A systematic review of the effect of breakfast on the cognitive performance of children and adolescents

Abstract

Breakfast is recommended as part of a healthy diet because it is associated with healthier macro- and micronutrient intakes, BMI and lifestyle. Breakfast is also widely promoted to improve cognitive function and academic performance, leading to the provision of breakfast initiatives by public health bodies. Despite this positive and intuitive perception of cognitive benefits, there has been no systematic review of the evidence. Systematic review methodology was employed to evaluate the effects of breakfast on cognitive performance in well-nourished children and nutritionally at-risk or stunted children. Acute experimental studies, school feeding programmes and studies of habitual breakfast intake are reviewed. Comparisons of breakfast v. no breakfast and breakfasts differing in energy and macronutrient composition are discussed. Included are forty-five studies described in forty-one papers published between 1950 and 2008. The evidence indicates that breakfast consumption is more beneficial than skipping breakfast, but this effect is more apparent in children whose nutritional status is compromised. There is a lack of research comparing breakfast type, precluding recommendations for the size and composition of an optimal breakfast for children's cognitive function. Few studies examined adolescents. Studies of school breakfast programmes suggest that such interventions can have positive effects on academic performance, but this may be in part explained by the increased school attendance that programmes encourage. The present systematic review considers methodological issues in this field and makes recommendations for future research design and policy priorities.

Introduction

Breakfast consumers tend to have higher micronutrient intakes, partly because of the fortification of breakfast cereals, and a better macronutrient profile than breakfast skippers⁽¹⁾. Regular breakfast cereal consumers have healthier body weights but also tend to engage in healthier lifestyle behaviours than those who skip breakfast⁽²⁾. Similarly, children who regularly eat breakfast tend to have a lower BMI and are less likely to be overweight than those who eat breakfast less frequently⁽³⁾. Studies in children suggest that breakfast eaters are more likely to meet daily nutrient intake guidelines compared with children who eat breakfast infrequently or skip breakfast⁽⁴⁾. Despite this, breakfast skipping increased in the USA from 14 % to 25 % between 1965 and 1991⁽⁵⁾. Moreover, the percentage of children eating breakfast in the UK has declined along with the nutrient quality of breakfast foods selected⁽⁶⁾, with implications for nutrient status and energy intake.

A good deal of research has considered the importance of breakfast consumption for cognitive performance⁽⁷⁾. Much of this research has been undertaken in healthy young adults, particularly undergraduate students. The premise that acute interventions can enhance mental performance in this population, in whom cognitive function is well protected, is now being reconsidered⁽⁸⁾. Increasingly, interest has turned towards groups who may be more vulnerable to nutritional deficits or cognitive impairment, for example, children and the elderly⁽⁹⁾.

Breakfast consumption, as with other meals, provides fuel for preferential oxidation of glucose. In children aged between 3 and 11 years, the brain has been shown to account for more than 50 % of body oxygen consumption⁽¹⁰⁾. Children have a higher ratio of brain weight to liver weight (1·4–1·6 v. 0·73 in adults) and a 50 % greater metabolic rate per unit brain weight. Thus children exert greater demands on glycogen stores during overnight fasts which are often longer than in adults. The child's relatively small muscle mass, in turn, limits the availability of glucogenic amino acids for hepatic gluconeogenesis⁽¹⁰⁾. Average global cerebral blood flow and O₂ utilisation are 1·8 and 1·3 times higher in children than adults, respectively⁽¹¹⁾. Positron emission tomography studies of thirty children, aged 0–18 years, demonstrate that the higher cerebral metabolic rate of glucose utilisation gradually declines from the age of 10 years, stabilising at age 16–18 years^{(12, 13)}. The higher metabolic turnover of children, their rapid growth rates and the importance of their cognitive function for academic achievement underlie the need for optimal nutrition.

Previous reviews give mixed support for the effects of breakfast provision in experimental and school settings on cognitive or scholastic performance^(14–16). Pollitt & Mathews⁽¹⁷⁾ concluded that there was no convincing evidence for ‘…either long or short term effects of breakfast on cognition and school learning…’ (p. 804S). Common to these reviews is the inclusion of studies of variable scientific quality, the lack of a clear classification of effects across particular cognitive domains and the failure to identify confounds. These reviews have not systematically evaluated all available published studies selected on the basis of the quality of research design and do not acknowledge the very likely publication bias which results in publication of a greater proportion of studies with positive findings.

