Problematic child behaviour has become a critical issue in Korea, where at least one-third of children have been reported to have some type of behavioural problem(1). According to nationwide studies of 75 643 adolescents in 2011, 42 % suffered from constant depression for more than 2 weeks during the last 12 months and 33 % of them suffered from severe stress. The prevalence of behavioural problems in Korean children is approximately 10 % greater compared with their counterparts in the USA(1, Reference Eaton, Kann and Kinchen2).
What a person chooses to eat and drink influences the electrochemical activity of the brain and has an impact on that person's behaviour and cognitive functions(Reference Rosales, Reznick and Zeisei3–Reference Benton6). Early childhood is a time of rapid and dramatic brain development, especially during the pre-school years, when there is a transition from the maternal-selected diet to food more based on the child's selection(Reference Rosales, Reznick and Zeisei3, Reference Georgieff4). The impact of diet on the brain has strong and long-lasting effects for children in early childhood(Reference Rosales, Reznick and Zeisei3, Reference Georgieff4).
Poor nutrition, determined by levels of certain foods or nutrients such as animal products, sugar, glucose and micronutrients (i.e. Fe and thiamin), has an influence on cognition and behaviour in children, although the results are not consistent(Reference Rosales, Reznick and Zeisei3, Reference Bourre5, Reference Brown and Pollitt7–Reference Neumann, Bwibo and Murphy12). Obesity has also been linked to behavioural problems. However, the pattern, degree and direction of these associations are not clear(Reference Datar and Sturm13–Reference Rucklidge17).
One reason for this inconsistency could be significant associations partly due to chance inter-correlations among many components in the diet. Considering this issue, dietary patterns draw our attention because they provide an overall view of the diet that is not observed when evaluating individual nutrients from foods. The roles of a ‘processed’ and a ‘Western’ diet (with high fat and sugar contents) and a micronutrient-rich ‘prudent’ diet have been reported with respect to physical and mental health in children(Reference Howard, Robinson and Smith18–Reference Feinstein, Sabates and Sorhaindo23). Investigating these dietary patterns further has important implications in public health because this dietary intake is visible and can subsequently be modified. The role of dietary patterns has been suggested in relation to obesity in young children(Reference Shin, Oh and Park24); however, the association between the Korean dietary pattern and behavioural problems in children of pre-school age has not been reported.
Methods
Participants and design
The Practical Approach for Better Maternal and Child Nutrition and Health Study recruited eleven urban pre-schools through multiple-step samplings. From the selected pre-schools, we chose 1724 children whose caregivers agreed to participate in our study from June 2001 to June 2005(Reference Shin, Oh and Park24). For the analyses of the present study, 1458 children with intake variables were selected for identifying dietary patterns. Among them, 1347 children who had complete records of dietary intake, weight, height and behavioural variables were included in the analyses with relation to behaviour.
Dietary intake
Dietary intake for each child was assessed by a validated semi-quantitative FFQ, including 100 food items. These food items were placed into nine non-overlapping categories according to the frequency of consumption (ranging from ‘rarely eaten’ to ‘more than three times per day’ during the preceding year) and portion size (small, average or large)(Reference Shin, Oh and Park24). The amount of each food item in the FFQ was converted into grams, from which we calculated the daily intakes of nutrients.
Child social-behavioural outcomes
A Korean version of the seventy-six-item Preschool and Kindergarten Behavior Scale (PKBS) was used to measure the problem behaviour and social skills of children(Reference Merrell25, Reference Chun26). The items on the PKBS were composed of two separate scales, each designed to measure separate domains: a thirty-four-item social skills scale and a forty-two-item problematic behaviour scale. Each of these two scales consisted of subscale structures. The social skills scale was composed of the following subscales: social cooperation (twelve items), social interaction (eleven items) and social independence (eleven items). The problem behaviour scale included two broad-band subscales: internalizing behaviours and externalizing behaviours. The internalizing broad-band scale consisted of two narrow-band scales, namely social withdrawal (seven items) and anxiety/somatic problems (eight items); and the externalizing broad-band scale was composed of three narrow-band scales, namely self-centred/explosive (eleven items), attention problems/overactive (eight items) and anti-social/aggressive (eight items). A four-point Likert scale, ranging from 1 for ‘never’ to 4 for ‘very often’, was used to measure each item. Correlation coefficients, for internal consistency and interrelations of PKBS subscales for Korean children, ranged from 0·62 to 0·97(Reference Chun26). In addition to the social skills scale, we used the problem behaviour scale by combining internalizing and externalizing scales due to a high correlation between these two behavioural variables (r = 0·66).
