Although traditionally nutritional research has focused primarily on single nutrients or individual foods, interest is growing in dietary patterns that consider the complexity of the overall diet(Reference Zarraga and Schwarz1, Reference Hu2). An exploratory approach using factor or cluster analysis empirically identifies patterns that represent actual eating behaviours of the study population. Typically, analyses extract two to six patterns that reflect different dietary compositions(Reference Newby and Tucker3).
Recently, several experimental, clinical and epidemiological studies have examined dietary patterns and found that patterns reflecting certain eating habits are associated with the risk of chronic diseases(Reference Hu2, Reference Cai, Zheng and Xiang4–Reference van Dam, Rimm and Willett7), including cancer(Reference Terry, Hu and Hansen8, Reference Terry, Suzuki and Hu9). Dietary habits are also closely related to other health-related behaviours. Current smoking is generally inversely related to a healthy diet(Reference Bamia, Orfanos and Ferrari10–Reference Schulze, Hoffmann and Kroke13); alcohol consumption is positively related to diets with high intakes of fat and meat, but inversely related to a prudent diet(Reference Park, Murphy and Wilkens14–Reference Imamura, Lichtenstein and Dallal17); and dietary supplement use is often associated with a healthy diet that reflects high intakes of skimmed milk, yoghurt, juice, cereals, rice, chicken, fruit and cod liver oil(Reference McNaughton, Mishra and Stephen11, Reference Beitz, Mensink and Hintzpeter18, Reference Kerver, Yang and Bianchi19). However, most studies of these factors have been conducted in North America or Europe, whereas only a few studies have investigated Asian populations(Reference Kim, Sasaki and Sasazuki15, Reference Shin, Oh and Park20), and the results may not apply to the Korean population whose dietary habits are different from patterns in Western countries.
Thus, the purpose of the present study was to evaluate Korean dietary patterns by using an FFQ and to assess the association of dietary patterns with lifestyle risk factors, such as smoking, alcohol consumption, physical activity and dietary supplement use.
Experimental methods
Study participants
The source population comprised 14 531 men and women who underwent cancer screening examinations at the Center for Cancer Prevention and Detection at the National Cancer Center in South Korea from August 2002 to May 2007. Participants who failed to complete information for all sixteen food groups in the FFQ or who were <30 years of age were excluded. After these exclusions, 11 440 participants (6434 men and 5006 women) remained for the analyses. Written informed consent was obtained from all participants, and all procedures were approved by the Institutional Review Board of the National Cancer Center.
Data collection
All participants were asked to complete a self-administered questionnaire about their sociodemographic characteristics (e.g. age, education and household income), cigarette smoking habit, alcohol consumption habit, physical activity, dietary supplement use, personal medical history and medication use. The average alcohol consumption amount (g/d) was calculated by summing the beverage-specific amount consumed as reported on the FFQ, average consumption frequency and the volume of one standard drink for each type of alcoholic beverage (beer, hard liquor, wine and traditional drinks, including soju and Korean rice wine).
At the time of screening, for each participant, height and weight were measured using the InBody 3·0 (Biospace, Seoul, Korea) body composition analyser, and BMI was calculated as weight (kg) divided by the square of height (m2). Three categories were constructed for smoking habits (non-, ex- and current) and alcohol consumption (non-, light and heavy), and two categories each were constructed for participation in leisure-time physical activity, dietary supplement use and past medical history of hypertension and diabetes (yes or no).