Two recent publications have systematically reviewed the effect of nutritional interventions, including school feeding programmes, on physical development, school attendance and performance. One considered a range of nutritional interventions, including breakfast, sugar intake on attention deficit–hyperactivity disorder, and fish oil supplementation⁽¹⁸⁾, concluding that there was insufficient evidence to identify any effect of nutrition on learning of children from the developed world. In contrast, a Cochrane review of school meal provision to disadvantaged children suggested some small benefits for physical and psychosocial health⁽¹⁹⁾. Although the application of systematic review methodology to children's nutrition is timely and desirable⁽²⁰⁾, there has not been a focused systematic review to evaluate the cognitive effects of breakfast in children. Such a review would inform government initiatives in the UK, the USA and elsewhere which aim to improve the diet of children with positive consequences for cognitive function.

The aim of the present review, therefore, was to provide a systematic examination of the best evidence from controlled studies of the effects of breakfast on the cognitive performance of school-aged children. The review examines whether the effects of breakfast on cognitive performance are consistent across populations of differing nutritional status with the aim of evaluating whether breakfast interventions can have an impact on the cognitive performance of well-nourished children in the developed world as well as those of or at risk of poor nutritional status. A further aim was to identify the nature of the breakfast which was associated with the clearest positive effects on cognitive function. Nutritional parameters which could account for breakfast effects on cognition include the macronutrient composition, energy provision and glycaemic properties of the breakfast meal.

Literature search

Search strategy and search terms

Electronic databases were searched on 20 January 2009. The databases queried were MedLine (1950 to January 2009), PsycInfo (1967 to January 2009) and Web of Science (1955 to January 2009). Table 1 provides the search strings used as text words and keywords in each database. Additional search strategies involved scanning reference lists of review articles identified. This yielded three further articles. Following removal of duplicates (n 253), 350 citations were retrieved for possible inclusion in the present review.

Table 1 Search terms

* $ denotes word truncation.

Inclusion and exclusion criteria

Papers were included or excluded according to the following criteria.

Participants

Studies of children or adolescents (aged 4–18 years) of either sex were included. Studies were excluded if they examined adult, elderly or patient samples.

Manipulations

Any type of breakfast manipulation, including studies comparing breakfast with no breakfast, and studies of different breakfast types were included. Studies of the effects of glucose ingestion or of intake at other mealtimes were excluded. Breakfast was defined according to the descriptions of the meals or foods consumed provided in the papers reviewed. These varied but generally considered breakfast to be the first food consumed that day although this was not the case for some interventions where prior intake was not controlled. Studies were not excluded on the basis of the content of the meal; for example, studies that included interventions using drinks and/or snacks were included.

Outcome measures

Studies including any standardised outcome measures of cognitive performance were included. Studies solely examining fatigue or employing only qualitative measures of cognitive performance were excluded. Studies involving teachers' subjective ratings of performance were excluded since these do not provide an objective measure of cognitive function. Acute (where performance was assessed within 12 h of breakfast consumption) and habitual effects of breakfast manipulations (for example, school breakfast programmes) were included.

Study selection process

Figure 1 details the study selection process and the number of papers retrieved and excluded at each stage. Of the 350 studies retrieved, 300 exclusions were made, most commonly because the studies examined adult or elderly samples; assessed breakfast programme delivery or efficiency only; measured breakfast behaviours only; took measures of appetite only. Also excluded were nine review papers. Some papers contained multiple studies^(21–24). Therefore, forty-one articles were extracted providing forty-five studies for review.

Fig. 1 Flow diagram of the study selection process.

Each study appears in the data tables only once, irrespective of whether the data were reported in more than one paper (Grantham-McGregor⁽¹⁵⁾ was also reported in Chandler et al. ⁽²⁵⁾; Pollitt et al. ⁽²³⁾ Study 1 was also reported in Pollitt et al. ⁽²⁶⁾; Simeon & Grantham-McGregor⁽²⁷⁾ was also reported in Simeon & Grantham-McGregor⁽²⁸⁾ and Simeon⁽²⁹⁾ Study 2).

