Changes in dietary intake during puberty and their determinants: results from the GINIplus birth cohort study

The present analysis was based on data collected at the 10- and 15-year follow-ups of the ongoing German birth cohort study GINIplus (German Infant Nutritional Intervention plus environmental and genetic influences on allergy development). Details on the GINIplus study design, recruitment and exclusion criteria have been described previously and can be found elsewhere [18]. In short, healthy full-term new-borns (n = 5991) were recruited from obstetric clinics in two different regions of Germany (Munich and Wesel). Infants were allocated to the study intervention arm (randomized to one of three hydrolysed formulae or to conventional cow’s milk) or to the non-intervention arm. Data on health outcomes and covariates were collected by means of identical questionnaires, completed by parents of all children at various time-points. Information on the relevant exposure variables and covariates is given below. To aid reporting of results, the 10-year time-point is hence forth referred to as baseline, and the 15-year time-point as follow-up.

This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the local ethics committees (Bavarian Board of Physicians, Board of Physicians of North-Rhine-Westphalia). Written informed consent was obtained from all subjects.

Dietary assessment at baseline and follow-up was carried out using a self-administered FFQ, designed and validated to measure 10-year-old children’s usual food and nutrient intake over the past year, and more specifically to estimate energy, fatty acid and antioxidant intake [19]. Due to the uncertain quality of dietary information collected from young children, the FFQ at baseline was addressed to the parents, who completed it along with their children. This was done in order to maximise accuracy by obtaining mutual impact from both the child and the parent [19]. At follow-up, the FFQ was addressed directly to the participants, who were asked to complete it themselves with support of whoever cooked at home, if needed. The FFQ comprised of eighty food items accompanied by several questions about preferred fat and energy contents, preparation methods, diets and food preferences, buying habits and dietary supplement use. To estimate how often food was consumed over the previous year, subjects could choose one of nine frequency categories, including ‘never’, ‘once a month’, ‘2-3 times a month’, ‘once a week’, ‘2-3 times a week’, ‘4-6 times a week’, ‘once a day’, ‘2-3 times a day’ and ‘four times a day or more’. In addition, common portion sizes were assigned for each food item to enable an estimation of quantities. For food items that are difficult to describe in common household measures, coloured photographs from the EPIC (European Prospective Investigation into Cancer and Nutrition) study showing three different portion sizes were included [20]. The 80 FFQ food items were allocated into 41 groups and combined to form 17 major food groups. The categorization systems of a number of sources were compared [21‐26] and adapted to the food items present in the FFQ. A list of the resulting food groups is displayed in Table 1. Further details on the development of the FFQ, including food item selection, dietary vitamins, supplement use, and validation methods, have been previously described [19, 27].

A quality control procedure was developed and applied to the FFQ data at both time-points (Fig. 1). This was done based on recommendations by Willett et al. for data cleaning in nutritional epidemiology [28]. Subjects were excluded if a complete block of food items, presented together under the same subheading, was empty (144 at baseline and 134 at follow-up). For each food item, if the intake frequency was provided, but portion size was missing, portion size was replaced by the median obtained from the remaining sex-specific populations. Subjects were excluded if responses to more than 40 food items (50 % of the FFQ) were missing (16 at baseline and 4 at follow-up). Intake frequencies and amounts were then combined to calculate average consumption in grams per day (g/d). Evidence suggests that the presence of intermittent blanks in an otherwise carefully completed FFQ, are best considered as no consumption of the missing food item [28]. Therefore, any remaining missing information on frequency of intake was regarded as “never”, and intake of the specific food item was defined as 0 g/d. Based on the German Food Code and Nutrient Database (BLS) version II.3.1 [29], the corresponding energy and nutrient content per daily grams of intake were calculated for each food item. Total daily energy and nutrient intake was obtained by the sum of daily energy and nutrients of all food items respectively. Intakes relative to total daily energy intake were calculated as the ratio of energy from each food item or macronutrient to the total daily energy intake, and multiplied by 100 to obtain percentage contributions towards total energy intake (%EI). Due to the lack of energy content of water and tea, these food groups were presented in g/day. Furthermore, vitamin intakes were presented in mg/day. Subjects were excluded if total daily energy intake was outside 500-3500 kcal or 800-4000 kcal for females and males respectively (38 subjects at baseline and 126 at follow-up), ranges suggested by Willett et al. in order to avoid substantial loss to follow-up [28]. Further exclusions were made if provided values for %EI of specific food items were implausible (1 subject at follow-up due to extreme rice values: 57 % of total daily energy intake from rice or 620 g/d). Only participants who completed the FFQ at both time-points were included (n = 1304). After excluding participants presenting extreme values for co-variables (1 subject), or reporting an illness affecting diet (22 subjects) or medical dietary indications (49 subjects), 1232 participants remained for inclusion in the analyses. Due to the extensive quality control applied at both time-points, the FFQ data in the present study differs from that in previously published papers using only the GINIplus 10-year follow-up dietary data [19, 27].