Dietary pattern derivation
Dietary information was collected using an FFQ. The FFQ consisted of sixteen food groups: cereals, salted vegetables and seafood, light-coloured vegetables, green-yellow vegetables, seaweeds, fruits, grilled meat and seafood, healthy protein foods, dairy products, bonefish, fried foods, cholesterol-rich foods, animal fat-rich foods, sweet foods, fast foods and caffeinated drinks (see Appendix). All participants were asked to record their current intake frequency for each food group according to the following categories: consumed rarely, once monthly, 2–3 times/month, once weekly, 2–3 times/week, 4–6 times/week, once daily or >2 times/d. In our previous study, the validity of this FFQ was compared to the 3 d dietary record method as a reference standard for 1401 participants from the source cohort. The cross-classification of the tertile categories by using these two methods showed good agreement (range: 38–96 %, depending on the food group). Generally, there was a trend that participants who reported a high frequency of intake on the brief FFQ also reported a high amount of intake of food in the 3 d dietary records. The highest agreement was observed for cereals (68·7 %), dairy foods (59·1 %), fruits (55·6 %) and caffeinated drinks (61·8 %)(Reference Shin, Lim and Sung21).
Statistical analyses
Dietary patterns were derived by using factor analysis, with the sixteen food groups entered into the analysis by the frequency of intake. The PROC FACTOR procedure in the SAS statistical software package version 9·1 (SAS Institute, Cary, NC, USA) was applied to perform the analysis. This procedure uses factor analysis and orthogonal rotation, using the Varimax option in SAS, to derive non-correlated factors and to render the results more easily interpretable. To determine which number of factors to retain, we examined both the scree plots and the factors themselves to see which set of factors most meaningfully described distinct food intake patterns. We considered components with an eigenvalue of >1·0. This served to limit the number of factors, as well as to better identify the three most meaningful factors. Factor loadings were calculated for each food group across the three factors. Factors were thereby interpreted as dietary patterns and named after the food groups with a loading of >0·2. A dietary factor score for each individual was then calculated by summing the consumption of food groups weighted by the factor loadings.
Dietary factor scores were categorized into quartiles separately for men and women based on the distribution of the study population, and linear regression analysis was performed to evaluate the association between dietary pattern categories and health behaviour variables with adjustments for age. P for trend was calculated from generalized linear models with adjustments for age as a continuous variable and from Mantel–Haenszel χ 2 tests with adjustments for age group as a categorical variable. To assess associations between dietary patterns and health behaviours, polytomous multiple logistic regression models were used to calculate the OR and 95 % CI for each quartile. P for trend in the OR was calculated with the dietary pattern categories as a continuous variable. Analyses were conducted using the SAS software version 9·1 (SAS Institute). All analyses were performed separately for men and women, and a two-sided P value of <0·05 was considered to be statistically significant.
Results
Among men and women aged ≥30 years who were recruited between 2002 and 2007 from the National Cancer Center in South Korea, 11 440 participants were included in the final analysis (6434 men and 5006 women). Table 1 shows the three main dietary patterns and the factor-loading matrix between food groups. The larger the loading of a given food group to the factor, the greater the contribution of that food group to a specific factor. Dietary pattern 1 was characterized by high consumption of fast foods, animal fat-rich foods, fried foods, grilled meat and seafood (barbecue), cholesterol-rich foods, sweet foods and caffeinated drinks. We named this pattern the ‘Western pattern’. Dietary pattern 2 was characterized by high consumption of green-yellow vegetables, seaweeds, healthy protein foods, bonefish, fruit and dairy products, and we named this pattern the ‘healthy pattern’. Dietary pattern 3 was characterized by high consumption of salted vegetables and seafood, cereals and light-coloured vegetables, and we named this pattern the ‘traditional pattern’. The major dietary patterns identified separately for men and women proved to be similar. Each of the three patterns explained 15·3 %, 13·0 % and 7·8 % of variance of food frequency consumption in men and 15·3 %, 12·2 % and 8·3 % in women, respectively.
Table 1 Factor-loading matrix for the major dietary patterns identified by factor analysis
For the sake of simplicity, factor loadings of <0·20 are not listed.
Demographic characteristics and lifestyle factors are presented as means (sd) or numbers and percentages stratified by quartiles of factor scores for each dietary pattern (Table 2). Among both men and women, participants with a higher healthy dietary pattern score tended to have a higher educational level and household income, to smoke less and to report more physical activity and dietary supplement use. Participants with a higher Western dietary pattern score were younger and more likely to smoke and drink, and men were more likely to have a higher BMI. Participants with a higher traditional dietary pattern score were more likely to have a higher educational level and household income and to be involved in regular leisure-time physical activity.
