Breakfast is widely considered to be an important component of a healthy lifestyle for both adults and children. Its consumption is promoted by the UK Government’s Change4Life public health campaign in England( 1 ). One of the suggested benefits of breakfast for children and adolescents is that eating breakfast regularly protects against overweight and obesity. To date, evidence for this consists largely of epidemiological studies( Reference Rampersaud 2 – Reference Timlin, Pereira and Story 9 ) and intervention studies have so far failed to demonstrate a causal relationship between breakfast habit and adiposity( Reference Dhurandhar, Dawson and Alcorn 10 – Reference van Nassau, Singh and Cerin 13 ). Another proposed benefit of breakfast consumption in children and adolescents is that it improves overall nutrient intakes( Reference Rampersaud 2 ).
Previous studies have been carried out into variations in nutritional intakes dependent on child breakfast habits, but many of these focus on consumption of a specific type of breakfast, for example breakfast cereal( Reference Albertson, Anderson and Crockett 14 – Reference van den Boom, Serra-Majem and Ribas 16 ) or a ‘good-quality’ breakfast( Reference Matthys, De Henauw and Bellemans 17 ). Some have examined overall differences in nutritional profile between children who eat breakfast and those that do not, including recent studies involving North American, Mexican and Australian populations( Reference Deshmukh-Taskar, Nicklas and O’Neil 18 – Reference Afeiche, Taillie and Hopkins 22 ), and have reported that breakfast consumption compared with breakfast skipping was associated with improved nutrient intakes( Reference Deshmukh-Taskar, Nicklas and O’Neil 18 , Reference Barr, DiFrancesco and Fulgoni 20 – Reference Sampson, Dixit and Meyers 30 ). However, the authors are not aware of any recent studies of UK or other European children examining differences in nutrient intakes between breakfast consumers and breakfast skippers. Therefore, our main aim in this study was to investigate associations between breakfast skipping in 4–18-year-olds and their nutrient intakes using data from the UK’s National Diet and Nutrition Survey Rolling Programme (NDNS RP). We compared daily nutrient intakes between children with different breakfast habits and also carried out a within-person analysis, comparing intakes of children on breakfast days with their intakes on non-breakfast days. This latter approach was adopted in order to minimise the impact of residual confounding inherent in cross-sectional studies.
Many epidemiological studies comparing the characteristics of individuals with different breakfast habits ask participants (or their parents) how frequently they (or their children) consume breakfast, but do not include a definition of this meal( Reference Szajewska and Ruszczynski 4 , Reference Williams 27 , Reference Antonogeorgos, Panagiotakos and Papadimitriou 31 , Reference Berkey, Rockett and Gillman 32 ). Where breakfast is defined, this definition varies widely( Reference Rampersaud 2 , Reference O’Neil, Byrd-Bredbenner and Hayes 33 , Reference Betts, Chowdhury and Gonzalez 34 ). Some studies classify any energy intake within a specified time period as breakfast( Reference Grieger and Cobiac 19 , Reference Fayet-Moore, Kim and Sritharan 21 ), others include only solid foods and not beverages in the definition, regardless of the energy content of the latter( Reference Betts, Chowdhury and Gonzalez 34 ). We employed an objective definition of breakfast, based on a minimum energy intake within a specified time period. We hypothesised that intakes of micronutrients of public health interest, which have a key role in children’s healthy development and growth, may be particularly affected by breakfast skipping, as many breakfast items consumed widely by UK children, such as breakfast cereals, fruit juice and dairy products, are important sources of micronutrients for this age group( 35 ).
Methods
Population
The data analysed in this study were collected between 2008 and 2012 as part of the NDNS RP( 36 ). Each year the NDNS RP gathers dietary and nutritional data from approximately 1000 randomly sampled individuals living in private UK households, comprising equal numbers of adults (aged 19 years and over) and children (aged 1·5–18 years)( 35 ). These individuals are sampled using The Royal Mail’s Postcode Address File, comprising all UK addresses, to randomly select addresses. Where there is more than one household at a selected address, one of these households is randomly selected. At each selected household either one adult and one child (if present) are selected, or one child, resulting in roughly equal numbers of adults and children in the final sample. Chapter 2 of the report on the NDNS RP results from years 1–4 (combined)( 35 ) gives fuller details of its methodology. This study focused on the sub-population of children in the NDNS RP 2008–2012 sample aged 4–18 years, consisting of 802 children aged 4–10 years and 884 children aged 11–18 years (1686 in total). Ethics approval for the NDNS RP was obtained from Oxfordshire A Research Ethics Committee( 35 ).
