obesity reviews
doi: 10.1111/obr.12325
Supplement Article
Fat, sugar and water intakes among families from the IDEFICS intervention and control groups: first observations from I.Family L. Arvidsson1†, L.-H. Bogl2†, G. Eiben1, A. Hebestreit3, P. Nagy4, M. Tornaritis5, L. A. Moreno6, A. Siani7, T. Veidebaum8, S. De Henauw9, L. Lissner1, on behalf of the IDEFICS and I.Family consortia
1
Section for Epidemiology and Social Medicine
Summary
(EPSO), Department of Public Health and
Background: The objective of this paper is to investigate differences in diets of
Community Medicine, Institute of Medicine,
2015
families in intervention versus control communities 5 years after the Identification and Prevention of Dietary- and Lifestyle-Induced Health Effects in Children and Infants intervention ended. Methods: Altogether, 4,691 families from the I.Family study with at least one participating parent and one child are included in this analysis. Diet quality indicators, defined as propensities to consume fat, sugar, water and fruit and vegetables, are calculated from a 59-item food frequency questionnaire. Multilevel linear models with random intercepts for study centre are used to determine whether mean diet indicators, calculated at the family level, differed as a function of previous exposure to the intervention. Results: Families in the intervention communities reported a significantly lower sugar propensity (19.8% vs. 20.7% of total food items, p < 0.01) and a higher water propensity (47.3% vs. 46.0% of total beverages, p < 0.05) compared with families in the control communities, while fat and fruit and vegetables propensities were similar. No significant diet differences between intervention and control children were present at the Identification and Prevention of Dietary- and Lifestyle-Induced Health Effects in Children and Infants baseline. Discussion: This result indicates better diet quality in intervention families, which was not present in children when their diets were assessed before the intervention, and gives some cause for optimism regarding the sustainability of some aspects of the diet intervention.
Address for correspondence: Louise
Keywords: Diet intervention, family diet, I.Family study, IDEFICS study.
Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden, 2
Department of Public Health, University of
Helsinki, Helsinki, Finland, 3 Leibniz Institute for Prevention Research and Epidemiology – BIPS, Bremen, Germany, 4 Department of Pediatrics, University of Pécs, Pécs, Hungary, 5 Research and Education Institute of Child Health, Strovolos, Cyprus, 6 GENUD (Growth, Exercise, Nutrition, and Development) research group, University of Zaragoza, Zaragoza, Spain, 7
Epidemiology & Population Genetics, Institute
of Food Sciences, CNR, Avellino, Italy, 8
Research Centre, National Institute for Health
Development, Tallinn, Estonia, and 9
Department of Public Health, University of
Ghent, Ghent, Belgium Received 31 March 2015; accepted 30 August
Arvidsson, University of Gothenburg, Section for Epidemiology and Social Medicine (EPSO) Box 453, S-405 30 Göteborg, Sweden. E-mail:
[email protected]
Abbreviations: body mass index (BMI); Children’s Eating Habits questionnaire (CEHQ); food frequency questionnaire (FFQ); Identification and prevention of Dietary- and lifestyle-induced health EFfects In Children and infantS (IDEFICS).
†
L. Arvidsson, L.-H. Bogl share first authorship.
obesity reviews (2015) 16 (Suppl. 2), 127–137
Introduction Children’s eating habits, whether healthy or unhealthy, are likely to track into adult life (1--3), underscoring the importance of introducing healthy diets early in life. The ecological © 2015 World Obesity
systems theory includes the family environment as an important determinant of child health (4). Hence, a healthpromoting family environment could be of great value for preventing obesity in children (5,6). Eating habits are shaped 127 16 (Suppl. 2), 127–137, December 2015
