Introduction

Childhood obesity is a major public health issue and a leading risk factor for health problems worldwide [1]. In 2019, the World Health Organization (WHO) estimated that 38.2 million children globally under age 5 years were overweight or obese [2]. In 2020, the Australian Institute of Health and Welfare (AIHW) reported that a quarter of Australian children and adolescents aged 2–17 years were classified as overweight or obese [3]. Obesity is likely to persist from childhood into adulthood and has been associated with a broad range of poor health outcomes in later life [1]. Common risk factors of childhood obesity include poor diet, physical inactivity, increased screen time, and insufficient sleep duration [4]. Nutrition in early life has been found to be one of the key contributing factors of childhood obesity in many Western countries, including Australia [5, 6]. Emerging studies revealed that dietary behaviours are established in early childhood and track across life stages [7]. Furthermore, early nutrition and dietary behaviours may contribute to the programming of long-term health, including obesity. Thus, understanding the dietary determinants of obesity in early life is critical for obesity prevention and early health promotion.

Consuming breakfast is an important dietary habit, which may track from childhood to adulthood [8, 9]. A good quality breakfast replenishes the body with energy and essential nutrients after overnight fast, and it has been associated with body weight control [10]. It has been postulated that consuming breakfast can reduce the chance of overeating later in the day via stabilising blood glucose level and facilitating appetite control [11, 12]. Emerging cross-sectional studies have found that regular breakfast consumption and better breakfast quality were associated with lower obesity risk at various ages during childhood to adulthood [13,14,15,16]. A previous cross-sectional study in American children aged 9–13 years has demonstrated that poor breakfast quality was linked with increased glycaemic response and appetite, which may in turn contribute to obesity [12]. Similarly, results from another cross-sectional study of Chinese children aged 6–13 years revealed that children with high breakfast quality showed lower fasting glucose and lower body mass index (BMI) [17]. While these cross-sectional studies provided important information on the role of breakfast quality and its association with obesity outcomes, longitudinal studies are needed to further understand the temporal order of the relationship between breakfast quality and obesity outcomes. Elucidating the changes in breakfast quality over time will also inform the critical time points for intervention. Moreover, a few studies have explored the longitudinal association between breakfast frequency or skipping with childhood obesity. One longitudinal study described breakfast consumption trends in 2–18-year-old German children, and revealed that regular breakfast consumption decreased with age [9]. Timlin et al. examined the longitudinal relationship between breakfast frequency among 2216 American adolescents aged 14 years and their weight change 5 years later [18]. They found that daily breakfast consumers were less likely to change their body weight gain over time. Furthermore, an Australian study investigated the longitudinal association between breakfast skipping and child and maternal BMI among 2–5-year-olds [19]. Although young Australian children who skipped breakfast were more likely to be childhood overweight and obesity in 3 years later and they appeared to be with a higher maternal BMI association, it was still unclear to identify the association between breakfast quality and childhood obesity outcomes over time. To date, no studies have assessed longitudinal associations between breakfast quality and obesity outcomes in early childhood. Knowledge on how breakfast quality changes in early childhood and its influence on obesity development in early childhood is vital to inform when and where to target early dietary and obesity prevention interventions.

Considering early childhood as a critical period for obesity prevention and the potential adverse effects of poor breakfast quality, it is important to study breakfast quality in early childhood and to identify its relationship with body weight development. Therefore, this study aimed to describe changes in breakfast quality index (BQI) (i.e. trajectories) across three time points at ages 1.5, 3.5 and 5.0 years and secondly to assess the association between BQI trajectories and obesity outcomes at age 5 years, in a cohort of Australian children.