Tabulation of studies

Tables 2–5 were produced to summarise the main characteristics of each study. The studies were categorised according to the intervention duration (acute, long term or habitual) and the study participants (well nourished or of differing nutritional status). Representation by sex and age range as well as mean and sd and were included if details were provided. The study design was classified as repeated measures (where participants received each intervention) or independent groups (where participants were assigned to receive one intervention arm only). Randomisation, counterbalancing or cross-over strategies were noted if sufficient detail was evident to confirm that these experimental controls were employed. The duration of the test battery and the time post-breakfast consumption of administration were included in the tables if these could be determined. Explicitly stated fasting periods were recorded.

Table 2 Acute experimental intervention studies in well-nourished children (twenty-one studies, reported in nineteen papers)

QA, quality assessment; SES, socio-economic status; BF, breakfast; RM, repeated measures; GL, glycaemic load; I&D, immediate and delayed (memory); BBC, British Broadcasting Corporation; CHO, carbohydrate; CT, cognitive test(s); CPT, Continuous Performance Test; EEG, electro-encephalography; IG, independent groups; MFFT, Matched Familiar Figures Test; MAST, Memory and Search Test; GI, glycaemic index; SRT, simple reaction time; CRT, choice reaction time; np, quality assessment not possible; RTE, ready-to-eat; SBP, school breakfast programme; IQ, intelligence quotient; HCI, Hagen Central Incidental Task; RAVLT, Rey Auditory-Verbal Learning Test.

Table 3 Acute experimental intervention studies in children with differing nutritional status (seven studies)

QA, quality assessment; NCHS, National Center for Health Statistics; RM, repeated measures; BF, breakfast; CHO, carbohydrate; CT, cognitive test(s); CRT, choice reaction time; IG, independent groups; ANCOVA, analysis of covariance; SES, socio-economic status; IQ, intelligence quotient; I&D, immediate and delayed (memory); CPT, Continuous Performance Test; np, quality assessment not possible; SBP, school breakfast programme; RT, reaction time; MFFT, Matched Familiar Figures Test; HCI, Hagen Central Incidental Task.

Table 4 School breakfast programmes (SBP) in well-nourished children and children of differing nutritional status (thirteen studies)

QA, quality assessment; NCHS, National Center for Health Statistics; IG, independent groups; BF, breakfast; CT, cognitive test(s); SES, socio-economic status; IQ, intelligence quotient; DoH, Department of Health; RT, reaction time; np, quality assessment not possible; CHO, carbohydrate; MFFT, Matched Familiar Figures Test.

Table 5 Habitual breakfast in well-nourished children (four studies)

QA, quality assessment; IG, independent groups; BF, breakfast.

Cognitive tests were listed with their respective outcome measures (dependent variables) if specified. The cognitive domain assessed by each test is provided where description of the test features was sufficiently detailed.

Where enough detail was provided, the energy content of the breakfasts was calculated from the macronutrient composition of the interventions. The statistical analysis performed on the data by each study is also provided. The comments provided for each study in the tables consider study quality in terms of design, analysis and conclusions drawn and indicate the sponsor or funding body where provided.

Quality assessment

Each study was rated for quality using pre-defined assessment criteria by two of the authors independently. The inter-rater correlation for ratings was r 0·85 and discrepancies were discussed as a panel with the third author to reach consensus. Validated tools for the assessment of clinical trials (for example, Jadad et al. ⁽³⁰⁾) do not lend themselves to the design features and experimental manipulations or comparisons made by breakfast studies. Therefore, we devised an eighteen-item tool which covered key elements of study aims and design, sample selection, breakfast manipulation, controls, analysis and outcomes (see Appendix). This tool was based on others developed to assess barriers to healthy eating in children⁽³¹⁾ and parental decision making for child health⁽³²⁾. All criteria were equally weighted and a score of 1 was obtained if the criterion was satisfied. Quality assessment (QA) ratings appear in Tables 2–5 next to each study reference. Due to insufficient details, five studies could not be assigned a QA rating (see Tables 2–5). We chose not to exclude studies on the basis of quality threshold because there was a limited number of studies in some categories and a large variation in the adequacy of descriptions provided. Hence we provide a quality assessment critique for each type of study.

