the diet in the rural area and explained 13.7% of the total variance. The second factor was labeled as ‘vegetable and fruit pattern’, which had a higher factor loading of
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
4
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7
Table 2 Factor loadings for the relationship between food groups and factors representing dietary patterns.a Diet items
Rice Wheat Corn Tuber-crops Cruciferous vegetables Deep-green vegetables Deep-yellow vegetables Citrus fruits Non-citrus fruits Tofu and dry bean-curd Dry-beans Nuts and seeds Red meat White meat Fresh eggs Preserved vegetables Pickled vegetables Salted meat Salted eggs Animal fat Plant oil Tea Liquor Beer Variance explained VAR (%) Cumulative explained VAR (%)
Prudent pattern 0.42 0.45
0.25 0.24 0.54 0.62 0.52 0.41 0.32 0.47 0.14 0.23 0.15 0.26 0.33 0.13
Vegetable and fruits pattern
0.12 0.12 0.13 0.80 0.79 0.51 0.64 0.68 0.11 0.13 0.13
0.14 0.21 0.08 0.28
Processed food pattern
Alcohol drinking pattern
0.12
0.18
0.10
0.22 0.21 0.29 0.24 0.26
0.12 0.11
0.20 0.68 0.46 0.64 0.41
0.11
0.12 0.11 0.25 0.15 0.11
0.18 0.11 0.22 0.36 0.72 0.58
13.77
9.56
9.09
8.64
13.77
32.42
32.42
41.06
a Principle component and factor analysis was performed based on 24 dietary items. With orthogonal rotation, the factor loading scores are identical to the correlation coefficients. The magnitude of each loading indicates the importance of the corresponding items to the factor. Loadings 0.40 were shown in bold typeface and loadings < 0.10 were suppressed for clarity.
cruciferous vegetables, dark-green vegetables, dark-yellow vegetables, citrus fruits and non-citrus fruits explaining 9.6% of the total variance. The third factor was labeled as ‘processed food pattern’, which had high loadings of pickled vegetables, preserved vegetables, salted meat and salted eggs, explaining 9.1% of the total variance. The fourth factor, as ‘alcohol drinking pattern’, had a high loading of beer and liquor, which explained 8.6% of the total variance. In the multivariate analysis, the processed food pattern and alcohol drinking pattern were associated with an increased risk of ESCC (Table 3). The highest quartile of pattern scores was related to a nearly 3-fold risk for the processed food pattern (OR: 2.84, 95% CI: 2.13e3.80, p for trend < 0.001) and the alcohol drinking pattern (OR: 2.69, 95% CI: 1.95e3.71, p for trend < 0.001), in comparison with the lowest quartiles of the scores. The highest quartile of the prudent pattern scores was related to a lower risk by 33% (OR: 0.67, 95% CI: 0.53e0.92). Similarly, the highest quartile of the vegetable and fruits pattern scores was related to 30% lower risk (OR: 0.70, 95% CI: 0.53e0.92, p for trend < 0.005). When all the four patterns and major confounding variables were entered simultaneously into a same model (Table F in S1 File), the positive or inverse associations remained, with the exception of the prudent pattern, of which association with the risk did not reach a statistical significance in the composite model. In stratified analysis by gender (Table 4), both processed food pattern and alcohol drinking pattern were associated with an increased risk in males. In the comparison of the highest with the lowest quartile of scores, OR was 2.58 (95% CI: 1.83e3.64) for the
Table 3 Associations between dietary patterns and ESCC risk. Dietary pattern