Dietary assessment
Each survey participant was visited in their home by a survey interviewer, who placed a 4-d estimated (unweighed) food diary (with written instructions) to be completed on 4 consecutive days by the participant, or their parent for those aged 11 years and under( Reference Lennox, Fitt and Whitton 37 , Reference Bates, Lennox and Prentice 38 ). Follow-up checks were made by the interviewer to optimise completeness of record keeping in the diary( 35 ). Within the sample analysed for years 2008 to 2012 of the NDNS RP 1686 children aged 4–18 years completed at least 3 diary days (98·2 % of these completed the full 4 diary days). Home visits were carried out continuously throughout each year, from February 2008 to August 2012, thus allowing for seasonal variations in diet( 39 ).
The diary entries were then recorded and analysed by a dietary assessment system using food composition data from the Department of Health’s NDNS Nutrient Databank to estimate energy and nutrient intakes( Reference Lennox, Fitt and Whitton 37 ). The NDNS Nutrient Databank is based on data from McCance and Widdowson’s The Composition of Foods series( Reference Lennox, Fitt and Whitton 37 , 40 ). Non-milk extrinsic sugars (NMES) were defined as comprising all free sugars (added monosaccharides and disaccharides, together with naturally occurring sugars in honey, syrups and fruit juices) and 50 % of fruit sugars from stewed, dried or canned fruit( 41 ). Dietary fibre intakes were of NSP, defined by the Englyst method( Reference Englyst, Quigley and Hudson 42 ). Mean daily energy intakes were expressed as a percentage of the estimated average requirement (EAR) for each child, as specified by the Scientific Advisory Committee on Nutrition( 43 ). Mean daily intakes of the micronutrients folate, vitamin C, Ca, Fe and I were compared with dietary reference values (DRV) (reference nutrient intakes (RNI) and lower reference nutrient intakes (LRNI)) for each child, as set by the Committee on Medical Aspects of Food Policy( 44 ). The results of the above analysis of the diary entries form part of the published core sample data for the NDNS RP 2008–2012( 36 ) and were further analysed in this study as described below.
For each day the food diary entries had been split into seven different time periods: 06.00–08.59, 09.00–11.59, 12.00–13.59, 14.00–16.49, 17.00–19.59, 20.00–21.59 and 22.00–05.59 hours. Microsoft Excel for Mac 2011 (version 14.4.6) and IBM SPSS Statistics (version 23) were used to calculate the total energy intake for each diary day between 06.00 and 08.59 hours and identify those days on which at least 418 kJ (100 kcal) were consumed between 06.00 and 08.59 hours (breakfast days) and those on which <418 kJ (<100 kcal) were consumed (non-breakfast days). The threshold of 418 kJ (100 kcal) was chosen to allow for consumption of a milky drink which was not considered sufficient to be classed as breakfast( Reference Pot, Richards and Prynne 45 ). Using these data, the children were split into three categories: those consuming breakfast every diary day, those consuming breakfast on at least one but not all diary days, and those not consuming breakfast on any diary day. A subset of the group of children with an irregular breakfast habit, that is who consumed breakfast on at least one but not all diary days, was created by eliminating all diary days that comprised a Saturday or a Sunday and then identifying all children who still had an irregular breakfast habit based on weekday diary days alone. This was to correct for a possible shift to later breakfast time and/or a different pattern in nutrient intakes at the weekend. For each child with an irregular breakfast habit mean daily nutrient intakes for breakfast days and non-breakfast days, respectively, were calculated.
The nutrients selected for analysis in this study were the macronutrients protein, fat and carbohydrate plus a selection of other nutrients linked with an unhealthy (NMES and Na) or a healthy diet (dietary fibre). We also analysed intakes of a selection of micronutrients that Public Health England has identified as being of particular interest to public health, namely: folate, Fe, vitamin C and Ca( 35 ). To this list we added I: there is some evidence of I deficiency in UK adolescent girls giving rise to public health concerns due to its vital role in fetal neurodevelopment( Reference Vanderpump, Lazarus and Smyth 46 ). The mean daily nutrient intake values were expressed as a percentage of total energy intake for protein, fat, carbohydrate and NMES and as mean intakes per 4184 kJ (1000 kcal) of energy intake for the remaining nutrients. This was to allow for possible differences in daily energy intakes between breakfast and non-breakfast eaters.