Materials and methods

Study design and participants

The Melbourne Infant Feeding Activity and Nutrition Trial (InFANT) was a 15-month cluster-randomised controlled trial focused on supporting first-time parents to improve young children’s dietary and activity behaviours [20]. At baseline in 2008, 542 parent-child pairs, with infants ~4 months of age, were recruited from 62 first-time parents’ groups across 14 local government areas within Melbourne, Australia [21, 22]. All participating parents provided informed consent. Participants were randomised into an intervention or control group. The intervention comprised six dietitian-delivered 2 h group-based sessions addressing nutrition, and active play, using an anticipatory guidance framework [22]. Participants were followed up post-intervention when children were aged 3.5 and 5.0 years to assess the sustainability of the intervention effects. The detailed study protocol and intervention outcomes have been described elsewhere [20, 21, 23]. Data from children at the end of the intervention (age 1.5 years) and the two post-intervention follow-ups (ages 3.5 and 5 years) were used in the current analysis. To maximise sample size, data from the intervention and control groups were pooled for the present analysis as no differences between groups were found for breakfast intakes of nutrients or food groups across the three different time points (Table S1). Also, intervention allocation was included as a covariate in analyses to account for potential confounding due to intervention allocation. The InFANT study was approved by the Deakin University Ethics committee in 2007 (ID number: EC 175-2007) and the Victorian Office for Children (Ref: CDF/07/1138).

Assessment of dietary intake

Child dietary intake at ages 1.5, 3.5 and 5 years was assessed using three 24 h recalls conducted with parents or main care givers by trained nutritionists [21, 24]. The recalls were collected over 3 non-consecutive days, including 3 weekdays and 1 weekend day, using a five-pass standard recall procedure based on methods used by the U.S. Department of Agriculture [25]. During the interview, parents were asked to recall all food and beverages their child consumed in the previous day (24 h). A food measurement booklet was provided to parents to assist with portion size estimation. The collected 24 h recall data were converted into food and nutrient intakes using the 2007 Australian Food and Nutrient (AUSNUT) database [26].

Assessment of breakfast consumption

When 24 h recall data were collected, parents reported their child’s food and drink consumption by time and not according to the individual meal occasion such as breakfast, lunch, dinner, and snack. This approach eliminates recall bias by eating occasion and promotes reporting of all foods consumed in 24 h [27]. For this reason, eating occasion durations that are widely used in the literature (15, 30 and 60 min from the first food/beverage consumed) were examined in the present study [28]. The 15 min cut off failed to capture breakfast foods and included only water. Moreover, a number of children (about 45% of children) had extended breakfast consumption over 30 min. An eating duration of 60 min most appropriately captured breakfast food and beverage intakes. For the breakfast timeframe, first eating occasions occurred between 5 am and 10 am in the InFANT study and all children reported some food or beverage intake during this time. As a result, this study defined breakfasted as the first eating occasion occurring between 5 am and 10 am, including all foods and beverages consumed across a 60 min duration.

Assessment of breakfast quality index (BQI)

BQI was calculated using a previously published nine-item BQI tool [29] which was adapted to the current sample based on dietary recommendations and nutrient reference values for young Australian children [30], scoring one point each for the consumption of cereals, wholegrains, dairy products, fruit, vegetables; one point for the intake of Calcium (Ca) > 167 mg for children aged 1.5- and 3.5 years (one-third of 500 mg/day, Recommended Dietary Intake; RDI) [30] and >230 mg of Ca for children aged 5.0 years (one-third of 700 mg/day RDI) [30]; one point for energy intake providing 20–25% of total daily energy intake; one point each for absence of added sugar (sugar, jam, honey), and absence of butter and margarine (Saturated Fatty Acid and trans-rich fats). Scores on the BQI ranged from 1 to 9. Detailed descriptions of each BQI food group are provided in Table S2. Multiple Source Method (MSM) was applied to combine the BQI scores from three non-consecutive days to derive a BQI score that reflected usual intake [31]. The proportions of children who received a score for each BQI item at three different time points over 3 non-consecutive days were assessed. BQI items were ranked from high to low based on the proportion of children meeting criteria for each item.