Case/control
Prudent Quartile 1 294/235 Quartile 2 242/236 Quartile 3 203/235 Quartile 4 203/236 p for trend Vegetable and fruits Quartile 1 317/235 Quartile 2 221/236 Quartile 3 205/235 Quartile 4 199/236 p for trend Processed food Quartile 1 139/235 Quartile 2 187/235 Quartile 3 200/236 Quartile 4 415/236 p for trend Alcohol drinking Quartile 1 167/235 Quartile 2 166/236 Quartile 3 162/235 Quartile 4 446/236 p for trend
Crude OR (95% CI)
Adjusted ORa (95% CI)
1.00 0.82 (0.64, 1.05) 0.69 (0.54, 0.89) 0.68 (0.53, 0.89) 0.002
1.00 0.80 (0.61, 1.05) 0.70 (0.53, 0.93) 0.67 (0.50, 0.88) 0.040
1.00 0.69 (0.54, 0.89) 0.64 (0.50, 0.83) 0.63 (0.49, 0.81) <0.001
1.00 0.80 (0.61, 1.05) 0.74 (0.56, 0.98) 0.70 (0.53, 0.92) 0.005
1.00 1.35 (1.01, 1.79) 1.43 (1.08, 1.89) 2.94 (2.26, 3.83) <0.001
1.00 1.25 (0.92, 1.71) 1.38 (1.01, 1.88) 2.84 (2.13, 3.80) <0.001
1.00 0.99 (0.75, 1.32) 0.97 (0.73, 1.28) 2.67 (2.07, 3.44) <0.001
1.00 1.06 (0.78, 1.45) 1.05 (0.76, 1.46) 2.69 (1.95, 3.71) <0.001
a OR was adjusted for age, gender, educational status (lower than primary school, primary school, higher than primary school), BMI (continuous, kg/m2), smoking status (no vs. yes), family history of cancer (no vs. yes), and total calorie intake (kcal, log 10-transformation).
processed food pattern, and 2.99 (95% CI: 2.07e4.31) for the alcohol drinking pattern. The highest scores of vegetable and fruits pattern were associated with a reduced risk by 46% (OR: 0.54, 95% CI: 0.38e0.78). Similar results for the processed food pattern and vegetable fruit pattern were obtained in females. No significant association with the prudent pattern was observed in either gender. In the stratified analysis by smoking status in males (Table 5), both processed food pattern and the alcohol drinking pattern were significantly associated with an increased risk, and the vegetable and fruit pattern was associated with a reduced risk in either smokers or non-smokers, in the comparison of the highest with the lowest quartiles of pattern scores. Likewise, the prudent pattern was not associated with ESCC risk in either subgroup. 4. Discussion In this study, four main dietary patterns were identified, i.e. the prudent, vegetable and fruit, processed food and alcohol drinking patterns. The processed food pattern and the alcohol drinking pattern were found to be associated with an increased risk of ESCC, whereas vegetable and fruit pattern was associated with a decreased risk. There was no significant association with the prudent pattern observed in the study subjects, however. Some components in the dietary patterns of this study were similar with those of the patterns defined in some other studies, although not all components were identical. For instance, the processed food pattern had higher factor loadings in four types of the processed foods, such as processed meat, pickled vegetables, preserved vegetables and salted eggs, which were similar to the western pattern labeled in studies from Uruguay (containing barbecued meat) [28], Iran (containing pickle and processed meat) [29], and Sweden (containing processed meat) [30]. These types of processed food were thought to be risk factors for digestive cancers [14,31e33]. Most of the components of the prudent pattern in the present study were similar to those of the prudent pattern reported in Uruguay [28] and the healthy pattern in Iran and Sweden [29,30].