Other measures
During the home visit, the interviewer measured the weight and height of the participant so that their BMI could be calculated (weight in kilograms divided by the square of height in metres)( 35 ). The calculated BMI and the British 1990 growth reference (UK90) charts( Reference Cole, Freeman and Preece 47 ) were used to categorise the children as normal weight, overweight (85th centile cut-off) or obese (95th centile cut-off). The interviewer also conducted a computer-assisted personal interview to collect further data on the individual and their household, including age, ethnicity (five main categories: white, mixed, black or black British, Asian or Asian British, and other), whether the individual was currently dieting to lose weight (a ‘yes’ or ‘no’ response, years 3 and 4 of the NDNS RP, for individuals 11 years and older) and household income (choice of thirteen income bands, ranging from under £5000 to £100 000 or more)( 48 ).
Statistical analysis
The energy-adjusted daily nutrient intake, equivalised household income (equivalised for different household sizes and composition using the McClements equivalence scale( 35 )) and BMI variables were checked for normality by inspecting frequency distribution histograms and skew and kurtosis values and transformed as necessary. The differences between the children grouped by three categories of breakfast habit (consumption of breakfast on every, some or no diary days) with respect to sex, age, ethnicity, weight status and equivalised household income were assessed by χ 2 analysis and ANOVA. χ 2 analysis was carried out on the dieting variable and used to examine the proportions of children meeting the RNI and LRNI for daily intakes of the selected micronutrients. χ 2 analysis was also used to compare the characteristics of children with missing BMI and/or income data to assess whether the relatively high incidence of missing data (n 270, 16 % of sample) might affect the results adjusted for these variables.
Daily nutrient intakes for the three groups of children with different breakfast habits were compared by one-way ANOVA and then ANCOVA, with adjustments made for the following covariates: sex, age, ethnicity, equivalised household income and BMI. Energy intake for this analysis was expressed as a proportion of EAR, to allow for variations in energy intake with age and sex. To investigate the potential effect of under- or over-reporting of dietary intakes a sensitivity analysis was carried out, rerunning the χ 2 analysis of children meeting the DRV and the ANCOVA analysis of energy intake after eliminating those children with energy intake as a proportion of EAR more than 2 sd from 100 % (as described by McCrory et al. ( Reference McCrory, Hajduk and Roberts 49 )). This was not considered necessary for the nutrient intakes, as these were all expressed as a proportion of energy intake. Due to inequality of variance (assessed using Levene’s test) for three of the nutrient variables: protein, NMES and Na, which can lead to an increase in type 1 error rate( Reference Rogan and Keselman 50 ), the ANOVA and ANCOVA analyses for these variables were carried out after equalising the numbers in the three groups of children by random sampling, to create three equal groups of 245 cases each (for large sample sizes and modest levels of variance heterogeneity ANOVA is generally robust to inequality of variances if group sizes are equal( Reference Rogan and Keselman 50 )).
Within-person analysis was carried out on the children with an irregular breakfast habit (n 879), comparing their mean energy-adjusted nutrient intakes for days on which they had consumed breakfast with those for days on which they had not. It was assumed that any degree of mis-reporting of dietary intakes by an individual would be similar across the diary days and would therefore not have a significant effect on the within-person analysis. Paired sample t tests were conducted, followed by repeated measures ANOVA, the latter adjusted for the covariates sex, age, BMI, ethnicity and equivalised household income. These tests were also carried out on the sample split into two age groups: 4–10 years (n 384) and 11–18 years (n 495) and on the subset of 4–18-year-olds with an irregular breakfast habit after removal of weekend diary days. Due to collinearity of nutrient variables, for all the tests carried out a P value of <0·01 was considered as statistically significant to allow for multiple testing, rather than using the potentially overly conservative Bonferroni method of adjustment( Reference Feise 51 ). We did not apply the weighting as provided with the NDNS 2008–2012 RP data set, as the calibration weights may not reflect the characteristics of the subsample of children that we analysed, which may not be representative of the UK population.
Results
Description of population
The characteristics of the children, split by breakfast habit, are set out in Table 1. There was a significant variation in breakfast habit between girls and boys (P=0·001), with 19·9 % of girls skipping breakfast every diary day compared with 14·5 % of boys. At 6·5 %, the proportion of the 4–10-year-olds skipping breakfast every diary day was less than a quarter of the proportion of 11–18-year-olds (26·8 %), and the proportion in the younger age group consuming breakfast every day was 45·6 %, more than twice that of the older age group (P<0·001). Mean equivalised household income varied significantly with breakfast habit (P=0·001), with a mean household income of £28 194 (sd £18 349) for those children consuming breakfast every diary day compared with a mean household income of £23 587 (sd £16 374) for those children skipping breakfast every diary day. There was no significant difference in breakfast habits based on ethnicity or weight status. Data on whether or not the children were dieting during the diary period were only available for 11–18-year-olds in years 3 and 4 of the NDNS RP. A significantly higher percentage of girls were dieting than boys (10·7 v. 3·4 %, P=0·003, n 430), but no significant variation in breakfast habit with dieting behaviour was observed for the children as a whole (P=0·456), or for the girls alone (P=0·419, n 224) (the χ 2 analysis was invalid for the boys as more than 20 % of the expected cell values were <5). There were no significant differences found in sex (P=0·457), age group (P=0·470) or breakfast habit (P=0·844) for children with missing BMI and/or income data compared with those with complete data.
