Skipping breakfast is detrimental for primary school children: cross-sectional analysis of determinants for targeted prevention

The present study was embedded in the baseline-measurements of the outcome evaluation of the school-based health promotion program “Join the Healthy Boat” in the state of Baden-Württemberg, Germany, in the year 2010. The “Baden-Württemberg Study” followed a cluster-randomized design with a waitlist control group. A more detailed description of the study can be found elsewhere [19].

All children in classes of teachers who agreed to participate in the outcome evaluation were eligible (3,159 pupils). Parents of 1,968 first and second graders in 84 schools (5.4 – 9.9 years old), 62% of all eligible children, gave their consent for participation in 2010. Response rate for parental questionnaires was 87% at baseline. Data from direct measurements at schools or from parental questionnaires were available for 1,943 children at baseline. Parents gave information on their health behavior and their socioeconomic background, as well as on the health and lifestyle characteristics of their children.

The parental education level was assessed and determined according to the CASMIN (Comparative Analysis of Social Mobility in Industrial Nations) educational classification [20]. CASMIN classifies tertiary level as higher educational levels (e.g. academically-oriented university education), while secondary level includes a range from intermediate vocational qualification to full maturity certificates (equals 12–13 years of school in Germany), and primary level comprises inadequately completed general education to basic vocational qualification. Family education level was defined as the highest level of two parents or the level of a single parent who mainly cares for the child. For analyses, family education level was dichotomized into tertiary versus intermediate and elementary level. Household income was assessed in several categories according to the German KiGGS survey (German Health Interview and Examination Survey for Children and Adolescents) [21] and dichotomized for analyses at a threshold of €1,750. A migration background was assumed if at least one parent was born abroad or at least one parent had mainly spoken a foreign language during the child’s first years of life.

Questions about children’s behavior were taken from the validated questionnaires of the German KiGGS survey [22]. Answers were offered on a 5-point Likert scale for frequency of consuming soft drinks, frequency of playing outside (nearly every day, 3–5 times a week, 1–2 times a week, less than 1 time a week, never), and time spent with screen media (never, about 30 min/day, about 1–2 h/day, about 3–4 h/day, more than 4 h/day). Variables were dichotomized for analyses (soft drinks > 1/week, playing outside > 60 min/day, screen media > 1 h/day). Information on the frequency of participation in club sports were retrieved in an open question and results were dichotomized for analyses (> 1/week). Parents gave information on the frequency of breakfast before school for their children on a 4-point scale, the results were subsequently dichotomized for analyses (never, rarely vs. often, always). Furthermore, parents stated how many days a week their children were physically active on a moderate to vigorous level (starting to sweat and/or get out of breath) for at least 60 min a day, as recommended by the World Health Organization (WHO) [23]. Results were dichotomized for analyses at the median (physically active ≥ 4 days/week ≥ 60 min/day). Lastly, parents were asked whether they were smokers, and they rated their health awareness on a 4-point scale, the latter variable being then dichotomized for analyses (not at all, little vs. strong, very strong).

Trained staff conducted the anthropometric measurements of the children according to ISAK-standards [24]. The children’s height was measured to the nearest 0.1 cm (Stadiometer, Seca®, Germany), and body weight to the nearest 0.1 kg using calibrated and balanced portable digital scales (Seca®, Germany). Waist circumference (WC) was measured midway between ileac crest and lower costal arch to the nearest 0.1 cm using a flexible metal tape (Lufkin Industries Inc., Texas, USA). The children’s BMI was calculated as weight divided by height squared [kg/m²]. Excess weight and obesity were defined at or above the 90^th and 97^th age- and gender-specific BMI percentiles according to German reference data [25]. Waist- to-height ratio (WHtR) was calculated as the ratio of WC and height in centimeters, participants with a WHtR ≥ 0.5 were then categorized as abdominally obese [26]. Because there was some doubt the WHtR threshold of 0.5 was sensitive enough, a lower threshold (0.47 for girls and 0.48 for boys) as proposed by Nambiar et al. [27], correlating with the 95^th percentile for %body fat, was added.

Parental BMI was calculated with self-reported weight and height data from the questionnaires, and categorized as overweight (BMI ≥ 25.0) and obese (BMI ≥ 30.0), according to the international classification of the WHO [28]. Parental WHtR was calculated as self-reported WC divided by height in centimeters and abdominal obesity was defined as WHtR ≥ 0.5 [29].

Missing data are a frequently occurring problem in observational studies, possibly leading to biased results [30]. Therefore, potential significant baseline differences between cases with and without missing values for the full regression model were examined.

Differences between boys and girls, as well as between participants with and without missing values, were tested for their statistical significance. Depending on scale level and distribution of the data, the Mann–Whitney-U test or t-test for continuous data and Fisher’s exact test for categorical data were applied. The significance level was set at α = 0.05 for two-sided tests. These analyses were carried out using the statistical software packages IBM SPSS Release 21.0 for Windows (SPSSInc, Chicago, IL, USA).