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7 Table 4 Associations between dietary patterns and ESCC risk by gender. Dietary patterns
Males Cases/ controls
Prudent Quartile 1 201/168 Quartile 2 167/168 Quartile 3 154/168 Quartile 4 153/168 p for trend Vegetable and fruits Quartile 1 175/168 Quartile 2 218/168 Quartile 3 174/168 Quartile 4 105/168 p for trend Processed food Quartile 1 103/168 Quartile 2 137/168 Quartile 3 147/168 Quartile 4 285/168 p for trend Alcohol drinking Quartile 1 85/168 Quartile 2 81/168 Quartile 3 128/168 Quartile 4 378/168 p for trend
Females ORa (95% CI)
Cases/ controls
ORa (95% CI)
1.00 0.83 (0.60, 1.16) 0.82 (0.58, 1.15) 0.78 (0.56, 1.08) 0.196
87/68 73/67 57/67 53/68
1.00 0.73 (0.45, 1.19) 0.63 (0.38, 1.04) 0.60 (0.36, 1.01) 0.041
1.00 1.28 (0.92, 1.77) 0.91 (0.65, 1.27) 0.54 (0.38, 0.78) <0.001
92/68 62/67 49/67 62/68
1.00 0.79 (0.48, 1.29) 0.93 (0.56, 1.53) 0.61 (0.36, 0.84) 0.639
1.00 1.35 (0.94, 1.96) 1.33 (0.92, 1.91) 2.58 (1.83, 3.64) <0.001
34/68 54/67 54/67 128/68
1.00 1.41 (0.80, 2.49) 1.49 (0.85, 2.63) 3.16 (1.87, 5.35) <0.001
1.00 0.77 (0.51, 1.16) 1.10 (0.75, 1.62) 2.99 (2.07, 4.31) <0.001
68/68 51/67 70/67 81/68
1.00 0.65 (0.38, 1.10) 0.99 (0.59, 1.64) 1.27 (0.77, 2.09) 0.189
a OR was adjusted for age, educational status (lower than primary school, primary school, higher than primary school), BMI (continuous, kg/m2), family history of cancer (yes vs. no), and total calorie intake (kcal, log 10-transformation).
Salted meat, which was found to increase the risk of ESCC [34], was dominant components of the processed food pattern. On the other hand, red meat and fresh eggs, which were two main components of the prudent in this study, were labeled as the western pattern in other studies [28e30]. It was reported that red meat and processed
Table 5 Associations between dietary patterns and ESCC risk by smoking status in males. Dietary patterns
Non-smoker Cases/ controls
Prudent Quartile 1 24/77 Quartile 2 22/76 Quartile 3 22/76 Quartile 4 34/77 p for trend Vegetable and fruits Quartile 1 31/77 Quartile 2 37/76 Quartile 3 19/76 Quartile 4 15/77 p for trend Processed food Quartile 1 19/77 Quartile 2 25/76 Quartile 3 20/76 Quartile 4 38/77 p for trend Alcohol drinking Quartile 1 18/77 Quartile 2 23/76 Quartile 3 17/76 Quartile 4 44/77 p for trend
Smoker a
OR (95% CI)
Cases/ controls
ORa (95% CI)
1.00 0.71 (0.35, 1.42) 0.73 (0.37, 1.46) 1.18 (0.62, 2.23) 0.537
182/92 143/91 126/81 119/92
1.00 0.88 (0.60, 1.30) 0.78 (0.53, 1.14) 0.68 (0.46, 1.01) 0.041
1.00 1.20 (0.66, 2.17) 0.66 (0.33, 1.29) 0.49 (0.24, 0.97) 0.019
153/92 173/91 156/91 88/92
1.00 1.16 (0.79, 1.69) 1.03 (0.70, 1.52) 0.53 (0.35, 0.80) 0.005
1.00 1.15 (0.56, 2.33) 1.08 (0.52, 2.23) 1.87 (1.05, 3.63) 0.065
92/92 106/91 130/91 242/92
1.00 1.21 (0.79, 1.86) 1.41 (0.93, 2.15) 2.73 (1.83, 4.05) <0.001
1.00 1.36 (0.66, 2.8) 1.09 (0.51, 2.33) 2.65 (1.38, 5.12) 0.004
86/92 89/91 103/91 292/92
1.00 1.10 (0.71, 1.70) 1.32 (0.86, 2.02) 3.97 (2.66, 5.92) <0.001
a OR was adjusted for age, gender, educational status (lower than primary school, primary school, higher than primary school), BMI (continuous, kg/m2), family cancer history (yes vs. no), and total calorie intake (kcal, log 10-transformation).