Characteristics of children by breakfast eating habit showing sex, age, ethnicity, weight status and household income (n 1686) (Numbers and percentages; mean values and standard deviations)
* Differences between children with different breakfast habits analysed using Pearson’s χ 2 test.
† P≤0·01 is considered significant.
‡ The BMI measurements for the children were compared with British 1990 growth reference (UK90) charts to assess whether children were normal weight, overweight (85th centile cut-off) or obese (95th centile cut-off), n 1619 (sixty-seven missing values).
§ Differences between children with different breakfast habits analysed using ANOVA.
|| Equivalised household income (using McClements equivalence scale), n 1470 (216 missing values).
Breakfast habits and micronutrient Dietary Reference Values
Significant increases (P<0·001) were noted in the proportion of children meeting their RNI for each of the micronutrients folate, vitamin C, Ca, Fe and I as breakfast frequency increased (Table 2). The same trend was observed in the proportion of children meeting the LRNI for folate, Ca, Fe and I (P<0·001). No children who consumed breakfast daily had a folate intake below their LRNI, compared with 7·3 % of those who did not eat breakfast on any diary day. The proportions of children not meeting their LRNI of Ca, Fe and I in the daily breakfast group were 2·9, 4·4 and 3·3 %, respectively, compared with 19·0, 31·5 and 21·5 %, respectively, in the breakfast skipping group. Similar results were obtained for the sensitivity analysis including only plausible reporters of energy intake (n 1505) (invalid LRNI χ 2 results for folate and vitamin C as more than 20 % of cells had expected count of <5).
Comparison of number and percentage of children below reference nutrient intakes (RNI) (as set by Committee on Medical Aspects of Food Policy (COMA)) and lower reference nutrient intakes (LRNI) (as set by COMA) for folate, vitamin C, calcium, iron and iodine depending on breakfast habitFootnote * (n 1686) (Numbers and percentages)
* P values for χ 2 analysis all <0·001, except for vitamin C LRNI where results invalid as 50 % of cells had expected count <5.
Breakfast habits and nutrient intakes
After adjustment for covariates, we observed significant increases in mean intakes of energy (as a percentage of EAR) (P=0·009) and carbohydrates (as a percentage of energy) (P=0·01) and decreases in mean intakes of fat (as a percentage of energy) (P=0·005), with increasing number of breakfast days (Table 3). However, after eliminating implausible reporters of energy intakes (energy intake as a percentage of EAR more than two SD from 100 % (i.e. outside the range 54–146 %)), the increases in energy intakes became non-significant (P=0·088, n 1271). The percentage of plausible reporters in the sample was 89 %, with a significantly higher proportion of plausible reporters within the 4–10-year-olds than 11–18-year-olds (97 v. 83 %, P<0·001).
Comparison of daily nutrient intakes for children aged 4–18 years with different breakfast habits (Mean values and standard deviations)
EAR, estimated average requirement of energy as set by the Scientific Advisory Committee on Nutrition; CHO, carbohydrate; NMES, non-milk extrinsic sugars.
* ANOVA adjusted for the covariates: age, sex, BMI, ethnicity (white/non-white) and equivalised household income (using McClements equivalence scale); sample size reduced by 270 cases due to missing values for BMI and equivalised income.
† P≤0·01 is considered significant.
‡ Due to inequality of variances for these variables ANOVA and ANCOVA performed after random sample selection to form three equal groups of 245 cases (n 735).
§ Natural logarithm transformation applied to calculate significance; means and standard deviations shown for untransformed variables.
|| Square root transformation applied to calculate significance; means and standard deviations shown for untransformed variables.
In the ANCOVA analysis we observed no statistically significant variations in intakes of protein, NMES, vitamin C or Na, but energy-adjusted intakes of fibre, folate, Ca, Fe and I all increased significantly with increasing frequency of breakfast consumption (P<0·001).
Within-person comparison of nutrient intakes on breakfast days v. non-breakfast days
For 4–18-year-olds, after adjustment for covariates, we observed significantly higher mean intakes of energy, carbohydrate, folate, Ca and I and significantly lower mean intakes of protein and Na for days on which breakfast was consumed compared with days on which it was not eaten (Table 4(a)). The mean energy intake for days on which breakfast was consumed was 364 kJ (87 kcal) higher than for days on which it was not consumed (95 % CI 218, 506 kJ (52, 121 kcal); P<0·001). No significant differences were observed when comparing breakfast with non-breakfast days for mean intakes of fat, NMES, fibre, vitamin C or Fe.
