This study revealed differences in weight development between age 5 and 10 years between ethnic and SES groups in the large multi-ethnic ABCD cohort, consisting of 1 765 children, indicating that these groups follow different patterns of weight development. In Dutch and high SES children we observed a decreased weight development, while in non-Dutch and low/middle SES children weight development increased. Across the range of bodyweight categories at age 5, we observed a conversion to normal weight, which was stronger in Dutch and high SES children but less pronounced in non-Dutch and low/middle SES children. Dutch and high SES children were more often underweight or normal weight, at both ages and less often overweight/obese compared to non-Dutch and low/middle SES children. Contrary to our expectations, high healthy pattern scores were postitively associated with weight development in Dutch, Moroccan and all SES categories of children and high full-fat pattern scores were negatively associated with weight development in Dutch and high SES children.
In line with our results, other studies also showed ethnic [8, 32] and SES [24, 33] differences in overweight and obesity. Prevalence of overweight/obesity was nearly double in children of non-western origin compared to children of western origin [8, 32]. However in our study, we also observed clear differences in the prevalence of overweight/obesity within the non-native group. At age 10, 44% of the children from Turkish origin was overweight/obese while 30% of children of African Surinamese and 24% of Moroccan origin were overweight/obese. Ethnicity also carries information about SES. Maternal educational level (in our study used as a proxy for SES) explains part of the ethnic differences in childhood overweight [14]. Non-native groups are more often from lower SES groups than native groups. In our study, 23% of Dutch children was from the low/middle SES group, compared to 69% in African Surinamese, 87% in Turkish, 82% in Moroccan children and 43% in children of “other” ethnicities. Despite comparable SES levels, we clearly observed that Turkish children developed more often overweight/obesity between the age of 5 and 10 years (44%) than Moroccan children (24%). These results suggest that the non-native group is a diverse group with possibly different causes and mechanisms that contribute to childhood overweight/obesity in each ethnic group.
We observed a higher tendency to underweight and normal weight in Dutch and high SES children. Another Dutch study also observed higher percentages of underweight in Dutch children, while percentages of underweight in children of Turkish and Moroccan origin were at most only half as high [34]. But unlike our results, this study did not observe SES differences in underweight rates [34]. A study in 4-5 year old children in Norway, found comparable percentages of underweight between children of European (10.4%) and Middle East/North African origin (12.7%) [32]. These differences could be possibly due to unintended side effects of current obesity prevention programs, and current delivery of these programs to lower SES groups is either not adequately reaching into these groups [35].
Aditionally we examined weight development per ethnic and SES groep, stratified by BMI category at age 5. We observed a conversion to normal weight in most groups, which is not in line with a study in 6-12 year old Norwegian children, where they found that overweight children tended to gain more weight than normal-weight children [24]. But the conversion to normal weight differed per ethnic and SES group. It was stronger in Dutch and high SES children that were more likely to experience negative weight development, regardless of BMI category at age 5. Observing weight development stratified per BMI category at age 5, showed that none of the children switched from underweight to overweight/obesity or from overweight/obesity to thinnes. We also obtained more specific information of the differences in weight development between the non-native groups. Weight development of Moroccan children, seemed to be more in line with the structure of weight development in Dutch and high SES children. In normal weight Dutch and high SES children, weight development was negative while in normal weight African Surinamese, Turkish, “other” ethnicities and low/middle SES children, weight development was positive. In normal weight Moroccan children, weight decreased; however results were not statistically significant. In overweight/obese children, we observed a further increase of weight in Turkish children and children of “other” ethnicities.