Table 2 Characteristics of study participants according to quartiles of factor scores for each pattern
*P for trend was calculated from a generalized linear model with adjustments for age for continuous variables, and a Mantel–Haenszel χ 2 test with adjustments for the age group for categorical variables.
†Age and BMI data are mean and standard deviation.
‡Unit is thousand won (
)/month.
§Hypertension was defined as the use of antihypertensive medication or a history of hypertension. Diabetes was defined as the use of any diabetes medication or a history of diabetes.
The results of multivariate logistic regression analyses of the association between dietary patterns and health behaviours such as smoking, alcohol consumption, regular exercise and dietary supplement use are presented in Table 3 for men and in Table 4 for women. Current smoking was positively associated with the Western dietary pattern (P for trend <0·001) and the traditional dietary pattern (P for trend = 0·002) in men. However, current smoking was inversely associated with the healthy dietary pattern in both genders (P for trend <0·001) and with the traditional dietary pattern (P for trend = 0·001) in women. Alcohol consumption was positively associated with all patterns in men, with the exception of light drinking, which was inversely associated with the Western dietary pattern (P for trend = 0·044). In women, heavy drinking was positively associated with the Western dietary pattern (P for trend = 0·007), and light drinking was positively associated with the traditional dietary pattern (P for trend = 0·001), while there was no association observed with the healthy pattern. Physical activity and dietary supplement use were positively associated with all patterns in both genders, with the exception of physical activity, which was inversely associated with the traditional dietary pattern in women.
Table 3 Dietary patterns according to lifestyle risk factors (smoking, alcohol consumption, physical activity and dietary supplement use status), by polytomous logistic regression analysis, in men
Ref., reference category.
*Adjusted for age.
†Non-drinkers (0 g/d); light drinkers (<12 g/d); and heavy drinkers (≥12 g/d).
Table 4 Dietary patterns according to lifestyle risk factors (smoking, alcohol consumption, physical activity, dietary supplement use status), by polytomous logistic regression analysis, in women
Ref., reference category.
*Adjusted for age.
†Non-drinkers (0 g/d); light drinkers (<12 g/d); and heavy drinkers (≥12 g/d).
Discussion
In the present study of Korean adults, we identified three major dietary patterns, Western, healthy and traditional, and we evaluated their relationship with health behaviours. The three dietary patterns identified in the present study were similar to the dietary patterns found by previous studies conducted among Asian and Western populations using factor analysis(Reference Bamia, Orfanos and Ferrari10, Reference Park, Murphy and Wilkens14, Reference Kim, Sasaki and Sasazuki15, Reference Yang, Kerver and Song22).