Within-person difference in daily nutrient intakes for days on which breakfast consumed compared with days on which breakfast not consumed, 4–18-year-olds (n 879) (Mean values and standard deviations; mean differences and 95 % confidence intervals)
CHO, carbohydrate; NMES, non-milk extrinsic sugars.
* Repeated measures ANOVA adjusted for the covariates: age, sex, BMI, ethnicity (white/non-white) and equivalised household income (using McClements equivalence scale). Sample size reduced by 145 cases (n 734) due to missing values for BMI and equivalised income.
† P≤0·01 is considered significant.
‡ Natural logarithm transformation applied to calculate significance; means, standard deviations and mean differences shown for untransformed variables.
§ Square root transformation applied to calculate significance; means, standard deviations and mean differences shown for untransformed variables.
For 4–10-year-olds there was no significant difference between mean energy intake on days on which breakfast was eaten and mean intake on non-breakfast days (Table 4(b)). In comparison, in the older age group (11–18-year-olds), after adjustment for covariates, energy intake was significantly higher on breakfast days (494 (95 % CI 276, 707) kJ (118 (95 % CI 66, 169) kcal); P<0·001) (Table 4(c)). For both age groups carbohydrate intake was significantly higher on breakfast days compared with non-breakfast days. There was a significant decrease in protein intake on breakfast days compared with non-breakfast days for the older but not the younger age group. For 4–10-year-olds mean energy-adjusted intakes of fibre, folate, vitamin C, Ca and I were significantly higher on days on which breakfast was eaten, but for 11–18-year-olds out of these five nutrients a significant increase was only observed for Ca intakes. Whereas for the 4–10 years age group there was no significant difference in mean Na intakes comparing breakfast with non-breakfast days, Na intake was significantly lower for the 11–18 years age group on breakfast days.
Within-person difference in daily nutrient intakes for days on which breakfast consumed compared with days on which breakfast not consumed, 4–10-year-olds (n 384) (Mean values and standard deviations; mean differences and 95 % confidence intervals)
CHO, carbohydrate; NMES, non-milk extrinsic sugars.
* Repeated measures ANOVA adjusted for the covariates: age, sex, BMI, ethnicity (white/non-white) and equivalised household income (using McClements equivalence scale). Sample size reduced by sixty-two cases (n 322) due to missing values for BMI and equivalised income.
† P≤0·01 is considered significant.
‡ Natural logarithm transformation applied to calculate significance; means, standard deviations and mean differences shown for untransformed variables.
§ Square root transformation applied to calculate significance; means, standard deviations and mean differences shown for untransformed variables.
Within-person difference in daily nutrient intakes for days on which breakfast consumed compared with days on which breakfast not consumed, 11–18-year-olds (n 495) (Mean values and standard deviations; mean differences and 95 % confidence intervals)
CHO, carbohydrate; NMES, non-milk extrinsic sugars.
* Repeated measures ANOVA adjusted for the covariates: age, sex, BMI, ethnicity (white/non-white) and equivalised household income (using McClements equivalence scale). Sample size reduced by eighty-three cases (n 412) due to missing values for BMI and equivalised income.
† P≤0·01 is considered significant.
‡ Natural logarithm transformation applied to calculate significance; means, standard deviations and mean differences shown for untransformed variables.
§ Square root transformation applied to calculate significance; means, standard deviations and mean differences shown for untransformed variables.
For 4–18-year-olds with an irregular breakfast habit after removal of weekend diary days, there were significantly higher intakes of energy (661 (95 % CI 389, 933) kJ (158 (95 % CI 93, 223) kcal)), folate, Ca and I for days on which breakfast was consumed compared with days on which it was not (Table 5). In contrast to the analysis of all diary days, no significant differences in intakes of any of the macronutrients (protein, fat or carbohydrate) were observed and there was no significant difference in Na intake.
Within-person difference in daily nutrient intakes for days on which breakfast consumed compared with days on which breakfast not consumed, no Saturdays or Sundays, 4–18-year-olds (n 365) (Mean values and standard deviations; mean differences and 95 % confidence intervals)
CHO, carbohydrate; NMES, non-milk extrinsic sugars.
* Repeated measures ANOVA adjusted for the covariates: age, sex, BMI, ethnicity (white/non-white) and equivalised household income (using McClements equivalence scale). Sample size reduced by sixty-seven cases (n 298) due to missing values for BMI and equivalised income.