5
meat increased the risk of ESCC, especially when meat is barbecued, fried and rotated at higher temperature [35e37]. These components were classified into the prudent pattern, because many traditional foods were prepared with red meat and eggs in Chinese population [38,39]. Moreover, average intake of red meat (32 g/day) and salted meat (30 g/day) were much higher than intake of poultry (3 g/day) and seafood (1 g/day) in the study area [10]. Red meat and fresh eggs were two major food sources that provided protein, energy, some minerals and other nutrients in the rural area with a lower economic status in China. The components of the vegetable and fruit pattern were similar to same patterns reported from Australia [40] and the U.S. [41], and the healthy pattern from Uruguay [42]. The alcohol pattern in the study was similar to the drinker pattern reported from Uruguay men [28,42], the smoking/ alcohol pattern from the U.S. [41] and the alcohol drinker pattern from Sweden [30]. The components of the dietary patterns, however, were somewhat different in studies conducted in different countries, possibly due to differences in the dietary assessment methods, food availability, and ethnical and cultural differences [43]. We observed that the vegetable and fruit pattern was associated with a reduced risk by nearly 30%. Similar results were also obtained in the stratified analyses. The result was consistent with some other studies [40e42]. The World Cancer Research Fund report in 2007 [5] and two updated systematic review and metaanalysis [44,45] state that there is strong evidence to support that fruits and vegetables have a protecting effect against ESCC. This study provided further supportive data for that. Smoking and alcohol drinking were significantly associated with an increased risk of ESCC in this study, which was consistent with existing evidence [5]. However, we did not find an evident association of drinking beer with ESCC, possibly because of a very small number of the subjects who drank beer. Alcohol drinking is a well-established risk factor for ESCC [5,46,47]. This study observed an increased risk associated with the alcohol drinking pattern, which was consistent with the results from systematic review and meta-analysis [7]. In the stratified analysis by gender, a significant association with the alcohol drinking pattern was only observed in males, but not in females. An explanation was a small number of alcohol drinkers in the females (13.7% in the cases and 11.5% in the controls). Some studies showed that processed food, including preserved vegetables, pickled vegetables, and salted meat, was associated with an increased risk [13,14,48]. Pickled vegetables, preserved vegetables and salted meat were commonly consumed by Yanting residents. This study integrated the different kinds of processed food into the processed food pattern. The pattern was associated with 2.6-fold risk in males (OR: 2.58, 95% CI: 1.83e3.64) and 3.2fold risk in females (OR: 3.16, 95% CI: 1.87e5.35). Similar result related to a labeled western pattern was reported from an Iran study [29]. Processed foods are sources of nitrites and N-nitroso compounds [49,50], which could exert an effect on carcinogenesis [51] and contribute to the development of EC [52]. In this study, the prudent pattern was observed to be related to a 33% decreased risk of ESCC, when the highest and the lowest quartiles of pattern scores was compared. This was consistent with a study conducted in Uruguay, in which the prudent pattern was associated with a decreased risk by 50% [28]. However, there was no significant association with the prudent pattern in a composite model that included all interested patterns and confounders. Therefore, the role of prudent pattern in ESCC needs to further study. Factor analysis is a major posteriori method in identifying dietary patterns in studying the dietedisease relationship [53]. Factor analysis is a data-driven method and its reliability is largely