Basic characteristics of the study population stratified by sex are displayed in Table 2. Parental education was mostly high, especially in females (71.4 and 62.9 in females and males respectively). More females (46.4 %) than males (10.9 %) had reached the onset of puberty at baseline, and pubertal development at follow-up was more advanced in females then in males. Mean baseline energy intake was significantly higher in males than females (2105.4 kcal/d (standard deviation = 567.7 kcal/d) in males and 1831.4 kcal/d (488.1 kcal/d) in females), with similar macronutrient proportions in both sexes. Follow-up energy intake was also higher in males, but protein and fat intake was greater in males whereas females consumed more carbohydrates. More females (13.7 %) than males (6.3 %) started or stopped a diet between assessments.

Average intakes of food groups changed significantly in both males and females. Both sexes presented a decrease in intakes of starchy vegetables, dairy and margarine, and an increase in total water intake. Meat intake increased in males and decreased in females (mainly due to subjects changing towards a vegetarian or vegan diet). Males also reduced fruit intake and increased caloric drinks, while females increased intakes of whole and refined grains, vegetables, oils and tea. As in our study, a study in Swedish adolescents aged 15 at baseline, and followed up at ages 17 and 21, reported that changes in food group intakes in males were less frequent than in females, suggesting a greater tendency in females to modify their diet during pubertal maturation and throughout adolescence [38]. Nevertheless, these changes did not seem to impact the overall intakes of macronutrients and vitamins in females, who presented only decreased retinol intakes. As meat and dairy are sources of this vitamin [42], the reduced consumption of these food groups in females might explain the lower retinol intakes. Males however, significantly increased protein and fat intakes, as well as n3 and n6 PUFAs, vitamin C and alpha-tocopherol, and decreased carbohydrates. Furthermore, food groups presenting changes in the previously mentioned study were similar to those in our study: females decreased fat spread, milk and meat intakes, and increased pasta intake from 15 to 17 years. At 21 years females had further reduced their meat intake and males had reduced fruit intake [38]. An increased consumption of caloric drinks in adolescence has also been observed previously in Norwegian [13] and German populations, especially in males [40].

Dietary tracking assessed the stability of food intakes within the study population. Females and males presented “fair” levels of tracking for all food groups, except for butter, margarine, and tea, which revealed stronger tracking; and oil, fat, carbohydrate and retinol in females and alpha-tocopherol in males, which showed only slight tracking. Previous studies on tracking of dietary behaviour in females and males during puberty have reported similar (slight to moderate) tracking levels for food groups such as fruit and vegetables [13, 16], caloric drinks [13, 31], dairy [31] or meat [11], among others. The present results suggest a possible overlap of dietary behaviours observed in other countries, although this may be limited due to sociocultural differences. We noted that food groups indicating greater stability were also amongst those presenting highly significant changes in average intakes. For example, average margarine intake decreased significantly over time, but margarine also presented the highest tracking coefficients in both females and males. These results are not necessarily contradictory as it is possible for a child to significantly modify his/her intake of a specific food group, while remaining in the same position relative to others in the sample (indicative of tracking). We performed further sensitivity analyses to determine if specifically non-tracking participants were responsible for the observed changes, but this was not the case. These results suggest that in our study sample, average intake changes observed during puberty in food groups such as margarine, starchy vegetables, fruit and caloric drinks, follow sex-specific secular trends, where the “order of the children by intake” remains but the overall median intakes are altered.