Based on our data, we defined the 15th and 85th percentiles as respective cut-off points for the behavioural scores from a statistical point of view, because these points are close to −1 sd and +1 sd. Children with social skills scores lower than the 15th percentile were considered to have a potential risk of poor social skills. For the problem behaviour scale, we identified children receiving scores greater than the 85th percentile as having a potential risk of this behavioural problem. The pre-school teacher responsible for each child completed the PKBS.
Assessment of other variables
The BMI of each child was calculated using his or her height and weight measures assessed by trained graduate students majoring in nutrition. Underweight and overweight status were defined respectively by the 5th and 85th percentiles of BMI for specific age and gender groups based on the Korean paediatric growth standards(27). Other variables included in data analyses were the child's age, gender and daily nutrient intakes obtained by FFQ assessment, household monthly income (in one of three categories; 1 = less than ≈$US 1700, 2 = ≈$US 1700 to less than ≈$US 4300, 3 = greater than ≈$US 4300), mother's occupational status (1 = simple work, 2 = sales, 3 = office work, 4 = administrative position, 5 = professional, 6 = no job) and pre-school region, as described elsewhere(Reference Shin, Oh and Park24). We divided the maternal occupation variable into two groups (yes/no) and used it in data analyses.
Statistical analysis
In the present study, the dependent variables were the child's behaviours, including social skills and problem behaviour, and the major independent variables were the dietary patterns and weight status. Dietary patterns were identified based on thirty-three foods/food groups from the 100 food variables, adjusting energy intake by the residual method. Using factor analysis with the food/food group intake frequency variables, we developed the ‘Korean healthy’, ‘animal foods’ and ‘sweets’ dietary patterns. We used factor analysis because it has been suggested as a reliable method for dietary pattern analysis and is more commonly used in related studies(Reference Newby and Tucker28, Reference Hearty and Gibney29).
The ‘Korean healthy’ dietary pattern indicated a relatively higher intake of vegetables, seaweed, beans, dried and fresh fish, kimchi, potatoes and nuts (Table 1). The ‘animal foods’ indicated a higher intake of fast foods, organ meat, poultry, pork, beef, processed fish cake, noodles and ramyeon. The ‘sweets’ pattern showed a higher intake of sugary foods, chocolate, sweet baked goods, sweet drinks, bread and ice cream (but a lower intake of rice). In each dietary pattern, we divided the children into four groups according to their dietary pattern scores.
Table 1 Factor-loading matrix for the three dietary patterns and their foods or food groups identified in 1458 Korean pre-school children using intake frequency values with adjustment of energy intake, Practical Approach for Better Maternal and Child Nutrition and Health Study, 2001–2005
Cumulative eiganvalue for the three dietary patterns was 21·8 %.
The associations of background characteristics and nutrients with the dietary factor and behaviour scores were evaluated by Pearson correlations and univariate logistic analysis. Multiple logistic regression analyses were used to evaluate the associations of dietary patterns and weight status with behavioural outcomes while minimizing a possible role of family and child characteristics in these associations(Reference Cody and Smith30). Potential confounding variables included pre-school (eleven schools), household monthly income, maternal occupation and the child's age, gender and BMI (only for dietary pattern variables). Except for the child's age, all confounding variables were categorical variables. We included total energy intake as a confounder in analysing the association between weight status and behaviour. Odds ratios and their 95 % confidence intervals were computed to assess the strength of the associations(Reference Cody and Smith30). We performed data analysis using the statistical software package SAS version 9·3.
Results
The mean age of the children was 5·2 years and the proportions of boys and girls were similar (Table 2). The respective percentages of underweight and overweight children were 18·5 % and 16·0 % which indicated a similar level of overweight, but a higher level of underweight in our children, because we used 5th and 85th percentile values based on the Korean child growth standards as cut-off points for underweight and overweight status, respectively (Table 2). About 70 % of the children were from middle-class households and about half of the children had working mothers. Regarding nutrients, 60–95 % of the children met the Korean Dietary Reference Intake for micronutrients(31). Mean carbohydrate, protein and fat intakes were comparable to the Korean Dietary Reference Intakes. Mean scores for social skills and problem behaviour were similar to those found in another study on Korean children who were selected from the general population(Reference Lee32) (Table 2).
Table 2 Background information of the Korean pre-school children, Practical Approach for Better Maternal and Child Nutrition and Health Study, 2001–2005
KDRI, Korean Dietary Reference Intake; RE, retinol equivalents.
†References are the EER (estimated energy requirement) for energy and the EAR (estimated average requirement) for other nutrients based on the KDRI.