The Western pattern and the healthy pattern (or ‘prudent pattern’) have been reported by many other studies. The main contributors to the Western dietary pattern are typically meats, fats, fast food, sweets, grains, butter, eggs, potatoes and sugar-containing foods(Reference Terry, Hu and Hansen8, Reference Slattery, Boucher and Caan23, Reference Fung, Hu and Fuchs24). The main contributors to the healthy pattern are vegetables, fruit, fish and poultry(Reference Terry, Suzuki and Hu9, Reference Pryer, Cook and Shetty12, Reference Slattery, Boucher and Caan23–Reference Chen, Ward and Graubard26). Several studies have also reported a traditional dietary pattern(Reference Pryer, Cook and Shetty12, Reference Kim, Sasaki and Sasazuki15, Reference Shin, Oh and Park20, Reference Yang, Kerver and Song22, Reference van Dam, Grievink and Ocke27, Reference Engeset, Alsaker and Ciampi28). In a study of 1441 Korean children, in which thirty-three food groups were created and entered into a factor analysis, the traditional (Korean) pattern included vegetables, seaweeds, beans, fruits, milk and dairy products(Reference Shin, Oh and Park20). A study of the dietary pattern of 637 Korean Americans found that the traditional Korean pattern was characterized by high intake of traditional Korean dishes, such as soyabean paste stew, anchovies, Korean-style grains, tofu, vegetables, kimchi, salted/fermented fish, seaweeds, Korean-style soups, red meat, other seafood (shrimp, squid, clams, oyster, etc) and fish(Reference Yang, Kerver and Song22). In Japan, the traditional pattern includes pickled vegetables, salted fish and roe, fish, rice and miso soup(Reference Kim, Sasaki and Sasazuki15). In contrast, in the Netherlands, the traditional pattern is characterized by higher consumption of red meat, potatoes, highly saturated added fats, coffee and beer and lower consumption of soya products, low-fat dairy, breakfast cereals, tea and fruits(Reference van Dam, Grievink and Ocke27). A traditional Norwegian dietary pattern consists mainly of two or three similar cold meals, usually open sandwiches, and one hot meal with either fish or meat served with potatoes and vegetables(Reference Engeset, Alsaker and Ciampi28). The traditional dietary pattern in the present study included salted foods, such as pickled vegetables, kimchi and rice.
Studies of dietary patterns conducted in other populations have noted the importance of evaluating associations of dietary patterns with health behaviours, such as smoking, alcohol consumption, physical activity and dietary supplement use. For smoking, our study found that current smoking was strongly associated with the traditional and Western dietary patterns in men, but was inversely associated with the healthy pattern in both genders. In general, previously conducted studies have found that current smoking is inversely related to a healthy dietary pattern(Reference McNaughton, Mishra and Stephen11, Reference Pryer, Cook and Shetty12), whereas former smokers tend to have a healthier dietary pattern(Reference Knudsen, Rasmussen and Haraldsdottir29). For example, in a population-based Dutch study of 4244 men, men who smoked >20 cigarettes/d had significantly lower intake of β-carotene and especially ascorbic acid compared with men who never smoked, which was attributed to an almost 60 % lower fruit intake among smokers(Reference Zondervan, Ocke and Smit30). The European Prospective Investigation into Cancer and Nutrition (EPIC)–Potsdam study found that the ‘fruit and vegetable’ dietary pattern was negatively associated with smoking (β = −0·188 in men and −0·225 in women)(Reference Bamia, Orfanos and Ferrari10). The Monitoring of Trends and Determinants in Cardiovascular Disease (MONICA) population survey of 976 men aged 45–64 years also found a negative association of smoking status (P < 0·001) with a healthy dietary pattern(Reference Perrin, Dallongeville and Ducimetiere31). Finally, the Multiethnic Cohort Study of 195 298 participants residing in Hawaii and Los Angeles reported that current smokers showed a positive association with the fat and meat pattern (OR = 1·67, 95 % CI 1·62, 1·72) and inverse associations with the vegetable (OR = 0·66, 95 % CI 0·64, 0·68) and fruit and milk patterns (OR = 0·53, 95 % CI 0·52, 0·55)(Reference Park, Murphy and Wilkens14). Consistent with these findings, another study has shown that smokers are less likely than non-smokers to consume vegetables and milk or dairy foods. Smokers may consume more caffeinated or alcoholic beverages and meat products in order to enhance the taste of smoking(Reference McClernon, Westman and Rose32).
For alcohol consumption, we observed that heavy alcohol consumption was positively associated with the Western pattern in both genders, whereas light alcohol consumption was inversely associated with the Western pattern in men. Among 15 073 university graduates enrolled in the Seguimiento Universidad de Navarra study in Spain, higher adherence to the Western dietary pattern was less likely to decrease participants’ alcohol consumption during follow-up, whereas participants with higher adherence to the Mediterranean (healthy) dietary pattern were less likely to increase their alcohol consumption (OR = 0·66, 95 % CI 0·46, 0·95)(Reference Sanchez-Villegas, Toledo and Bes-Rastrollo16). In contrast, the Multiethnic Cohort Study in Hawaii found that the fat and meat dietary pattern was positively associated with BMI, smoking and alcohol consumption (≥1 drink/week; OR = 1·40, 95 % CI 1·37, 1·43)(Reference Park, Murphy and Wilkens14).