Assessment of child anthropometrics

Each child’s birth weight was transcribed from their health record at baseline (child age ~4 months). Trained staff used standardised protocols to measure children’s height and weight at ages 1.5, 3.5, and 5.0 years. Height was measured using a portable stadiometer (Invicta IP0955, Oadby, Leicester, UK) and weight was measured using a calibrated scale (Tanita 1592, Tokyo, Japan). The average of two measurements was used in the present analysis. Body mass index z-score (zBMI) was derived using WHO age- and sex-specific growth standards [32]. International Obesity Task Force (IOTF) age- and sex-specific cut-offs were used to categorise weight status [33].

Assessment of child and maternal covariates

Parents reported child and maternal characteristics through self-administered questionnaires at baseline which included questions on child sex and gestational age (<37 weeks vs ≥37 weeks), as well as maternal educational level (university or non-university), pre-pregnancy height and weight and country of birth (Australia vs others). Maternal education level was used as a proxy for Socio Economic Status (SES) as this has been considered as a good proxy for SES in previous studies [34,35,36]. Further, maternal pre-pregnancy BMI was calculated using self-reported height and pre-pregnancy weight and categorised into underweight/healthy weight (<25 kg/m2) vs overweight/obesity (≥25 kg/m2). Information on breastfeeding duration (<6 months vs ≥6 months) and timing of solid food introduction (before age 6 months vs at or after age 6 months) was collected at baseline and 6 months-follow-up (child age ~9 months).

In line with previous analysis [36], the 6-months cut off was used for both breastfeeding duration and the timing of solid food introduction to reflect infant feeding guidelines to exclusively breastfeed for the first 6 months of life and introduce solid foods around 6 months of age [37].

These child and maternal factors were considered as potential covariates in the present analysis as they have been associated with child dietary intakes and obesity risk in young children in previous studies [21, 36, 38].

Statistical methods

Identifying BQI trajectories

Data analyses were performed using Stata 16.0 (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC.) with the significance level at P < 0.05. Group-based trajectory modelling was conducted to identify BQI trajectory groups over three time points at ages 1.5, 3.5 and 5 years using the “Traj” command. This method utilises all available data and does not exclude participants with missing data [39, 40]. The analysis included children with two or more BQI measures over three time points.

Censored normal models with linear, quadratic, and cubic functions of child’s age with two to four groups were conducted. The selection of the optimal number of trajectory groups was based on model parsimony, average posterior portability (>0.7), proportion of sample in each trajectory group (>5%) and distinctive and clinical interpretable visual inspection of the trajectories. In addition, the highest (less negative) Bayesian Information Criterion (BIC) indicates better model fit. For BQI trajectories, three groups showed the best BIC, but included a small group (5.5% of total sample). Two groups were therefore chosen for model parsimony. After removing the non-significant cubic function, the two-group trajectory model with quadratic (2 2) was chosen (BIC = −1148.14, AIC = −1129.49) as the final model (Table S3).

Summarising descriptive analysis between BQI trajectories

Descriptive analyses were conducted to summarise cohort characteristics by identified BQI trajectory groups. T-tests and Chi-squared tests were used to compare continuous and categorical child and maternal variables by BQI trajectory groups, respectively.

Association between overweight risks and BQI trajectories

Mixed effects multivariable linear and logistic regressions specifying parent groups as random effects to account for potential clustering were performed to examine associations between identified BQI trajectory groups and zBMI and overweight status at age 5 years, respectively. The crude models adjusted for zBMI and overweight status at age 1.5 years. Further, the models additionally adjusted for child sex, maternal country of birth, maternal education, maternal pre-pregnancy BMI and child total energy intake at age 1.5 years (Model 1). Additional models (model 2) adjusted for breastfeeding duration (≥6 months).