obesity reviews
128 IDEFICS intervention and diet in families L. Arvidsson et al.
by a combination of genetic and environmental factors where the family food environment serves as a model for children’s intake (7). Important determinants of children’s eating habits are child-feeding practices, food availability and parental role modelling (8--12). By involving parents in interventions, i.e. parent’s presence and participation in nutrition and behaviour training sessions, children’s eating habits and overall health can be improved (13). The community-based Identification and Prevention of Dietary- and Lifestyle-Induced Health Effects in Children and Infants (IDEFICS) intervention was carried out at community, school, family and individual levels (14,15). The diet intervention aimed to improve dietary habits by increasing daily consumption of water, fruits and vegetables (16) and thereby decreasing intake of added sugars. Educational materials were provided to parents in order to improve behavioural skills, social support, accessibility and availability of water, fruit and vegetables at home and to increase awareness and self-efficacy in parents. The educational information was developed and provided at the community level, with the aim to help the parents promote and encourage a healthy food environment (14). More details on the intervention design can be found in Pigeot et al. (17). Findings presented elsewhere in this supplement show no short-term effects of the IDEFICS intervention on changes in children’s dietary intake (18). However, this was not examined on a family level. Given the importance of family environment on children’s eating habits, the present paper aims at investigating the diet quality of the families 5 years after the intervention ended by comparing diet quality in families who were in the IDEFICS intervention or control groups. In addition, we aimed to investigate parents and children separately regarding diet quality and to compare diet quality in families from different educational strata. Finally, we compared baseline diets in children before the intervention in both groups.
children from the original IDEFICS cohort ( from here on referred to as ‘index’ children) participated. At this time, siblings and parents or other caregivers (less than 1% were other caregivers, e.g. grandparents) of the index child were also invited to participate such that additional 2,429 siblings and 7,794 parents were examined. Ethics approval was obtained from responsible committees in each country. Parents and children older than 16 years provided written informed consent. Younger children gave oral consent for examinations and sample collection. The present study included 4,691 families (i.e. at least one parent and one child), whereof 2,548 (54%) had at least one child exposed to the intervention. These families were composed of 7,739 index children and siblings and 6,631 parents. The sample for the baseline analysis included 4,914 index children who participated in I.Family and met our inclusion criteria of at least one participating parent (or other caretaker) in I.Family, information on parental education and valid dietary data at both time points.
Socio-demographic information Highest parental educational achievements, household size and children’s sex and age were reported by one of the parents (or other caretakers). Household size was obtained from an interview. Data on parental education were attained from IDEFICS baseline parental questionnaires. However, some children joined the study later, and for these children, educational data were collected from the followup measures. Education level was based on the International Standard Classification of Education for cross-country comparability (20) and used to determine the maximum of the parents’ education, a proxy for socioeconomic status. Levels 1–3 represent upper secondary school and are classified as lower education level while levels 4–6 represent postsecondary education and are classified as higher education level.
Methods Weight status Participants The present paper includes participants from the community-based IDEFICS study (16,19) and the followup study I.Family. Both studies examine the same cohort of children, recruited in eight countries in Europe (Belgium, Cyprus, Estonia, Germany, Hungary, Italy, Spain and Sweden), with the general purpose to understand how to prevent overweight and obesity in children and to identify determinants of eating habits, lifestyle choices and health in European children and their families. From September 2007 to May 2008 (the baseline time point), 16,228 children (2 to 9.9 years old) were included in the baseline survey of the IDEFICS study. In 2013, I.Family started, and 7,126 16 (Suppl. 2), 127–137, December 2015
Weight of all participants was measured to the nearest 0.1 kg by a Tanita BC 420 SMA scale (TANITA, Tokyo, Japan), and height was measured to the nearest 0.1 cm by a SECA 225 Stadiometer (Seca GmbH & Co. KG., Hamburg, Germany). Measurements were performed in the morning, when the children and parents were fasting and wearing only light clothes. Age-specific and sex-specific body mass index (BMI) and BMI z-scores for children and adolescents developed by the International Obesity Task Force (21,22) were calculated and categorized as normal weight (including thin) and overweight (including obese). BMI of the parents was categorized according to the definition of the World Health Organisation where © 2015 World Obesity
obesity reviews
BMI < 25 was classified as normal weight and BMI ≥ 25 as overweight.