In the present study, the association of dietary intake changes, with selected indicators of socio-economic status and individual characteristics differed amongst females and males for different food groups, macronutrients and vitamins, and according to baseline intake levels. Studies on the determinants of changes in dietary intake during puberty are limited. Wang et al. [11], reported that children’s dietary intake patterns can be predicted by family income, urban-rural residence, maternal education and baseline dietary intakes. In our study we observed significant associations of dietary intake with parental education, family income, child education and BMI. Given that the consumption of nuts, whole grains, vitamin C and retinol are frequently associated with health-benefits [42‐44], our findings suggest that higher SES in females, represented by higher parental education and family income, may promote an increased consumption (nuts) or at least the maintenance of higher intakes (whole grain, vitamin C and retinol) of certain healthier foods and nutrients during puberty. On the other hand, our results also indicate that females with lower family income were more likely to increase whole grain intakes than those with high income. Despite typically being more expensive [45], increasing whole grain products may be an attainable goal in children with less resources making efforts to improve their diet as they grow older. In males, higher parental education was associated with maintenance of low egg and n3 PUFA intakes as opposed to increasing intakes. Egg intake has been previously associated with unhealthy lipid profiles in humans [46]. Adolescent males with higher educated parents may be more informed with regards to dietary advice [47], and eggs may hence be eaten sparingly. Eggs are also a source of n3 PUFA, which may in turn remain low in the same male subgroup of parental education, even though n3 PUFA has been associated with beneficial health effects [48]. Higher child education in males was associated with tracking low intakes of sugar-sweetened foods, rather than increasing them. Those with higher education levels may be more aware of the negative relationship between health and carbohydrate-rich diets, especially sugar [49, 50], and hence attempt to lower their intakes [51].

Higher BMI was associated with tracking of high protein intake in females. In males BMI was positively associated with increasing margarine, and vitamin C, and with maintenance of n3 PUFA levels in the lowest baseline intake tertiles. High BMI is often associated with unhealthy dietary behaviours [52‐54], however in the present study BMI does not seem to be a predominant predictor of unhealthy dietary change during adolescence. This could be due to common underreporting of unhealthy foods in overweight subjects [55, 56] (margarine may be regarded as healthy and hence not underreported, given its lower content of saturated fats compared to butter [57]). The lack of associations with BMI could also be explained by possible earlier influences of the exposure variable on food intake at baseline. Dietary behaviours already established before the baseline assessment could indicate an intake threshold was reached before puberty, impeding further change in that direction, e.g. higher BMI was associated with high starchy vegetable, meat, water and protein intakes at baseline (see Additional file 1: Table S1). Similarly, parental education level, child education and screen-time also showed significant differences in intakes of a number of food groups at baseline (e.g. higher parental education associated with higher intakes of grains, butter and oil and lower intakes of starchy vegetables and margarine; higher child education with higher fruit, wholegrain and butter intakes and lower intakes of starchy vegetables, meat and sugar-sweetened foods; and higher screen-time associated with lower fruit, vegetables, wholegrain and beta-carotene in both females and males). However earlier influences must be interpreted with caution, as these were cross-sectional associations and reverse causality cannot be excluded as there is no previous dietary data available for longitudinal analyses before the 10 year assessments. We hence highlight the importance of longitudinal analyses in investigating associations with dietary intake changes.