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
6
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7
dependent on sample size [24]. This study included 942 pairs of the cases and controls, which met the suggested criteria (>300) [24]. Although sample sizes might have reduced in stratified analysis, overall and individual values of KMO were all higher than 0.55, suggesting that factor analysis was acceptable in this study [24]. A limitation of the factor analysis may be a subjective decision in defining dietary patterns, including food/food groups selected, number of dominant factors to be retained and the interpretation of factors. However, the reproducibility and validity study of the FFQ [20] suggested PCFA was stable and robust in identifying dietary patterns. There was a significant difference in BMI and calorie intake between cases and controls. Due to the retrospective nature of this study, recall bias cannot be avoided. However, the food frequency questionnaire used in this study was proved to be reproducible and valid in exploring the dietary intake [20]. BMI, which was suggested to have critical effect on the occurrence of ESCC, was related to the total caloric intake [54]. High BMI might be due to the high level of calorie intake. Patients who suffered from tumors, especially the digestive tumors, might have an abnormal diet intake and then lose their weight during the progress of the disease [55]. However, in this study, the subjects were required to report the anthropological and diet information five years before being diagnosed as ESCC for the cases and prior to the interviews for the controls, which made sure that the influence on BMI by the disease progression was small. Moreover, these two variables as well as other confounders were adjusted via regression model to reduce the possible confounding effects. This study is among the few, if any, to assess the associations of dietary patterns and the risk of ESCC in China, where dietary patterns were largely different from those of the western countries. The strengths of the study included a matched caseecontrol design, high response rates and relatively large numbers of cases and controls. There were no differences in major socialdemographic characteristics between the recruited and nonrecruited cases and controls. All cases were histological confirmed incidence cases, thus largely reduced disease misclassification. In addition, detailed information on potential confounding factors was collected and adjusted in the analyses. However, because of a retrospective nature of the caseecontrol design, information bias, in particular recall bias, could be a major concern. However, we made an effort to minimize recall bias by blinding the cases, controls, and interviewers about the study hypothesis and objectives. In addition, the cases were interviewed within one week once diagnosed as ESCC, so to minimize the likelihood of changing a dietary habit following the diagnosis and treatment of the disease. In addition, Standard training was provided to the interviewers to handle cases and controls in the same way. The FFQ used in our study is reasonably reproducible and valid to collect information on the consumption of common foods and drinks [20]. Another limitation is that there was a small number of females and in some subgroups (e.g. male nonsmokers), which reduced the statistical power to detect a significant association in specific subgroups. In conclusion, this study identified four major dietary patterns in the subjects of the study. The results suggested that the processed food pattern and alcohol drinking pattern were associated with an increased risk of ESCC, whereas the vegetable and fruit pattern reduced the risk. The study provided further support to the current recommendation on decreasing the consumption of processed food and alcohol drinking and increasing the vegetable and fruit intake for EC prevention. Further prospective studies are needed to verify the roles of the dietary patterns in the development of ESCC in Chinese population.