In bivariate analyses, the ‘Korean healthy’ pattern was positively associated with protein and micronutrients such as Fe, β-carotene, folate and vitamins A, C and E (Table 3). The ‘animal foods’ pattern was relevant to a greater intake of protein and showed an inverse, but weaker relationship with antioxidant nutrients compared with the ‘Korean healthy’ pattern. The ‘sweets’ pattern had a positive relationship with fat intake and negative associations with protein and Fe intakes. Intake of carbohydrate was negatively associated with the dietary patterns because we used energy-adjusted intake variables to identify the dietary patterns. The risk of ‘Korean healthy’ diet intake was lower in girls than boys and lower among those who were underweight (Table 3). The ‘animal foods’ pattern was associated positively with age.
Table 3 Univariate associations of dietary patterns and behavioural problems with background information variables among Korean pre-school children, Practical Approach for Better Maternal and Child Nutrition and Health Study, 2001–2005
*P < 0·05, **P < 0·01 for Pearson correlation coefficients using continuous variables.
†Dietary pattern variables were divided into two groups based on median values.
‡Behavioural variables were divided into poor (15 % of the children) and normal (85 % of the children) groups.
Behaviour variables were not associated with nutrient intakes in bivariate analysis (Table 3). The two behavioural problems were lower in older children and showed a lower risk in girls than boys. The poor social skills scale was associated positively with being underweight while the problem behaviour scale showed the opposite relationships. A higher risk of behavioural problems was found in the children who had working mothers (Table 3).
In multivariate analyses, children were divided into four groups (Q1–Q4) for each dietary pattern and two groups (poor v. normal behaviour) for each behaviour variable. The ‘Korean healthy’ dietary pattern showed a significant and inverse association with poor social skills in the second highest group (OR = 0·42; 95 % CI 0·21, 0·82) compared with the lowest intake group for boys (Table 4). For girls, the ‘sweets’ pattern was positively associated with a greater risk of poor social skills (OR = 3·41; 95 % CI 1·29, 9·01 at Q4 v. Q1) and problem behaviour (OR = 2·80; 95 % CI 1·05, 7·43 at Q4 v. Q1). The ‘animal food’ pattern had no association with behavioural problems.
Table 4 Association between dietary patterns and behavioural problems in Korean pre-school children by gender using multivariate logistic analysis, Practical Approach for Better Maternal and Child Nutrition and Health Study, 2001–2005
ref., referent category.
Respective numbers of children belonging to the low v. high groups of the poor social skills and problem behaviour scores were 475 v. 112 and 453 v. 132 for boys and 504 v. 56 and 511 v. 47 for girls.
Adjusted for household income, maternal occupation, pre-school, child's age, sex and BMI, and total energy intake.
Both overweight and underweight boys had a higher risk of poorer social skills than normal-weight boys (Table 5). The problem behaviour scores did not differ according to weight status for both genders.
Table 5 Association between weight status and behavioural problems in Korean pre-school children by gender using multivariate logistic analysis, Practical Approach for Better Maternal and Child Nutrition and Health Study, 2001–2005
ref., referent category.
Respective numbers of children belonging to the low v. high groups of the poor social skills and problem behaviour scores were 532 v. 129 and 518 v. 141 for boys and 553 v. 60 and 561 v. 52 for girls.
Adjusted for household income, maternal occupation, pre-school, child's age and gender, and total energy intake.
Discussion
We found that the ‘Korean healthy’ pattern was associated with pro-social behaviours in boys, while the ‘sweets’ pattern was associated with anti-social behaviours in girls. For weight status, both underweight and overweight boys had a higher risk of poor social skills than normal-weight boys.
Our findings support the results of other studies on Western children suggesting a considerable role of dietary patterns on behaviour in children(Reference Wiles, Northstone and Emmett15, Reference Howard, Robinson and Smith18–Reference Feinstein, Sabates and Sorhaindo23). In a prospective cohort study, a poor diet, similar to the ‘sweets’ dietary pattern in the present study, at 3 years of age was negatively related to intelligence quotient at 8·5 years of age, while the opposite association was found for a better diet similar to our ‘Korean healthy’ dietary pattern(Reference Northstone, Joinson and Emmett20). The magnitude of this association was greater in the poor diet than in the better diet(Reference Northstone, Joinson and Emmett20). In other studies, children with a higher intake of the Western diet, a combination of the ‘animal foods’ and ‘sweets’ dietary patterns, also had a greater risk of anti-social behaviours(Reference Wiles, Northstone and Emmett15, Reference Howard, Robinson and Smith18, Reference Oddy, Robinson and Ambrosini22).