To identify a possible clustering effect of data in schools, the intraclass correlation coefficient (ICC_Logit) for a generalized linear mixed model with binary outcome was calculated according to Eldridge et al. [31]. Depending on the magnitude of the ICC_Logit, the appropriate regression analysis, a logistic or a generalized linear mixed model, was conducted subsequently, considering the variables described above. A closer examination of decisive factors for target groups and target behaviors was realized in two sub-models. These analyses were performed with the statistical software R Release 3.2.5 for Windows (http://​cran.​r-project.​org/​).

Primary school children who took part in this research had a mean age of 7.1 ± 0.6 years, 51.2% of them were boys. Table 1 shows the baseline characteristics of the participants. Significant differences between boys and girls occurred in anthropometric variables where boys had a slightly higher waist circumference but were less frequently at or above the 95^th percentile for %body fat. They also had a different distribution of WHtR with a lower variance and range than girls. Boys more often played outside and reached higher levels of physical activity. Girls more often skipped breakfast than boys.

All variables shown in Table 1 were considered for their potential association with the outcome variable in the regression analyses.

Unadjusted, crude odds ratios (OR) for skipping breakfast for all variables in the subsequent generalized linear mixed regression models are illustrated in Table 2.

In the bivariate analysis, female gender, migration background, variables of overweight and obesity, single parenthood, lower household income, parental overweight and obesity, parental smoking, one or more soft drinks per week, and screen media use exceeding 1 hour per day had higher odds for skipping breakfast. On the other hand, tertiary family education level, parental health awareness, playing outside more than 60 min per day, and physical activity of more than 60 minutes on 4 days and more per week showed lower odds for skipping breakfast. The ICC_Logit of skipping breakfast was 0.045, indicating that 4.5% of the total variance is due to clustering of data in schools. This would lead to differences in the ORs between a logistic regression model and a generalized linear mixed model. Therefore, the model presented here is the generalized linear mixed model for binary outcomes, which accounts for the clustering of data in schools. Table 3 shows both the results of the full model and of sub-models for groups and behaviors for targeted preventive measures.

A higher level of physical activity and tertiary family education level showed lower odds for skipping breakfast. On the other hand, migration background, single parenthood, female gender, percentage of body fat at or above the 95^th percentile, and frequent soft drink and screen media consumption had higher odds.

No significances were lost after dividing the full model into two separate sub-models for target groups and target behaviors. ORs in the sub-models were more pronounced than in the full model.

Further sub-models were calculated for boys and girls, they are depicted in Table 4.

The sub-model for girls showed higher odds for migration background, single parenthood, percentage of body fat at or above the 95^th percentile, and frequent soft drink and screen media consumption. In the sub-model for boys lower odds were found for a higher level of physical activity and tertiary family education level.

Participating children with missing data, significantly, more often had a migration background, lower results in the 6 min run test, and they differed in all anthropometric measurements from the ones with complete data. Participants with missing data had significantly higher BMI percentiles, more often were overweight and obese, and more often had a higher waist circumference and a higher WHtR. They also were more often abdominally obese and more often had a percentage of body fat at or above the 95^th percentile than their counterparts. Finally, their parents were significantly more often single parents, less frequently had a tertiary educational level and more often a household income at or below €1,750.

Although the importance of breakfast consumption to young children’s health is generally known, there is an increasing prevalence of children skipping breakfast [34, 35]. Food behaviors established in childhood are often continued into adulthood [36]. Therefore, it is necessary to identify the determinants of skipping breakfast. Based on these determinants, interventions for preventing skipping breakfast and promoting healthy dietary behaviors among children can be developed.

The parental role in the development of children’s healthy breakfast behaviors is not questioned [37]. Pearson et al. report in their review on family correlates of breakfast consumption that parental breakfast intake is associated with the breakfast intake of their children [38]. Furthermore, they found out that living in two-parent families also has a positive influence on children’s and adolescent’s breakfast consumption [38]. The research from Wendy & Campbell also shows that children with single parents are more likely to skip breakfast than those with two parents [39]. These findings confirm the results of the present study that also found evidence for a relation between family structures and skipping breakfast. Therefore, it is important to consider family structures of the children when designing programs to promote healthy breakfast behavior’s.

Associations between parenting and children’s breakfast consumption were found for permission to skip breakfast and parental self-efficacy of skipping breakfast which were negatively associated with children’s breakfast consumption [40]. Another study by Fugas et al. shows further reasons for skipping breakfast: lack of time, not being hungry in the morning and feeling unwell at the time of having breakfast are identified as explanations for skipping breakfast before going to school [41]. Similar to our findings that girls are more likely to skip breakfast than their male peers, other studies showed that breakfast consumption was more frequent among boys [42, 43]. This result could be explained by the likelihood that even young girls care more about their appearance and a slim figure [44]. A further study found out that adolescent boys and girls are more likely to skip breakfast if they perceived that their mothers often skip lunch [45]. Equal results are available for girls skipping breakfast with regard to their best friend’s meal skipping behaviors. On the contrary, those who reported exemplary maternal healthy eating behaviors were less likely to skip breakfast [45]. These findings underline the importance of interventions to address parents, children and their peers simultaneously.