Results

Studies are grouped into four categories according to the type of investigation undertaken. Of the forty-five studies, twenty-eight were examinations of the acute effects of breakfast v. no breakfast or breakfast type on cognitive performance. This is broken down into studies in well-nourished children (n 21) and in children of differing nutritional status (n 7). Of the studies, thirteen examined the long-term effects of school breakfast programmes and breakfast clubs on cognitive performance. Lastly, four studies examined the effect of the quality of habitual breakfast intake on cognitive performance. Of the forty-five studies, only ten examined participants over the age of 13 years.

Studies of acute effects of breakfast in well-nourished children

There were twenty-one studies of the acute effects of breakfast in well-nourished children identified and included in the review (see Table 2). Of the studies, nine were performed in the USA and six were from the UK. Of these studies, four came from the same two research groups. The rest were conducted in Europe, Israel, China and Venezuela. Details provided by each article varied considerably.

A repeated-measures design was employed by thirteen studies. A large number of cognitive measures were employed. Some cognitive domains were examined more frequently; for example, six studies employed continuous performance tests which assess sustained attention, and nine studies used other measures of attention such as visual search and checking tasks. Of the studies, nine employed tests of verbal memory and seven employed spatial memory. Digit span was used in six studies. Not all studies assessed more than one aspect of cognitive function and the timing of tests post-ingestion also varied. Studies often considered more than two treatments, commonly including breakfast and no-breakfast conditions within the design. The inclusion of multiple breakfast conditions within the same (often small) sample has implications for the statistical analysis such that degrees of freedom and power are reduced.

Effects of breakfast v. no breakfast

The majority of these studies demonstrate positive effects of breakfast compared with no breakfast. However, effects vary over cognitive domain. Benefits of breakfast consumption were most evident on measures of memory and in terms of fewer errors on attention tasks especially later in the morning when performance decrements become apparent on the no-breakfast conditions. Effects on memory and attention are also clearest because more studies have used the same or comparable measures across these domains.

When verbal memory was assessed following breakfast–no-breakfast interventions, six studies report null findings and four studies show positive effects of breakfast. Of the six studies which assessed spatial memory, three report a benefit of breakfast consumption, two show better performance in the no-breakfast condition and one shows no difference. In the study by Wesnes et al. ⁽³³⁾, it is not possible to distinguish verbal and non-verbal memory performance because a composite factor ‘quality of episodic memory’ (derived by principal components analysis) is reported. On this factor, there is a notable advantage for breakfast over no breakfast but immediate or delayed verbal performance cannot be distinguished from other memory measures. However, factors were derived from studies in adults and factor scores were not weighted for loading on the factors. Hence the factor structure and loadings may not be the same in children.

There are three well-designed studies that showed no effects of breakfast v. no breakfast comparisons on verbal memory^{(22, 34, 35)}. Each used a repeated-measures design and the breakfasts provided between 753 kJ⁽²²⁾ and 2238 kJ⁽³⁵⁾. Mahoney et al. ⁽²²⁾ demonstrated positive effects of breakfast using a spatial memory task but not with story recall (verbal memory). Vaisman et al. ⁽³⁶⁾ showed positive effects of breakfast at school compared with breakfast at home or no breakfast on immediate verbal recall. However, this study was an unbalanced design with no consideration of change over time or time of testing post-consumption, which differed for breakfast consumed at school or home.

Wesnes et al. ⁽³³⁾ also documented an effect of breakfast v. no breakfast on his ‘quality of episodic memory’ factor. Age was not included as a covariate in the analysis although the study included children aged from 9 to 16 years. The matching familiar figures test was not sensitive in Cromer et al.'s⁽³⁷⁾ study of children with average IQ but more errors were made on the harder version of the task⁽³⁸⁾ and in children with lower IQ⁽¹⁷⁾ after no breakfast. Visual perception might only be susceptible to nutritional intervention in more vulnerable samples.

Therefore, breakfast v. no breakfast comparisons show some positive benefits particularly if testing occurs later in the morning, with the effect more easily discernible where tests are more demanding and consider error rates.

Comparisons of different breakfasts

Fewer studies find effects on cognitive function when different breakfasts are compared. In total, nine studies compared at least two breakfasts that provided solid food, for example, cereals, doughnuts or toast. However, published studies do not always provide sufficiently detailed descriptions of the breakfasts that were administered to permit calculation of exact energy and macronutrient composition. Two studies^{(22, 33)} compared cereal breakfasts with similar energy contents but which varied in terms of carbohydrate quality. Wesnes et al. ⁽³³⁾ included cereals that varied in amount of complex carbohydrate and Mahoney et al. ⁽²²⁾ compared an oatmeal breakfast against a ready-to-eat cereal. The two breakfast comparisons in each study were advantageous relative to the no-breakfast and glucose-drink conditions but performance did not differ between the cereal breakfasts in either study.