For physical activity and dietary supplement use, our study found both of these factors to be positively associated with all dietary patterns in both genders, with the exception of physical activity, which was inversely associated with the traditional dietary pattern in women. The EPIC–Potsdam study found that physical activity level was positively associated with the ‘fruit and vegetable’ dietary pattern in men (β = 0·182) and was negatively associated with the ‘bread and sausage’ pattern in women (β = 0·099)(Reference Bamia, Orfanos and Ferrari10). The Multiethnic Cohort Study found that physical activity (≥3 times/week) was positively associated with the vegetable (OR = 1·73, 95 % CI 1·69, 1·77) and fruit and milk patterns (OR = 1·44, 95 % CI 1·40, 1·47), but not with the fat and meat pattern (OR = 0·98, 95 % CI 0·96, 1·01)(Reference Park, Murphy and Wilkens14). In addition, participants with the healthy pattern have been reported to have higher physical activity levels and to be more likely to take dietary supplements daily(Reference van Dam, Grievink and Ocke27). Likewise, in our study, physical activity was more strongly associated with the healthy dietary pattern than other dietary patterns in men (OR = 2·16, 95 % CI 1·86, 2·53) and women (OR = 2·03, 95 % CI 1·70, 2·41). Among 64 252 women in the French E3N-EPIC cohort, supplement use was positively associated with the fruit/vegetable pattern and inversely associated with the processed meat/starchy foods and alcohol/meat products patterns (P for trend for all associations <0·001)(Reference Touvier, Niravong and Volatier33). In the German Nutrition Survey, comprising 7124 men and women, significant differences in food consumption between regular vitamin and mineral supplement users and non-users were observed, indicating a tendency towards a healthier diet among regular users of supplements(Reference Kerver, Yang and Bianchi19). In a study by McNaughton et al.(Reference McNaughton, Mishra and Stephen11), dietary supplement use was positively associated with the fruit/vegetable pattern and inversely associated with a Western dietary pattern, and dietary supplement users tended to have a healthier lifestyle and diet than non-users.
Most dietary pattern studies have been conducted in North America or Europe, and only a few studies have been conducted in Asia and Korea. Thus, the results of the present study are important as they reflect dietary habits common in Asian countries. It is important to note that our study participants were volunteers recruited at one cancer screening centre; therefore, the results may not be generalizable to other Korean populations. However, the relatively high socio-economic status of our study population may contribute to more accurate responses regarding risk factors and may have helped to reduce errors related to internal validity. Yet the present study has a limitation. The FFQ focused on eating habits rather than actual food intake amounts. The FFQ did not ask about portion size, and therefore information on nutrient intake was not available. In addition, the use of an FFQ containing only sixteen food groups may restrict the number of food categories used to characterize the usual dietary intake. It is not clear to what degree this FFQ represents habitual food intake.
In summary, the present study found three different dietary patterns in a middle-aged Korean population and indicates that these dietary patterns are strongly associated with health behaviours. The healthy dietary patterns were mainly associated with other healthy lifestyle behaviour factors, including not smoking, low alcohol consumption, participating in physical activity and dietary supplement use. The possible confounding effect of other risk behaviours should be appropriately considered when conducting nutritional epidemiological studies of the association between dietary patterns and disease outcomes.
Acknowledgements
The present study was supported by grants from the National Cancer Center, Korea (Grant numbers 0610552 and 0910222). There are no conflicts of interest. A.S. and J.K. designed the study; E.R.C. and S.-Y.L. conducted the statistical analysis; and E.R.C. wrote the manuscript. All authors critically reviewed and approved the manuscript.
Appendix
Food list for the sixteen food groups in the FFQ of the present study