Results

Sample characteristics

Of the 542 children participating at baseline, 393 children participated at age 1.5 years and were included in the current analysis. A further 65 children who had no or only one BQI measure at age 3.5 or 5.0 years were excluded, resulting in 328 children being included in the group-based trajectory modelling analyses to identify BQI trajectory groups. Of 328 children, 145 children had two BQI measures (ages 1.5 and 3.5 or 5.0 years) and 183 had three BQI measures (ages 1.5, 3.5, and 5.0 years). Characteristics of those children included and excluded had no statistically significant differences (all P value > 0.05) (Table S4). For the included children, there were similar proportions of boys (53%) and girls (47%). Most children had birth weight ≥2.5 kg (93.8%), were introduced to solid foods at or after 6 months of age (88%) and were breastfed for ≥6 months (59.3%). Most mothers were born in Australia (80.8%), had a healthy pre-pregnancy BMI < 25 kg/m2 (65%) and had university-level education (59.5%).

For analysis between identified BQI groups and obesity outcomes, 62 children without anthropometric data at 5.0 years were excluded. This resulted in a sample of 266 children (with complete data on covariates including child sex, maternal country of birth, maternal education, pre-pregnancy BMI, and child total energy intake at 1.5 years) being included in the adjusted model 1 analysis. For the adjusted model 2 analysis, which additionally adjusted for breastfeeding duration, the final sample was reduced to 244 children (22 children excluded for missing data on breastfeeding duration). The flowchart outlining the sample sizes for each analysis are shown in Fig. S1.

Average BQI scores

Mean BQI (mean ± SD) at ages 1.5, 3.5 and 5.0 years was 4.8 ± 0.9, 4.8 ± 0.8, 2.7 ± 1.6 points, respectively (possible maximum of 9 points). The average BQI scores for children at 1.5 and 3.5 years of age were similar but declined significantly at age 5 years.

Table 1 shows the proportion of children meeting the criteria for each BQI item across three time points and items were ranked from high to low. The top three highest ranking BQI items were grains, absence of butter and margarine, and dairy products, which were consistent across all three time points. Rankings of BQI items for ages 1.5 years and 3.5 years were identical. Relative to 1.5 and 3.5 years, higher proportions of children consuming wholegrains and calcium were observed at 5 years. In contrast, proportions of children meeting the item “absence of added sugar” and consuming fruit were slightly lower at 5.0 years. Across three time points, the two lowest ranked items were consumption of 20–25% of total energy intake and vegetable intake. Detailed proportions of children meeting the criteria for each BQI item across three time points are shown in Table S5.

Table 1 Ranking of breakfast quality index (BQI) items from high to low by proportion of children meeting the criteria for each BQI item at ages 1.5, 3.5 and 5.0 years.
Full size table

BQI trajectories

Two BQI trajectory groups were identified from the group-based trajectory modelling (Fig. 1). Both groups showed similar BQI scores at age 1.5 years, with a decrease in BQI at age 5 years. Most children (74%; n = 244) showed a smaller decrease from 5.0 to 4.0 points and were classified as the “High BQI” group. About a quarter (26%; n = 84) of the children showed a decrease from 4.8 to 1.2 points and were classified as the “Low BQI” group. There were no statistically significant differences for any child or maternal factors between the high and low BQI groups (all P values > 0.05; Table 2).

Fig. 1: Comparison of average observed and predicted BQI from ages 1.5, 3.5, 5.0 years of the low BQI and high BQI trajectory groups.
figure 1

Both trajectory groups decreased across three different ages. The group with less decreasing BQI values was referred as the high BQI group (n = 244), and the group with relatively large decreasing BQI values was referred as the low BQI group (n = 84). Black lines represent high BQI group and grey lines represent low BQI group. Solid lines represent observed BQI values and dotted lines represent expected BQI values. There is no significant difference between the actually observed BQI value and the expected value.

Full size image
Table 2 Characteristics by trajectory groups using a two-group model (N = 328).
Full size table

Association of BQI trajectory with BMI z-score and weight status

Associations between BQI trajectory groups and obesity outcomes are presented in Table 3. The prevalence of overweight was 10.2% at age 1.5 years and 15.2% at age 5.0 years. Across all models, no statistically significant differences in zBMI or overweight status were revealed between the two BQI groups. However, the low BQI group showed a tendency for a higher risk of overweight at age 5 years than the high BQI group in all models (adjusted OR: 1.39, 95% CI 0.63, 3.10; model 1).