Only one study has explicitly compared the glycaemic index (GI) of two breakfasts⁽³⁹⁾, but details of GI determination were not provided. GI may be calculated or estimated from international tables of GI values⁽⁴⁰⁾, but potential differences in the glycaemic responses of adults and children are not well understood since GI studies have not been conducted in children. Glycaemic load (GL) of breakfasts was considered in one study of 6- to 7-year-old children⁽⁴¹⁾. GL is calculated by multiplying the amount of available carbohydrate in a food item by the GI of the food and dividing this by 100. The three breakfasts provided by Benton et al. ⁽⁴¹⁾ varied in GL from 2·5 to 17·86 but more importantly the energy content of the breakfasts prescribed also varied from 657 to 820 kJ and actual intake was not consistent between participants.

Time of testing post-consumption may be important. The strongest effects have been found in the late morning, at+130 min⁽³⁹⁾ and+200 min⁽³³⁾. Mahoney et al. ⁽²²⁾ tested only at 1 h post-consumption and found an effect of breakfast v. no breakfast but no strong effects between two types of breakfast. Pollitt et al. ^{(17, 35)} administered one cognitive test battery at+180 min and found effects in children with lower IQ only. Benton et al. ⁽⁴¹⁾ tested children between 140 and 210 min post-ingestion, with the possibility that the period between ingestion and testing varied between conditions and children.

There is less robust evidence for effects across other cognitive domains or between breakfasts providing similar energy but varying in macronutrient composition, GL or GI. It is difficult to confirm effects of breakfast quality in well-nourished children in mainstream education based on these acute studies.

Studies of acute effects of breakfast in children of differing nutritional status

There are seven studies that examined the effects of providing breakfast to children of differing nutritional status (see Table 3). Four studies were performed in South America, two in Jamaica and one in India. These studies compared the effects of breakfast in well-nourished and stunted and/or wasted children or children considered nutritionally at risk. Nutritional status was classified on the basis of height for age ( − 1 sd) and weight for age ( − 0·5 sd) in four studies. These studies showed that cognitive performance was better following breakfast in the at-risk or undernourished group, with few if any effects on the well-nourished and not-at-risk control children. López et al. ⁽⁴²⁾ reported more errors in stunted children irrespective of treatment. However, some authors (for example, Grantham-McGregor⁽¹⁵⁾) do not support the view that nutritional status is an important determinant of the effect of breakfast on cognitive function. In studies where details were provided, breakfasts delivered a substantial proportion of the energy requirements for these children.

Cognitive tasks susceptible to nutritional intervention in nutritionally vulnerable children appear to be verbal fluency and memory tasks, particularly short-term recognition (a hippocampal task), Sternberg memory search, as well as the matching familiar figures test (a measure of visual perception). It would appear that nutritionally at-risk children are more vulnerable in terms of memory performance and that these effects are not evident in geographically matched well-nourished control children.

Studies of long-term effects of school breakfast programmes and breakfast clubs

There are thirteen studies that examined breakfast provision at school (see Table 4). Generally, these evaluated government-funded breakfast provision which was free to low-income children. Of the studies, seven were conducted in the USA in children of low socio-economic status, one study took place in the UK and the rest were undertaken in South America (in undernourished, at-risk children), South Africa and Jamaica. The interventions employed were predominantly school breakfast v. no school breakfast (either breakfast at home or no breakfast). The duration of the school breakfast programme during these studies ranged from 4 weeks to 3 years but the majority of evaluation studies had a duration of 6–12 weeks. The children in these studies tended to be younger (between 3 and 8 years of age) than in other studies included in the present review. No studies considered effects in adolescents.

Scholastic achievement tests were used as measures of cognitive function in seven studies. These studies, taken together, showed improvement mainly in mathematics or arithmetic scores post-intervention. School breakfast programmes were associated with increased attendance or decreased absenteeism, a possible explanation for the improved performance. This is especially likely where scholastic achievement tests are employed as post-intervention measures. Benefits were not greater or confined to undernourished or at-risk groups in studies that also included well-nourished controls with one exception⁽⁴³⁾.