† P≤0·01 is considered significant.
‡ Natural logarithm transformation applied to calculate significance for untransformed variables.
§ Square root transformation applied to calculate significance for untransformed variables.
Discussion
Our findings suggest that the overall nutritional profile of the children in terms of fibre and micronutrient intake was superior in regular breakfast consumers, for whom higher energy-adjusted intakes of fibre, folate, Ca, Fe and I were observed compared with breakfast skippers. These findings are supported by our observation that significantly higher proportions of breakfast-consuming children met their RNI of folate, vitamin C, Ca, Fe and I compared with breakfast skippers.
In accordance with previous studies( Reference Fayet-Moore, Kim and Sritharan 21 , Reference Rampersaud, Pereira and Girard 23 , Reference Affenito, Thompson and Barton 25 ), for both the between and within-person analyses and both age groups a lack of breakfast was associated with lower Ca intakes. A recent Mexican study of breakfast dietary patterns among 4–13-year-old children (n 3760) found that the differences in overall daily nutrient intake profile between breakfast consumers and skippers varied with the type of breakfast consumed, but all breakfast types were associated with a higher daily intake of Ca than breakfast skipping( Reference Afeiche, Taillie and Hopkins 22 ). This is a particularly important finding given that Ca is a vital nutrient for bone growth and needs in childhood are high due to rapid growth and bone mass accretion( Reference Cashman 52 ). In the within-person analysis, higher intakes of fibre, folate, vitamin C and I were observed on the days that the 4–10-year-olds ate breakfast, but no significant difference in intakes of these micronutrients was noted for the 11–18-year-olds. This suggests that more foods rich in these micronutrients (e.g. fortified cereals, milk and fruit juice) are consumed for breakfast in the younger compared with the older age group, possibly due to greater parental supervision of the younger children’s meals. A Spanish study of 8–17-year-old children (n 4332) evaluating the utility of a breakfast quality index, in which points were awarded for the consumption of cereals, fruits, vegetables and dairy products at breakfast, found that the score decreased with age (P=0·001)( Reference Monteagudo, Palacin-Arce and del Mar Bibiloni 53 ). For the within-person analyses including all diary days, intakes of Na are significantly higher on no breakfast days for 4–18-year-olds and 11–18-year-olds, but not 4–10-year-olds, suggesting that the older age group may compensate for the lack of breakfast by the consumption of salty snacks later in the day (protein intakes are also higher on non-breakfast days for the older but not the younger group). However this association was no longer significant when weekend diary days were removed from the within-person analysis (and is not present in the between-person analysis), so may simply be a reflection of a different dietary pattern at weekends (e.g. late or no breakfast with greater daily amounts of protein rich and salty foods) in the older age group.
We did not find any evidence to support the oft-quoted hypothesis that breakfast skipping leads to increased overall daily energy intake due to compensatory overeating later in the day( Reference Timlin, Pereira and Story 9 , Reference Miech, Kumanyika and Stettler 54 ). On the contrary, in our analyses we observed that breakfast skipping was associated with either no difference or a significantly lower daily energy intake. These results are in accordance with those of most (but not all) recent observational studies of children( Reference Timlin, Pereira and Story 9 , Reference Deshmukh-Taskar, Nicklas and O’Neil 18 , Reference Barr, DiFrancesco and Fulgoni 20 – Reference Afeiche, Taillie and Hopkins 22 ). Similarly, small cross-over trials in children have reported no significant differences in overall daily energy intakes on breakfast and non-breakfast days( Reference Leidy and Racki 55 , Reference Kral, Whiteford and Heo 56 ). The findings on energy intake in this and other studies might suggest caution when recommending breakfast consumption as a weight management tool in children, despite the large body of epidemiological evidence linking breakfast skipping and excess weight in children. Also, they undermine the argument for a causal link between breakfast skipping and overweight and obesity based on excess energy intake. Interestingly, a large, longitudinal US study of breakfast habits and weight gain in 9–14-year-olds found that, although breakfast skipping was associated with higher BMI overall, overweight breakfast skippers tended to lose weight over the study period compared with overweight breakfast eaters, whereas the reverse was true for normal weight breakfast skippers compared with normal weight breakfast eaters( Reference Berkey, Rockett and Gillman 57 ). Alternative theories for the link observed between adiposity and breakfast skipping include the presence of confounding factors such as sleep duration and circadian rhythms( Reference Dhurandhar 12 ), and lower physical activity levels in breakfast skippers( Reference Szajewska and Ruszczynski 4 , Reference Schembre, Wen and Davis 58 , Reference Betts, Richardson and Chowdhury 59 ).