The investigations of Keski-Rahkonen et al. and Utter et al. show an association between skipping breakfast and having a higher BMI [43, 46]. Research on breakfast intake and abdominal obesity parameters are rare. Alexander et al. report that eating breakfast was associated with lower visceral adiposity in overweight Latino youth, aged 10 to 17 years [11]. In a survey of Iranian children and adolescents aged 6 to 18 years the percentage of abdominal obesity in breakfast skippers was almost 5 percentage points higher than in non-skippers [47]. A French study in primary schoolchildren showed that those who regularly eat breakfast had the lowest waist circumferences [48].

There are many studies showing a positive relationship between skipping breakfast and a low socioeconomic status [42, 49]. In the present study, children with migration background were more likely to skip breakfast than their counterparts and a tertiary family education level was positively associated with having breakfast.

Children who spend more time in front of screen media and are less physically active are more likely to skip breakfast than their peers. These results are in line with the results of a study by Tin et al. in Hongkong with primary schoolchildren [49] and the study of Timlin et al. with adolescents [42]. Only one study found no relation between physical activity and skipping breakfast [43]. TV viewing during meals and the consumption of sugar sweetened beverages along with skipping breakfast were associated with significantly higher waist circumferences in French primary schoolchildren [48].

The probability of skipping breakfast increased with age [1]. The present study could not detect associations between age and skipping breakfast, probably because of the restricted variation in the age of the participating primary schoolchildren.

Last but not least, there should also be a focus on general meal patterns and obesity. Not just skipping breakfast is associated with obesity but a wide range of obesogenic behaviors influences weight status. Berg et al. showed an association between obesity and skipping breakfast, skipping lunch or eating at night. Larger self-reported portion size was also related to obesity. However, the investigation showed no significant relationship with intake of total energy [50]. Other authors emphasized that large, high energy-dense portions favor obesogenic eating behaviors in children [51]. Moreover, serving children large entrée portion sizes increases total energy intake but without decreases in intake of other foods. If children can self-select and limit their food intake, energy intake will decrease [52]. In this context, it seems necessary not only to consider breakfast skipping children when providing healthy breakfast at school. Attention should also be attached to children who had breakfast previously to avoid a higher intake of energy that might aggravate the situation and would be counterproductive in obesity prevention.

In summary, our study contributes to the body of evidence that exits for factors associated with skipping breakfast in primary schoolchildren. We identified vulnerable groups for targeted prevention and behavioral aspects, which need to be addressed in addition with preventing children from skipping breakfast. Therefore, the current results play an important role for developing targeted preventive measures for skipping breakfast in primary schoolchildren.

This research provides a valuable contribution in exploring determinants for the prevention of skipping breakfast in schoolchildren. However, some aspects should be considered when interpreting these findings.

Anthropometric measurements of the children were taken in a standardized manner according to a protocol by specially trained staff. Data management and statistical advice was provided by the Institute of Epidemiology and Medical Biometry at Ulm University. However, a limiting factor is the cross-sectional character of this research that precludes any causal interpretations of the results because cross-sectional studies do not allow conclusions about the direction of the detected associations.

Parents did not completely fill in questionnaires therefore missing values occurred. In observational studies, missing data are a frequently arising problem and may possibly bias the results. Therefore, the study examined the specific differences between those participants with missing and those with complete data. Considering these differences, children with missing data show several critical characteristics: They have higher rates of migration background and abdominal obesity, and less frequently a tertiary educational level. If these participants could have been included in the regression analysis for correlates of skipping breakfast, it might have accentuated the results. In this research, only schoolchildren whose teachers gave their agreement to participate were involved. Furthermore, due to the voluntary participation, only 62% of parents of eligible children gave their consent. Thus data from over a third of eligible children could not be collected. It may be assumed that there are differences between children who participated in the study and those who failed. This may contribute to the missing data bias.

In common with studies with an observational character, some unintentional bias may compromise the results. As already mentioned, only schoolchildren whose teachers and parents gave their agreement to participate were studied. Therefore, a twofold selection bias may have occurred on behalf of the teachers deciding to opt in and parents giving their consent for participation. Recall bias and social desirability bias may affect the parental report concerning the offspring’s patterns of physical activity, screen media use and consumption of soft drinks as well as breakfast habits. Parental breakfast was not assessed in the present study, but should be included in future research. Furthermore, the way skipping breakfast was assessed is a limitation of the study. Although the results are not representative for the whole of Germany, the sample size and the fact that this research comprises of data from an entire federal state of Germany are great strengths of the study. Response rates from participating parents with 87% at baseline were remarkably high. The study thus provides a valuable contribution for exploring determinants in the prevention of skipping breakfast in schoolchildren.