Conflict of interest The authors state that they have no conflicts of interest. Funding source and its roles The study was supported by the World Cancer Research Fund, International (No. 2010/240). The founder had no role in the design, analysis or writing of this manuscript. Statement of authorship XW conceived and designed the study; XDL, SL and QS collected the data, XDL analyzed the data and drafted the manuscript under the supervision of XW; XW, SL, JZ, QS, XQL, XS, and IY reviewed and edited the manuscript. All co-authors provided comments and approved the final version. Acknowledgments The authors would like to thank nurses and doctors in the Yanting Tumor Hospital for their cooperation in data collection; thank all study subjects for their participation. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.clnu.2015.11.009. References [1] Stewart BW, Wild C. World cancer report 2014. Lyon, France: International Agency for Research on Cancer; 2014. [2] Chen WQ, Zheng RS, Zhang SW, Zeng HM, Fan YG, Qiao YL, et al. Esophageal cancer incidence and mortality in China. Thorac Cancer 2010;2014(5):343e8. [3] Chen WQ, Hao J. Chinese cancer registry annual report, 2012 (in Chinese). Beijing: Military Medical Science Press; 2012. [4] Reszka E, Wasowicz W, Gromadzinska J. Genetic polymorphism of xenobiotic metabolising enzymes, diet and cancer susceptibility. Br J Nutr 2006;96: 609e19. [5] American Institute for Cancer Research, World Cancer Research Fund. Food, nutrition, physical activity and the prevention of cancer: a global perspective: a project of World Cancer Research Fund International. Washington, D.C.: American Institute for Cancer Research; 2007. [6] Newby PK, Tucker KL. Empirically derived eating patterns using factor or cluster analysis: a review. Nutr Rev 2004;62:177e203. [7] Liu X, Wang X, Lin S, Yuan J, Yu IT. Dietary patterns and oesophageal squamous cell carcinoma: a systematic review and meta-analysis. Br J Cancer 2014;110:2785e95. [8] Prentice RL, Caan B, Chlebowski RT, Patterson R, Kuller LH, Ockene JK, et al. Low-fat dietary pattern and risk of invasive breast cancer: the women's health initiative randomized controlled dietary modification trial. JAMA 2006;295: 629e42. [9] Feng X, Liao J, He Y, Li L, Sun D. Epidemiological characteristics and trends of malignant tumors in Chengdu (in Chinese) Chin J Evid-Based Med 2012;12: 1291e5. [10] Xiao P, Tao D, Huang C, Zheng S, Wang H, Du H. Analysis on dietary structure for residents in high-incidence area of esophageal cancer (in Chinese) Modern Preventive Medicine 2006;33:393e402. [11] Zhao L, Liu CL, Song QK, Deng YM, Qu CX, Li J. Association between dietary behavior and esophageal squamous cell carcinoma in Yanting. Asian Pac J Cancer Prev 2014;15:8657e60. [12] Song Q, Zhao L, Li J, Ren J. Fruit consumption reduces the risk of esophageal cancer in Yanting, People's Republic of China. Asia-Pac J Public Health 2015;27:469e75. [13] Lin S, Wang X, Huang C, Liu X, Zhao J, Yu IT, et al. Consumption of salted meat and its interactions with alcohol drinking and tobacco smoking on esophageal squamous-cell carcinoma. Int J Cancer 2014;137:582e9. [14] Song Q, Wang X, Yu IT, Huang C, Zhou X, Li J, et al. Processed food consumption and risk of esophageal squamous cell carcinoma: a caseecontrol study in a high risk area. Cancer Sci 2012;103:2007e11. [15] Yang CX, Wang HY, Wang ZM, Du HZ, Tao DM, Mu XY, et al. Risk factors for esophageal cancer: a caseecontrol study in South-western China. Asian Pac J Cancer Prev 2005;6:48e53.