The results of previous studies of children investigating the effect on overall macronutrient profile of breakfast habit vary( Reference Rampersaud, Pereira and Girard 23 ). In this study small but statistically significant variations in macronutrient profile were noted, with the proportion of carbohydrate consumed generally higher for breakfast consumers (for the between-person analysis) and on breakfast days (for the within-person analysis), at the expense of either protein or fat intake. However, no significant variations in macronutrient profile when comparing intakes on breakfast days with those on non-breakfast days were found for the within-person analysis after weekend diary days had been removed from the analysis. This may be due to a different dietary pattern for the children at the weekend. In studies that have analysed breakfast habit by type of breakfast consumed, different overall daily nutrient intake profiles have been associated with different breakfast types( Reference Barr, DiFrancesco and Fulgoni 20 , Reference Afeiche, Taillie and Hopkins 22 ). For example, in a large US cross-sectional study of breakfast consumption in 9–18-year-olds (n 9659), consumers of ready-to-eat cereals had higher intakes of carbohydrate and lower intakes of fat than breakfast skippers, but for consumers of other types of breakfast there was no significant difference in intakes of these macronutrients compared with breakfast skippers( Reference Deshmukh-Taskar, Nicklas and O’Neil 18 ).
No link was observed between overall NMES intake (as a percentage of energy) and breakfast consumption. In the within-person analysis we observed no significant difference in the proportion of fat or NMES consumed between breakfast and non-breakfast days, suggesting that intermittent breakfast skipping did not lead to an increase in consumption of poor quality, high sugar and high fat foods and beverages on non-breakfast days in this study population, as has been postulated elsewhere( Reference Miech, Kumanyika and Stettler 54 ).
In our analysis of the characteristics of the children in the sample we found no significant differences in the proportion of normal weight, overweight and obese children in each breakfast habit category, which is at odds with the findings of some but not all cross-sectional studies( Reference Szajewska and Ruszczynski 4 – Reference Kelly, Patalay and Montgomery 7 ) in children. In line with other studies in children( Reference Timlin, Pereira and Story 9 , Reference Rampersaud, Pereira and Girard 23 , Reference Hoyland, McWilliams and Duff 60 , Reference Adolphus, Lawton and Dye 61 ), we observed higher levels of breakfast skipping in girls and older children and lower mean household incomes for breakfast skippers. There is evidence that frequency of breakfast skipping in teenagers is related to dieting and other weight-control behaviours( Reference Timlin, Pereira and Story 9 ), which may explain its greater incidence in girls( Reference Cohen, Evers and Manske 62 ). A cross-sectional study of UK 11–16-year-olds (n 1019) found that almost twice as many girls reported dieting and those girls that were dieting were three times more likely to skip breakfast than non-dieters( Reference Lattimore and Halford 63 ). In our study a significantly higher proportion of girls in the 11–18-year-old age group stated that they were currently dieting than boys. However, reported dieting behaviour did not vary significantly with breakfast habit for the girls in our sample, but this could be due to the smaller sample size (n 224) resulting in lower statistical power (dieting data were only available for the last 2 years of the NDNS RP 2008–2012).
The greater level of breakfast skipping among older children may be influenced by a reduction in parental control enforcing a ‘healthy’ breakfast habit. It may also reflect the shift in circadian rhythms in adolescence to a later wake/sleep cycle( Reference Crowley, Acebo and Carskadon 64 ). During puberty an individual’s chronotype, that is their preference for an early or late wake/sleep cycle, shifts from early to late, with sleep schedules moving progressively later between the ages of 10 and 20 years( Reference Roenneberg, Kuehnle and Pramstaller 65 ). During the school week children’s wake/sleep cycle is dictated by the school routine but at the weekend the wake/sleep cycle is generally less restricted. A study of food logs of German adolescents (n 152, mean age 13·23 years) found that wake times at the weekend were on average 2h40min later, which translated to later breakfast times: the average breakfast time on weekdays was 06.36 hours, compared with 09.15 hours at weekends( Reference Fleig and Randler 66 ). Wake time data were not available for the study sample, so in our analysis we have defined breakfast in relation to a specific, fixed time period, namely 06.00–08.59 hours, rather than relate it to intake within a certain time of waking. However, because of this weekend shift in breakfast times a meal eaten shortly after waking may not fall within this fixed time period. To address this issue, a separate within-person analysis was carried out after removing weekend diary days.