Diet quality: sugar, fat and water propensity ratios For the evaluation of the IDEFICS intervention with respect to diet quality of the family 5 years later, three indicators ref lecting habitual diet during the last 4 weeks, namely propensity to consume fat, sugar and water, were used. Information was obtained from a 59-item food frequency questionnaire (FFQ) included in the Children’s Eating Habits Questionnaire (CEHQ). The CEHQ-FFQ describes consumption frequency of 59 foods and beverages on a typical week during the preceding 4 weeks by the following question: In the last month, how many times did you eat or drink the following food items? Possible response categories were never/less than once a week, 1–3 times a week, 4–6 times a week, 1 time per day, 2 times per day, 3 times per day or 4 or more times per day. Parents and teens selfreported usual consumption frequency of all foods and beverages. For children, parents or other caretakers reported foods and beverages consumed at home, i.e. excluding meals eaten at schools and kindergartens/pre-schools. To maintain comparability across countries, the same foods and beverages were translated into eight languages, but for some food items, country-specific foods were also noted as examples. Furthermore, the questionnaire was developed in English and translated into local languages. The quality of translations was checked by back translation. The CEHQ-FFQ has been found to be reproducible with mean kappa coefficients ranging from 0.41 to 0.60 and Spearman’s correlation higher than 0.5 for 81% of the food items (23), and a validation study of the CEHQ-FFQ against repeated 24-h dietary recall found that under 12% of the food groups were misclassified (24). The sugar and fat propensity ratios have previously been used to describe eating habits in children (25,26), and they have been found to correlate with the percentage of sugar and fat intake reported by a 24-h dietary recall (26). A preference for fat in taste tests was associated with a higher propensity to consume foods rich in fat, but this was not seen for sweet foods (25). Yet-unpublished results showed that a higher propensity to consume sugar was associated with higher odds of having higher counts of cariogenic microorganisms in Swedish children (27). Furthermore, high propensity to consume fat and sugar was associated with eating while watching TV, as well as with watching TV more than 60 min per day (26). The sugar and fat propensity ratios were calculated from the CEHQ-FFQ using the principle formula of ‘total intake frequency of sugar or fatty food items a week/total intake frequency of all food items a week’. Hence, the propensity ratios provide a description of the diet quality avoiding misclassification of participants as high-sugar or high-fat © 2015 World Obesity
IDEFICS intervention and diet in families L. Arvidsson et al.
129
consumers simply because they consume all types of food more frequently. It also corrects for variation in the number of meals and eating occasions away from home and thus not captured for children in the CEHQ-FFQ. For the present analysis, we excluded participants for whom more than 29 (49%) of the CEHQ-FFQ items were missing (28), resulting in a reduction of 2,636 (15%) participants from our original sample of 17,349. When creating composite scores, missing food items were treated as not consumed (28). The following individual food items were included in the fat propensity ratio: fried potatoes, whole fat milk, whole fat yoghurt, fried fish, cold cuts/sausages, fried meat, fried poultry, fried eggs, mayonnaise and mayonnaise based products, cheese, chocolate-based or nut-based spread, butter/margarine on bread, oil, nuts and seeds, salty snacks, savoury pastries, chocolate-based candies, cake/pudding/ cookies and ice cream. The sugar propensity ratio included the following: fresh fruit with added sugar, fruit juices, carbonated sugar-sweetened drinks, sugar-sweetened drinks not carbonated, sweetened coffee, sweetened tea, sweetened or sugar-added breakfast cereals, sweetened and/or f lavoured milk, sweetened and/or f lavoured yoghurt, sweet spreadable (jam, honey, chocolate or nut-based) and snacks (chocolate-based candies, non-fat candies, cake/pudding/ cookies or ice cream). Propensity to consume water was also calculated from the CEHQ-FFQ using the formula of ‘intake frequency of water a week/total intake frequency of all beverages a week’. The denominator included the following 12 beverages: water, fruit juices, carbonated sugar-sweetened drinks, sugarsweetened drinks not carbonated, diet carbonated drinks, artificially sweetened drinks not carbonated, coffee unsweetened and sweetened, tea unsweetened and sweetened and milk unsweetened and sweetened. Participants for whom more than six (50%) of the beverages were missing were excluded from the analysis. When creating composite scores, missing beverages were treated as not consumed. As an additional analysis, a propensity ratio for consuming fruits and vegetables was calculated from the CEHQFFQ using the formula of ‘intake frequency of fruits and vegetables per week/total intake frequency of all food items per week’. It included the following six items: legumes, boiled potatoes, other boiled vegetables, raw vegetables, fresh fruits without added sugar (also juice) and fresh fruits with added sugar (also juice). Participants for whom more than 29 (49%) of the CEHQ-FFQ items were missing were excluded from the analysis. Finally, we calculated fat, sugar and water propensity ratios for index children assigned to control and intervention groups at baseline using the same strategy as described in the previous text. However, because of the focus on obesogenic foods in the IDEFICS study, as opposed to the interest in general food intake in I.Family, the CEHQ-FFQ 16 (Suppl. 2), 127–137, December 2015
obesity reviews
130 IDEFICS intervention and diet in families L. Arvidsson et al.
at baseline consisted only of 43 food items; therefore, participants for whom more than 21 (49%) food items and more than three (50%) beverages were missing were excluded from the analysis. Because of large differences in number of food items included in the CEHQ-FFQ at baseline and in I.Family, no statistical comparisons were possible.
to evaluate whether index children from control and intervention groups already differed in propensity ratios at baseline. All analyses were performed with Stata 13.1 (StataCorp, College Station, TX, USA; http://www.stata. com) without adjusting for multiple testing. Thus, our results should be interpreted from a more exploratory point of view.