Our results indicate that a healthy dietary pattern at age 5 is postively associated with weight development whereas a full-fat dietary pattern at age 5 is negatively associated weight development. Oellingrath et al. [23] also observed a positive association between dietary patterns identified as healthy (a ‘varied Norwegian’ pattern characterized by intakes of fish and meat for dinner, brown bread, regular white or brown cheese, lean meat, fish spread, and fruit and vegetable) and BMI. In this study, 9 year old children with high scores on this 'varied Norwegian' pattern were more likely to be overweight or obese at 9 years of age [23]. Another Dutch birth cohort observed that better diet quality at age 1 and 8 years was associated with higher height, weight, and FFMI, but not with body fatness up to age 10 years. This was independent of diet quality at an earlier or later time point [36]. We did not find other studies in the literature that observed an association between a dietary pattern comparable to our full-fat pattern (high intakes of full-fat cheese and full-fat spreads, low intakes of low-fat cheese, low-fat spreads and low-fat dairy) and BMI. In line with other studies [23, 37, 38, 39, 40], we did not observe a clear association between the snacking pattern (a dietary pattern characterized by snacking items) and weight development. Overall, the observed patterns of association between dietary patterns and weight development were mixed, also after analyses were stratified by BMI category at age 5. We observed only a few statistically significant associations, mainly within the larger groups (e.g. Dutch, low/middle SES and high SES). Therefore based on these results, we cannot draw any conslusions about the role of dietary patterns at the age of 5 in weight development over the next 5 years.
Childhood obesity is a complex construct. A diversity of factors contribute as biological- and lifestyle factors but also the social- and obesogenic environment [41]. There is also interest in the role of dietary patterns in early childhood as a determinant of obesity risk [18]. Few studies observed associations between dietary patterns and overweight [22, 23, 24]. In our analyses we considered sex, children’s exact age at the 5-years health check, screen time at age 5 and maternal BMI at age 5 as major determinants that might influence the association between dietary patterns at age 5 and weight development.
Additionally, we studied the role of energy intake in the association between dietary patterns at age 5 and weight development. A previous study by Northstone et al [42] also failed to observe differences in outcome when adjusting for energy intake before entry into the PCA analysis. However, to exclude a possible role of energy intake, we included energy adjusted intake of food groups as an input in the PCA analyses to derive dietary patterns [43]. In sensitivity analyses, we also observed that additionally adjusting for energy intake in the association between dietary patterns at age 5 and weight development did not change the results. In the same Model, we observed that energy intake was not associated with weight development in the ethnic and SES groups.
Results of the association between dietary patterns and weight development stratified by ethnic and SES group, also give information about the possible moderating effect of ethnicity and SES. In another sensitivity analysis, we studied the intermediate effect of dietary patterns on the association with ethnicity / SES on weight development and we observed that the intermediate role of dietary patterns on weight development is limited. However, in our cohort, results were more pronounced in the ethnic groups than in the SES groups.
Methodological considerations
A strength of this study is the population based cohort-design that included a number of 1 765 children with complete data on dietary patterns at age 5 and BMI at age 5 and 10 years, of different ethnic and SES groups that are often excluded in epidemiological studies. Analyses stratified per BMI category gave an complete overview of weight development in these groups. Height and weight were measured by a team of trained researchers and health professionals according to standard protocols.
However, interpretation of the association between dietary patterns and weight development is limited due to the cross sectional nature of dietary intake assessment. Dietary tracking, the maintenance of a dietary pattern over a certain time period, has been observed during childhood and from childhood to adolescence [40]. Another limitation of our study is that the FFQ was based on food commonly consumed by the Dutch population as determined by the Food Consumption Survey 1997–1998 [44]. The FFQ was validated with the gold standard of doubly labelled water in a group of 4 to 6-year-old children, although this validation study did not include non-Dutch groups [45]. Thus the FFQ may not reflect the food intake of ethnic minority children. To account for this we included an open question at the end of the FFQ, asking participants to note commonly eaten foods that were missing from the questionnaire. However, in the analysis we decided to not include the food items mentioned in the open item section because it was used only by a few mothers and it was unclear whether the mentioned items were additional to what was already included in the FFQ. In sub-analysis we found that energy intake related to energy requirements (based on Schofield resting metabolism) was not different between Dutch and non-Dutch groups, implying that the FFQ is able to capture total food intake in this population. Smaller numbers in ethnic groups is inherent to the ABCD cohort design but it is possible that some biases may have been introduced into the analyses, particularly as the non-responders tended to come disproportionately from lower SES and ethnic minority groups. Response rates per ethnic and SES group were 27% for Dutch, 12% for African Surinamese, 10% for Turkish, 13% for Moroccan, 13% for other ethnicities, 11% for low/middle SES and 38% for high SES. A nonresponse analysis determining the degree of selective response and selection bias between pregnancy and birth outcomes, indicated that selective non-response was present in the ABCD cohort, but selection bias was acceptably low and did not influence the studied birth outcomes [46].