Table 3 Results of multivariable linear and logistic regressions of BQI trajectory groups with zBMI and overweight status of children at aged 5.0 years.
Full size table

Discussion

This study is the first to use novel longitudinal trajectory modelling to describe breakfast quality trajectories and their associations with obesity outcomes in early childhood. Two distinct breakfast quality trajectory groups from ages 1.5 to 5.0 years were identified, and both trajectories showed a decline in average BQI across three time points. No statistically significant association was found between identified BQI trajectory groups and obesity outcomes, but children in the low BQI group appeared to have a higher risk of being overweight at age 5.0 years.

Most previous research on breakfast consumption in young children has been cross-sectional and limited to school-aged children and adolescents [10, 29, 41, 42]. Few studies have examined changes in breakfast consumption and quality in young children [9, 43]. Consistent with our findings, a study (n = 1081) of 2–18-year-old German children found that the proportion of children meeting breakfast consumption guidelines (a proxy for breakfast quality) decreased between age 2–5 years (29%) and 13–18 years (23%) between 1986 and 2007 [9]. There are no studies to date that have examined longitudinal trajectories of breakfast quality in young children. The findings of the present study provide new longitudinal evidence on changes in breakfast quality in early childhood.

In the present study, across ages 1.5, 3.5 and 5 years, most children (>80%) consumed grains and dairy products at breakfast. In contrast, fruits and vegetables were consumed at breakfast by 30% and 3% of children, respectively. This is expected as breakfast cereals and milk are commonly consumed Australian breakfast foods [44], and the high proportion of children consuming both foods aligns with the Australian societal norm for breakfast. Congruent with the findings of the present study, Smith et al. [45] found that breakfast skippers showed lower dairy intake than breakfast consumers in Australian children aged 2–17 years. The selection of breakfast food items can be influenced by several factors such as food availability and convenience of preparation and consumption [46]. Compared to fruit and vegetables, breakfast cereals and milk have a longer shelf life and are more likely to be available at home for consumption. Moreover, given morning may be a busy time for families, pre-packaged breakfast cereals and milk may be preferred because they can be readily consumed with minimal preparation, whereas vegetables may require cooking, making them potentially difficult to consume in the morning. In addition, the low intake of vegetables can be attributable to young children’s preferences for sugary foods [47,48,49]. The findings of the present study suggest that increasing intakes of vegetables and fruits could improve breakfast quality among young children. Research has also shown that the eating environment (presence of family members and screen use) may influence eating behaviours of toddlers [50]. Also, maternal food choices for children may be influenced by informal social networks (neighbours and friends) [51]. Given parents have a primary role in preparing breakfast for young children, future research could focus on parental strategies to improve breakfast quality with consideration for eating environment and informal social networks.

Previous studies have also reported that older children and children from low-SES families were more likely to have low breakfast quality [9, 29, 52]. It is postulated that parents with a higher educational qualification or income may have higher nutrition knowledge [53]. Low breakfast quality observed in older children and adolescents may be due to various reasons such as less parental supervision at breakfast, insufficient time in the morning and personal reasons for diet/weight loss [29, 54]. The lack of difference in child and maternal factors between two BQI trajectory groups in our study might be attributed to the younger age group examined, with young children having less control of their eating habits. Furthermore, the over-representation of highly educated mothers in our sample may have reduced the potential to detect SES differences.

The present study is the first to evaluate the association between breakfast quality (trajectories) and obesity in early childhood. Although there were no statistically significant differences in obesity outcomes between the two BQI trajectory groups, the low BQI group did show a tendency for higher overweight risk at age 5 years. The lack of a significant association could be explained by the small sample size and limited statistical power.