One potential reason for the association between diet and behaviour is the nutrient content of the diet. The ‘Korean healthy’ pattern is an antioxidant-rich diet that includes vitamins A, C and E, β-carotene and folate (Table 2). Antioxidant nutrients protect against oxidative stress and cell damage from free radicals, preserving neural functions(Reference Rosales, Reznick and Zeisei3–Reference Bourre5). Similar to antioxidant nutrients, folic acid is required for neurotransmitters(Reference Rosales, Reznick and Zeisei3, Reference Gewa, Weiss and Bwibo10). Maternal folate status during early pregnancy has been reported to have an inverse association with emotional problems in the offspring in a large-scale cohort study(Reference Steenweg-de Graaff, Roza and Steegers33).
Another important finding is a negative role of the ‘sweets’ pattern in behavioural problems for girls. Sweet foods are relevant to sucrose that could produce a rapid rise of blood glucose, which in turn induces a rapid fall to a level that disrupts brain function(Reference Benton, Brett and Brain11, Reference Benton34). Less sugar from fruit snacks and/or lower vitamin C intake has been reported to have a positive association with the risk for attention-deficit hyperactivity disorder in school-aged Korean children(Reference Kim and Chang35), although the adverse influence of sugar is not clear(Reference Benton34).
Glucose may need a shorter period of time to respond to human body needs than micronutrients(Reference Benton34). The ‘Korean healthy’ pattern related to micronutrients suggests a relative long-term effect of nutrition on health compared with the ‘sweets’ pattern associated with glucose. These interpretations imply that chronic rather than acute response to nutrition would be meaningful in behaviour for boys while the opposite response would be important for girls according to the context of the Korean diet. However, this suggestion is limited because a simple sugars intake variable was not available in our data.
We found a ‘U’ shape in the relationship between weight status and behaviour. Besides obesity, being underweight presented a higher level of poor social behaviour in boys. Accumulating evidence suggests an important role of obesity on problematic behaviours, although the results are not consistent in terms of the direction of the association and/or gender(Reference Rosales, Reznick and Zeisei3, Reference Datar and Sturm13, Reference Anderson, He and Schoppe-Sullivan14, Reference Datar, Sturm and Magnabosco16). One of the reasons could be that being underweight limits exposure to the external environment and induces lethargy in young children, which could then weaken their pro-social behaviour(Reference Perren, Stadelmann and von Wyl36).
A gender difference in the association between weight and behaviour requires attention. Previous studies have suggested a gender difference in adolescents but not children of pre-school age(Reference Perren, Stadelmann and von Wyl36). To our knowledge, our study is the first one to show the importance of weight status and gender in relation to behaviour in young children.
The present study has both limitations and strengths. The Korean version of the PKBS we used is not a diagnostic instrument. We assessed children with behavioural problems based on a statistical point of view rather than clinical considerations. It was a cross-sectional study of free-living children, thus evidence linking dietary pattern and weight status with behaviour may be inconclusive because unmeasured or unexamined variables (such as environmental pollution, parenting style, physical activities and genetic factors) could partly explain the outcome(Reference Kim, Cho and Kim37).
One of the strengths of the present study is that the examination of dietary patterns can overcome the problem of inter-correlations from many dietary components. Furthermore, the study is the first to present the importance of a Korean-style healthy diet on the behaviour of young children. The associations between the type of behaviour (e.g. pro- v. anti-social behaviours) and dietary patterns (healthy v. unhealthy diets) are meaningful because of a lack of related research for young children. Further research, such as a cohort study, may be needed to confirm our findings.
Conclusion
In conclusion, our analyses suggest that dietary patterns and weight status are important for the behaviour of young children. Healthy and unhealthy dietary patterns, underweight and overweight status, and gender differences should be considered for further studies.
Acknowledgements
Sources of funding: This work was supported by a research grant (#911047-1) from the projects for the ‘Globalization of Korean Foods’ and the Brain Korea 21 project. The funder, the Korea Institute of Planning & Evaluation for Technology in Food, Agriculture & Fisheries, had no role in the design, analysis or writing of this article. Conflicts of interest: The authors declare no conflicts of interest. Ethics: We used secondary data based on published research, thus ethical approval was not required. Authors’ contributions: S.-Y.O. designed the research. S.-Y.O., N.C. and M.-H.K. contributed to obtaining funding and conducted the research. S.-Y.O. and H.A. analysed the data. S.-Y.O. and J.V.O. wrote the manuscript. S.-Y.O. had primary responsibility for the final content. All authors read and approved the final version of the paper.
