Ecological factors and childhood eating behaviours at 5 years of age: findings from the ROLO longitudinal birth cohort study

The ROLO (randomized control trial of low glycemic index diet in pregnancy) study is an ongoing longitudinal birth cohort. The primary study took place in the National Maternity Hospital, Dublin, Ireland from 2007–2011 and focused on a low glycemic index diet during pregnancy, with the aim of preventing the recurrence of fetal macrosomia [33]. Findings from the primary ROLO study have been published elsewhere [33]. The mothers and infants from this study (n = 759) have been followed-up as part of the ROLO longitudinal birth cohort study. Participants have been followed-up at numerous time points including at 3 and 6 months, 2 years and when children were aged 5 years old for the ROLO Kids study. ROLO Kids consisted of 401 mother–child dyads and will be the focus for this cross-sectional analysis. Of the 401 children that returned at the 5-year-old follow-up, complete data on child eating behaviours at 5 years old was available for 306 participants, resulting in a final sample of 306 mother–child dyads. A study flow chart detailing the progression of the ROLO study and participant numbers at each follow-up can be seen in Fig. 2. Ethical approval was granted by Our Lady’s Children’s Hospital, Dublin (OLCHC) and the National Maternity Hospital (NMH), Dublin Ethics Committees (Ethics reference number: GEN/279/12).

Weight was measured to the nearest 0.1 kg, using a calibrated stand-on digital weighing scale (SECA 813, Hamburg, Germany). Standing height was measured, without shoes, with head aligned in the Frankfort plane, using a free-standing stadiometer (SECA 217, Hamburg, Germany) and measurements were recorded to the nearest 0.1 cm. Body mass index (BMI) was calculated as kilogram per meter squared (kg/m²). Children’s BMI scores were converted to standardized z-scores according to the 1990 UK age- and sex-specific reference data using Excel LMS Growth macro [34, 35]. BMI z-scores were categorized using the World Health Organization criteria for children aged 5 to 19 years [36]. All measurements were taken at the 5-year-old follow-up study day as per study protocol and were carried out by trained researchers.

Children’s eating behaviours were measured at 5 years old using the Children’s Eating Behaviour Questionnaire (CEBQ) [37]. The CEBQ is a 35-item, parent reported, validated psychometric tool, developed to capture individual differences in eating styles that may contribute to both underweight and overweight in children [37]. The CEBQ has been validated in a British cohort of children aged 4–5 years old as an accurate measurement of child eating behaviours, displaying good internal consistency and good test–retest reliability [38]. To date the CEBQ has not been validated in an Irish cohort. The CEBQ has eight subscales, four that measure food approach eating behaviours and four that measure food avoidant eating behaviours. The food approach eating behaviour subscales include: ‘Food Responsiveness’ (5 items), measuring heightened responsiveness to external food cues, ‘Enjoyment of Food’ (4 items) measuring a general liking for eating, ‘Emotional Overeating’ (4 items) measuring food intake in response to negative emotions such as anger or anxiety and ‘Desire to Drink’ (3 items) which measures an increased desire to drink, particularly of sugary drinks. The food avoidant behaviour subscales include: ‘Satiety Responsiveness’ (5 items) which measures the degree of self-regulation of food consumed by the child, based on the sensation of feeling full, ‘Emotional Undereating’ (4 items) measuring a decrease in food intake due to negative emotions such as anger or anxiety, ‘Slowness in Eating’ (4 items) measuring the speed at which a child eats, with slowness representing disinterest in food, and ‘Food Fussiness’ (6 items) which measures, a lack of interest in food, or for trying new foods. Responses to the CEBQ statements are scored on a 5-point Likert scale (1 = Never, 5 = Always) with five statements within the CEBQ being reverse scored due to opposite phrasing. To determine the score for each subscale, the items within the subscale were summed and its mean calculated by dividing by the number of items within the subscale. A higher score indicates the child is more likely to express this eating behaviour.

Screen time was assessed at 5 years old, using the Children’s Leisure Activities Study Survey (CLASS) questionnaire, which is validated for children aged 5–6 years old [39]. Screen time was defined as a combination of three activities; watching television/video games, computer games, and internet. Mothers were asked to report on the amount of screen time minutes for a typical week (Monday to Friday) and a typical weekend (Saturday and Sunday). Screen time was calculated in minutes per week, and analyzed in hours per week.