The association between breakfast quality and obesity remains controversial. There are several hypotheses supporting a potential link between breakfast quality and obesity. First, children with higher breakfast quality are more likely to have higher overall diet quality [41], which in turn may protect against obesity. Other hypotheses relate to the consumption of certain breakfast foods (such as wholegrains and dairy products) and their associated influence on obesity. For instance, the fibre contained in wholegrains may reduce fat absorption and/or promote satiety response, and thus lead to lower food intake [55]. In addition, calcium in dairy products has been shown to play a favourable role in inhibition of fat absorption and regulation of lipid metabolism [56]. In contrast, some studies have revealed that consumption of sugary breakfast cereals and high protein intake from dairy products may increase obesity risk in children [57,58,59]. Sugary breakfast cereals have a high glycaemic load and their consumption during breakfast after an overnight fast can rapidly raise blood sugar level, exacerbate insulin response, and in turn contribute to obesity risk [57]. High protein intake from dairy products may stimulate insulin-growth factor-I and promote obesity in children [58, 60]. In addition, the amount of energy consumed at breakfast may also contribute to obesity development. Consistent with our findings, two studies found that most children did not consume the recommended energy intake during breakfast (20–25%) [41, 61]. Evidence has shown inadequate breakfast energy intake was associated with unhealthy markers (i.e. blood sugar level, insulin, triglycerides, cholesterol) related to obesity [17]. Studies have reported that school-aged children with obesity tend to have low energy intakes at breakfast [62]. Additional studies are needed to understand the association between breakfast quality and obesity in early childhood.

The present study has several strengths. To best of our knowledge, it is the first study that examined breakfast quality trajectories in early childhood. Dietary intake was assessed using three 24 h recalls which are gold-standard approaches for assessing dietary intake [20, 23]. Furthermore, anthropometry was objectively measured by trained staff rather than self-reported as occurs in most existing studies. In addition, the utilisation of group-based trajectory modelling enabled a novel assessment of BQI trajectories. This method does not exclude participants with missing data and can examine various trajectory modelling using unbalanced longitudinal data [39, 40]. The method also allowed the exploration of the relationship between breakfast quality trajectories and obesity outcomes rather than breakfast quality at one time point, as evaluated in other studies. In terms of assessing breakfast quality, there is no standardised tool to define breakfast quality in the literature. Based on the existing literature, an adapted 9-item BQI including both nutrient and food groups based on breakfast recommendations was developed to evaluate breakfast quality. The inclusion of both nutrient and food groups provides a comprehensive evaluation of breakfast quality. Further breakfast research should consider both nutrient and food groups.

The present study has some limitations that warrant discussion. The observational nature of our study cannot infer causal relationships. Although a range of child and maternal covariates were considered when assessing the association between BQI trajectory groups and obesity outcomes, unmeasured and residual confounding by other factors is possible [41]. For instance, there are a range of other factors that may influence weight status in children such as physical activity, sedentary behaviour, and sleep duration [46, 63]. It would also be desirable to have other measures of SES, such as household income and maternal employment, despite maternal education being found as a good proxy for SES in previous literature [34,35,36]. In addition, the small sample size may have limited the statistical power to detect significant association between BQI trajectory groups and obesity outcomes. Moreover, the current sample consisted of a high proportion of highly educated mothers, which may limit the generalisability of the study findings to the Australian national population. Further research in a diverse range of populations is needed to better understand the contribution of breakfast quality in obesity and other health outcomes.

In conclusion, two BQI trajectory groups in early childhood were found in this study. Both trajectory groups showed a decline in breakfast quality from 1.5 to 5.0 years. Our study highlights the need for early health promotion interventions and strategies to improve breakfast quality in early life. Given that Australia’s current focus on breakfast promotion is limited to primary school children and adolescents [64], it might be useful to expand the current breakfast promotion program to early childhood education services or to develop successful nutrition programs for parents regarding healthy breakfast, particularly in early childhood. Our study reinforces the importance and need for healthy breakfast promotion and intervention in early childhood.