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009
X. Liu et al. / Clinical Nutrition xxx (2015) 1e7 [16] Tian D, He Q, Wang Y, Wang Y, Wang J, Wu Y, et al. Survey of eating habits of rural residents in high-incidence area of esophageal cancer and their cognition of esophageal cancer (in Chinese) J North Sichuan Med Coll 2010;26: 470e3. [17] Jiang Y, Du HZ, Zhu WY, Xiao HJ, Huang CY. Effects of a regional Chinese diet and its vitamin supplementation on proliferation of human esophageal cancer cell lines. Biomed Environ Sci 2008;21:442e8. [18] Jiang Y, Huang C, Lei J, Zhu W, Xiao H, Du H. Effects of a regional Chinese diet on proliferation of human esophageal cancer cell line Eca-109 by a serophysiology method. Nutr Cancer 2010;62:543e53. [19] Lee CH, Lee JM, Wu DC, Hsu HK, Kao EL, Huang HL, et al. Independent and combined effects of alcohol intake, tobacco smoking and betel quid chewing on the risk of esophageal cancer in Taiwan. Int J Cancer 2005;113:475e82. [20] Liu X, Wang X, Lin S, Song Q, Lao X, Yu IT. Reproducibility and validity of a food frequency questionnaire for assessing dietary consumption via the dietary pattern method in a Chinese rural population. PLoS One 2015;10: e0134627. [21] Knol MJ, VanderWeele TJ, Groenwold RH, Klungel OH, Rovers MM, Grobbee DE. Estimating measures of interaction on an additive scale for preventive exposures. Eur J Epidemiol 2011;26:433e8. [22] Chinese Society of Nutrition. Dietary guidelines for Chinese residents. Beijing: Tibet People's Publishing House; 2011. [23] Johnson RA, Wichern DW. Applied multivariate statistical analysis. Upper Saddler River, NJ: Prentice Hall; 2002. [24] Field AP. Discovering statistics using SPSS. 3rd ed. Los Angeles; London: SAGE; 2009. [25] J-o Kim, Mueller CW. Factor analysis: statistical methods and practical issues. Beverly Hills, Calif.: Sage Publications; 1978. [26] International Agency for Research on Cancer. Statistical methods in cancer research. Lyon: International Agency for Research on Cancer; 1980. [27] Yang YX, Wang GY, Pan XC. China food composition. 1.2nd ed. Beijing: Peking University Medical Press; 2009. [28] De Stefani E, Deneo-Pellegrini H, Boffetta P, Ronco AL, Aune D, Acosta G, et al. Dietary patterns and risk of cancer: a factor analysis in Uruguay. Int J Cancer 2009;124:1391e7. [29] Hajizadeh B, Rashidkhani B, Rad AH, Moasheri SM, Saboori H. Dietary patterns and risk of oesophageal squamous cell carcinoma: a caseecontrol study. Public Health Nutr 2010;13:1107e12. [30] Bahmanyar S, Ye W. Dietary patterns and risk of squamous-cell carcinoma and adenocarcinoma of the esophagus and adenocarcinoma of the gastric cardia: a population-based caseecontrol study in Sweden. Nutr Cancer 2006;54: 171e8. [31] Huang W, Han Y, Xu J, Zhu W, Li Z. Red and processed meat intake and risk of esophageal adenocarcinoma: a meta-analysis of observational studies. Cancer Causes Control 2013;24:193e201. [32] Islami F, Ren JS, Taylor PR, Kamangar F. Pickled vegetables and the risk of oesophageal cancer: a meta-analysis. Br J Cancer 2009;101:1641e7. [33] Ward MH, Pan WH, Cheng YJ, Li FH, Brinton LA, Chen CJ, et al. Dietary exposure to nitrite and nitrosamines and risk of nasopharyngeal carcinoma in Taiwan. Int J Cancer 2000;86:603e9. [34] Lin SH, Wang XR, Huang CY, Liu XD, Zhao J, Yu ITS, et al. Consumption of salted meat and its interactions with alcohol drinking and tobacco smoking on esophageal squamous-cell carcinoma. Int J Cancer 2015;137:582e9. [35] Salehi M, Moradi-Lakeh M, Salehi MH, Nojomi M, Kolahdooz F. Meat, fish, and esophageal cancer risk: a systematic review and doseeresponse meta-analysis. Nutr Rev 2013;71:257e67. [36] Qu X, Ben Q, Jiang Y. Consumption of red and processed meat and risk for esophageal squamous cell carcinoma based on a meta-analysis. Ann Epidemiol 2013;23:762e70.