Chronotype not only varies with age but it also depends on genetic and environmental factors( Reference Roenneberg, Kuehnle and Pramstaller 65 , Reference Roenneberg, Kuehnle and Juda 67 ). Chronotype may be an important confounding factor in the between-person analysis in this and other studies. There is evidence that not only are adolescents with later chronotypes more likely to skip breakfast, they are also more likely to have poorer overall diets( Reference Fleig and Randler 66 ) and lower levels of physical activity( Reference Schaal, Peter and Randler 68 ). The impact of chronotype and other possible residual confounding factors, which are always an issue in cross-sectional studies, should be less of a factor in the within- compared with the between-person analysis. Adjustments have been made in the ANOVA for age, sex, BMI, ethnicity and equivalised household income, however we were not able to adjust for physical activity level due to lack of complete and consistent data across all ages( 35 ).
We relied on data from 4-d estimated food diaries and an objective definition of breakfast to categorise children by their breakfast habit, rather than on responses to an eating habits questionnaire. This avoids issues resulting from inconsistent personal definitions of breakfast. Nevertheless, it is recognised that the 418 kJ (100 kcal) threshold chosen for our definition of breakfast is, to some extent, arbitrary. Also, the fixed time frame we used to define breakfast may result in a late weekend breakfast not being captured by the definition, however this was allowed for in the within-person analysis by removing weekend diary days.
In common with other dietary surveys, there is a possibility of mis-reporting of dietary intakes. Doubly labelled water techniques used in the NDNS RP to validate energy intakes for a sample of survey participants suggest that under-reporting may have been more prevalent for the children in the 11–18 years age group, who completed their own food diaries, in contrast to children in the younger age group, whose diaries were completed by their parents( Reference Coulthard and Pot 69 ). The results of our sensitivity analysis suggest that there were higher levels of mis-reporting in the older age group. Where appropriate, to assess the impact of mis-reporting we reran analyses omitting implausible reporters.
Many different methods of dietary data collection have been used in previous studies of breakfast habits in children( Reference Rampersaud, Pereira and Girard 23 ), with varying degree of reliability. Strengths of our study include the large sample size and the method of dietary assessment used, namely a 4-d estimated food diary. In other recent cross-sectional surveys of children’s breakfast habits and nutrient intakes( Reference Deshmukh-Taskar, Nicklas and O’Neil 18 – Reference Afeiche, Taillie and Hopkins 22 ), the 24-h recall method was used, which relies heavily on the accuracy of the child’s or their parent’s memory. We are aware of two small cross-over studies involving US children which report on the impact of breakfast skipping on overall daily energy intake( Reference Leidy and Racki 55 , Reference Kral, Whiteford and Heo 56 ) but, to our knowledge, there has not yet been a randomised, controlled trial assessing the impact of breakfast consumption in children on intakes of individual macro- and micronutrients. The approach we adopted of approximating a cross-over study design (thereby reducing residual confounding) in free living individuals by conducting a within-person analysis of subjects with an irregular breakfast habit has been carried out previously in adults( Reference Kant and Graubard 70 ), but not children.
In our study we examined how macro- and micronutrient intakes, plus energy intake, varied with children’s breakfast habit. A possible topic for future research would be to look at associations in UK children between breakfast habit and daily intakes of specific foods, incorporating a diet quality index( Reference Gorgulho, Pot and Sarti 71 ) to further investigate associations between breakfast habit and overall diet quality. It would also be interesting to investigate how the quality of the breakfast foods consumed by UK children varies with age, to ascertain if the differences we noted between the two age groups in our within-person analysis might be attributed to lower quality breakfasts in the older age group. Work has already been carried out on developing breakfast quality indices for use in relation to children and adolescents in a Mediterranean setting( Reference van den Boom, Serra-Majem and Ribas 16 , Reference Monteagudo, Palacin-Arce and del Mar Bibiloni 53 , Reference Gorgulho, Pot and Sarti 71 ); further work is required to adapt these for use in a UK population.
In conclusion, the connection between the consumption of breakfast and good health appears to involve many different factors, and is still some way from being fully elucidated. A causal link with obesity is, as yet, unsupported by the available evidence( Reference Brown, Brown and Allison 72 ). However, this study adds to the existing body of data linking breakfast consumption with higher quality dietary intake in school-age children, particularly the 4–10 years age group, supporting the promotion of breakfast as an important element of a healthy dietary pattern in children.
Acknowledgements
The authors would like to thank everyone involved in the National Diet and Nutrition Survey Rolling Programme (2008–2012), the data from which formed the basis of this study.
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
J. D. C. formulated the research question, prepared the data for analysis, analysed and interpreted the data and wrote the manuscript, all under the supervision of G. K. P. and L. P. All authors approved the final draft before publication.
None of the authors has any conflicts of interest to declare.