Statistics Descriptive characteristics of the families including age (mean age of the parents and mean age of the children), sex (percentage of female participants in the family), BMI (percentage of overweight participants in the family), household size and parental education were examined in an exploratory analysis using chi-square tests for categorical variable (presented as counts and percentages) and independent t-tests for continuous variables (presented as means ± standard deviation). A chi-square test was further used to compare parental education in the index children who participated in I.Family and met our inclusion criteria with the rest of the original IDEFICS baseline children. This was performed separately for the intervention and control groups. In order to analyse diet at the family level, propensity ratios were obtained by calculating the mean ratio of the participants belonging to a family. Multilevel models using the mean propensity ratio of the families as the dependent variable were used to assess whether there were significant differences between intervention and control families. Random intercepts for study centre and random slopes for the effect of the intervention were included to consider the clustered study design. The reduced model was adjusted for country, age (mean age of the parents and mean age of the children) and sex (percentage of female participants in the family). The full model was additionally adjusted for overweight (percentage of overweight participants in the family) and household size. Among a lower number of families with available parental education data, stratified analyses were performed by lower and higher education strata using the same models as described in the previous text. The family data were disaggregated into parents and children to see whether the differences in diet propensity ratios between intervention and control families could be confirmed in individuals. Multilevel models included an additional random effect for family membership. Random slopes were also assessed to allow the random intercept to vary between intervention and control groups for each study centre and family member. To adjust for potential confounding, the following covariates were considered in the reduced model: country (categorical), sex and age (continuous). The fully adjusted model further included BMI (categorical) and household size. Finally, we included educational level (two categories) as a covariate. The same multilevel models described in the previous text were used 16 (Suppl. 2), 127–137, December 2015
Results Characteristics Descriptive statistics are presented in Table 1. The number of families participating from each country highly varied.
Table 1 General characteristics of the families Family-level variables
Intervention group
Families (n, %) Families from each country* Italy (n, %) Estonia (n, %) Cyprus (n, %) Belgium (n, %) Sweden (n, %) Germany (n, %) Hungary (n, %) Spain (n, %) Number of parents* Mothers (n, %)† Fathers (n, %)† Number of children Girls (n, %) Boys (n, %) Mean age of the parents and children* Parents (mean ± SD) Children (mean ± SD) % Female participants in the family 25 and ≤ 50% (n, %) >50 and ≤ 75% (n, %) >75% (n, %) % Overweight in the family‡ 25 and ≤ 50% (n, %) >50 and ≤ 75% (n, %) >75% (n, %) Household size‡ 2–3 Family members (n, %) 4 Family members (n, %) >4 Family members (n, %) Parental education§ Lower-education families (n, %) Higher-education families (n, %)
2,548 (54.3) 474 334 562 49 257 383 374 115
Control group 2,143 (45.7)
(18.6) (13.1) (22.1) (1.9) (10.1) (15.0) (14.7) (4.5)
472 330 308 82 221 267 345 118
(22.0) (15.4) (14.4) (3.8) (10.3) (12.5) (16.1) (5.5)
2,360 (64.2) 1,318 (35.8)
1,978 (67.0) 975 (33.0)
1,980 (50.0) 1,983 (50.0)
1,622 (48.0) 1,754 (52.0)
42.51 ± 5.4 11.22 ± 2.5
41.85 ± 5.6 11.00 ± 2.2
241 1,120 663 524
(9.5) (44.0) (26.0) (20.6)
226 927 529 461
(10.6) (43.3) (24.7) (21.5)
858 843 364 359
(35.4) (34.8) (15.0) (14.8)
772 726 256 280
(38.0) (35.7) (12.6) (13.8)
517 (21.3) 1,217 (50.2) 690 (28.5)
461 (22.7) 969 (47.6) 604 (29.7)
916 (41.8) 1,278 (58.3)
860 (45.0) 1,052 (55.0)
Families with at least one parent and one child are included. SD, standard deviation. † Including male/female caregivers (