7
[37] Choi Y, Song S, Song Y, Lee JE. Consumption of red and processed meat and esophageal cancer risk: meta-analysis. World J Gastroenterol 2013;19: 1020e9. [38] Liu R, Dang S, Yan H, Li Q, Zhao Y, Liu X. Dietary patterns and nutrients intakes in rural residents in Hanzhong of Shaanxi Province. Wei Sheng Yan Jiu 2012;41:997e1003. [39] Gao J, Fei J, Jiang L, Yao W, Lin B, Guo H. Assessment of the reproducibility and validity of a simple food frequency questionnaire used in dietary patters study (in Chinese) Acta Nutr Sin 2011;33:452e6. [40] Ibiebele TI, Hughes MC, Whiteman DC, Webb PM. Dietary patterns and risk of oesophageal cancers: a population-based caseecontrol study. Br J Nutr 2012;107:1207e16. [41] Navarro Silvera SA, Mayne ST, Risch HA, Gammon MD, Vaughan T, Chow WH, et al. Principal component analysis of dietary and lifestyle patterns in relation to risk of subtypes of esophageal and gastric cancer. Ann Epidemiol 2011;21: 543e50. [42] De Stefani E, Boffetta P, Ronco AL, Deneo-Pellegrini H, Correa P, Acosta G, et al. Exploratory factor analysis of squamous cell carcinoma of the esophagus in Uruguay. Nutr Cancer 2008;60:188e95. [43] Hu FB, Rimm E, Smith-Warner SA, Feskanich D, Stampfer MJ, Ascherio A, et al. Reproducibility and validity of dietary patterns assessed with a foodfrequency questionnaire. Am J Clin Nutr 1999;69:243e9. [44] Norat T, Riboli E. Fruit and vegetable consumption and risk of cancer of the digestive tract: meta-analysis of published caseecontrol and cohort studies. IARC Sci Publ 2002;156:123e5. [45] Liu J, Wang J, Leng Y, Lv C. Intake of fruit and vegetables and risk of esophageal squamous cell carcinoma: a meta-analysis of observational studies. Int J Cancer 2013;133:473e85. [46] Islami F, Fedirko V, Tramacere I, Bagnardi V, Jenab M, Scotti L, et al. Alcohol drinking and esophageal squamous cell carcinoma with focus on lightdrinkers and never-smokers: a systematic review and meta-analysis. Int J Cancer 2011;129:2473e84. [47] Bagnardi V, Rota M, Botteri E, Tramacere I, Islami F, Fedirko V, et al. Light alcohol drinking and cancer: a meta-analysis. Ann Oncol 2013;24:301e8. [48] Song QK, Zhao L, Li J, He YM, Jiang CP, Jiang HD, et al. Adverse effects of preserved vegetables on squamous cell carcinoma of esophagus and precancer lesions in a high risk area. Asian Pac J Cancer Prev 2013;14: 659e63. [49] Kamangar F, Chow WH, Abnet CC, Dawsey SM. Environmental causes of esophageal cancer. Gastroenterol Clin North Am 2009;38:27e57. [50] Siddiqi MA, Tricker AR, Kumar R, Fazili Z, Preussmann R. Dietary sources of Nnitrosamines in a high-risk area for esophageal cancer e Kashmir, India. Relevance to human cancer of N-nitroso compounds. Tob Mycotoxins 1991;105:210e3. [51] Mayne ST, Risch HA, Dubrow R, Chow WH, Gammon MD, Vaughan TL, et al. Nutrient intake and risk of subtypes of esophageal and gastric cancer. Cancer Epidemiol Biomark Prev 2001;10:1055e62. [52] Jakszyn P, Gonzalez CA. Nitrosamine and related food intake and gastric and oesophageal cancer risk: a systematic review of the epidemiological evidence. World J Gastroenterol 2006;12:4296e303. [53] Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol 2002;13:3e9. [54] Renehan AG, Tyson M, Egger M, Heller RF, Zwahlen M. Body-mass index and incidence of cancer: a systematic review and meta-analysis of prospective observational studies. Lancet 2008;371:569e78. [55] Wu J, Huang C, Xiao H, Tang Q, Cai W. Weight loss and resting energy expenditure in male patients with newly diagnosed esophageal cancer. Nutrition 2013;29:1310e4.
Please cite this article in press as: Liu X, et al., Dietary patterns and the risk of esophageal squamous cell carcinoma: A population-based caseecontrol study in a rural population, Clinical Nutrition (2015), http://dx.doi.org/10.1016/j.clnu.2015.11.009