Maternal weight was measured at the same visit as her child’s and used the same methodology. Maternal education was self-reported and recorded at recruitment for the ROLO pregnancy study. Mothers selected one of the following categories; ‘no schooling’; ‘primary education only’, ‘some secondary level education’, ‘complete secondary level education’, ‘some third level education (certificate/diploma)’ or ‘complete third level education (higher-level degree)’.

Mothers reported breastfeeding duration retrospectively, at 6 months, 2 years and 5 years postnatally. At each timepoint, mothers reported if they had breastfed or not and for how long. At the 2 year and 5-year follow-up mothers reported the age (weeks) their infant had commenced solids. A variable was created to indicate if their child had started solids as per national recommendations [40], that is; that complementary foods are not introduced before four months (17 weeks), but should not be delayed beyond six months (26 weeks).

The Pobal Haase & Pratschke Deprivation Index (HP Index) was used to allocate a deprivation score as per the participants address or small area [41]. Participants addresses were obtained during the ROLO pregnancy study (2007–2011). The HP Index is derived from data from the ‘2011 Census of Population in Ireland’, and provides information on a combination of three dimensions of relative affluence and deprivation, specifically demographic profile, social class composition and labor market supply and demand. The HP Index data is normally distributed; with the mean fixed at 0. Thus, a negative score (below 0) is classified within the disadvantaged categories and a positive score (above 0) being grouped within the advantaged categories. SES was determined by creating a composite variable, using maternal education level and Pobal HP Index [41]. Four SES categories were created based on level of education and neighborhood deprivation index: ‘Third level and Advantaged’, ‘Third level and Disadvantaged’, ‘Less than third level and Advantaged’ and ‘Less than third level and Disadvantaged’.

Information on childcare was obtained from maternal reported lifestyle questionnaires at the 5-year follow-up. In the current analysis, childcare attendance was defined as non-parental care, which may have been in the home or outside the home. Mothers were asked an open-ended question about whether their child attended childcare or not. Categories provided for mothers to describe their childcare arrangements included; crèche or preschool (formal), childminder in the home, childminder outside the home, relative in the home, relative outside the home (informal) or other (asked to specify). Duration of childcare attendance was calculated by subtracting the age of childcare commencement from the age of the child at the 5-year examination. The childcare questionnaire asked whether food was provided, and this was classified into five groups; ‘Yes, all meals’, ‘Yes, main meals only’, ‘Yes, snacks only’, ‘No meals’ or ‘Don’t know’.

Continuous data were tested for normality using the Kolmogorov–Smirnov test and visual inspection of histograms. Normally distributed variables were reported as mean and standard deviation (SD). Non-parametric variables were reported as median and interquartile range (IQR). Eating behavior variables were normally distributed and parametric tests were used. One-way ANOVA with post-hoc Tukey’s tests were used to explore crude mean differences in maternal and child characteristics across SES groups. Maternal BMI was non-normally distributed; therefore, Kruskal–Wallis and Mann Whitney U tests were used to examine these characteristics across SES. Chi-squared test for independence was completed to examine breast feeding exposure across SES. Unadjusted and adjusted linear regression analysis were performed to examine associations between maternal SES and children’s eating behaviours at age 5. For this analysis dummy variables were created, with the largest category ‘Third level and Advantaged’ used as the reference variable. Conditions of independence, linearity, normality and homoscedasticity were tested and met prior to all linear regression analyses. Cronbach α was performed on each subscale of the CEBQ to assess the internal consistency of each subscale in our cohort.

Total screen time exposure was analyzed in hours per week. Adjusted and non-adjusted linear regression were completed to examine associations between children’s screen time and eating behaviours.

Independent sample t-tests examined differences in eating behaviours and child’s body composition for children who had attended childcare and those who did not. Meals provided in childcare were recategorized into two groups, ‘food provided’ and ‘no food provided’. Independent sample t-tests were completed to examine differences between these two groups. Spearman’s correlations were completed to assess relationships between duration of time spent in childcare and children’s eating behaviours. Type of childcare attendance was stratified into 3 groups – ‘formal’ (creche/preschool), ‘informal’ (nanny or relative inside or outside the home) and ‘mixed’ – combination of both. One-way ANOVA assessed differences in eating behaviour and children’s body composition across type of childcare.