In seven studies specific tests of cognitive function were employed. Of these, one study⁽⁴⁴⁾ showed improved memory, another reported improved concentration⁽⁴⁵⁾ while a third found no effect on a range of tests⁽²³⁾. Worobey & Worobey⁽²⁴⁾ found a positive effect of a school breakfast programme on a range of mainly spatial cognitive tests in two separate samples of school children. Cueto & Chinen⁽⁴⁶⁾ report acute effects of breakfast provision in full-grade and multiple-grade schools, confined to tests of memory, arithmetic and, to a lesser degree, reading in the multiple-grade schools. Multiple-grade schools include children of different ages within the same class. They are associated with more poverty and lower achievement than full-grade schools and are therefore more likely to include children who are nutritionally at risk.

Provision of breakfast at school seems to have positive effects in all but two studies, particularly in these younger children who were participating in a free, universal school feeding programme. Despite the lack of detail of the nutrient composition and energy provided by the breakfasts at school, the effects of breakfast seem positive. These effects could, however, be an artifact of the increased school attendance that such provision encourages. One pitfall of school breakfast programmes is the potentially negative impact of breakfast provision on class time and pupil–teacher contact. This depends on whether breakfast is provided before the school day or during time normally allocated for teaching. In some studies, particularly in the developing world, breakfast was provided during teaching time⁽⁴⁶⁾.

Studies of effects of habitual breakfast quality

There were only four studies that considered habitual breakfast intake (see Table 5). Of the studies, three were conducted in Spain, utilising food diaries to determine the quality of breakfast intake based on target food groups. These studies suggest a positive effect of breakfast quality on scholastic performance. It was indicated by one study that snack provision could ameliorate the negative effect of a poor-quality or no breakfast⁽⁴⁷⁾. The studies did not include measures of cognitive function other than school performance.

Discussion

The present review has identified relatively few good-quality studies that examine the effects of breakfast on the cognitive performance of school-aged children. We identified forty-five studies presented in forty-two articles published between 1950 and 2008 as suitable for inclusion, although many lacked scientific rigour. This is despite intense public and scientific interest and confident claims in the media on behalf of governments and the nutrition industry regarding the effects of breakfast. The majority of the studies reviewed were sponsored in whole or in part by industry. A recent development is the consideration of the impact of habitual breakfast on performance. The four studies in this category were all conducted in the last decade. Across all categories there is a predominance of studies in younger children and far fewer in adolescents in whom metabolic and cognitive effects could be different.

Overall, the quality of studies was poor. Some studies were not counterbalanced or allocation to condition was not randomised. It was sometimes unclear whether testing was performed blind to treatment condition where possible. Socio-economic status, if specified and not deliberately selected for, was predominantly middle class and monetary incentives were provided to parents in a number of studies. Studies included in Table 2 assume that the children were well nourished since they were all described as healthy and no consideration of the nutritional or weight status of the sample was provided. However, it is probable that samples included children across a range of body weights but these were not reported. Weight is likely to be positively skewed, reflecting the distribution of body weight in the populations from which they were recruited (i.e. predominantly white, middle class). Quality assessment scores reflected almost the whole range of possible scores (range 4–17). Differences in quality scores by category were small and decade of publication did not appear to influence study quality. However, habitual breakfast studies tended to score lower than other categories probably due to confounds inherent in these designs. Lifestyle factors are difficult to account for, socio-economic status may be associated with breakfast quality, and the free-living nature of the studies and reliance on food diaries reduce experimental control.

Arguably, studies in undernourished children are more difficult to design and execute. However, the studies of undernourished or at-risk children reviewed were equally well performed, scoring across a similar range to the acute healthy investigations (9–16 out of a maximum 18). The majority employed repeated-measures designs (five studies), details of breakfasts provided were good, the cognitive tests employed comparable and, where stated, the analysis was appropriate.

Difficulties also present in evaluations of school breakfast programmes. These are logistically challenging to conduct since it is difficult to match samples and the introduction of contamination between treatment arms can have a serious impact on the study. For example, Shemilt et al. ⁽⁴⁵⁾ experienced high attrition and contamination, such that participants in the intervention and non-intervention arms became aware of each other's condition. Therefore, the data could not be analysed using the intended method.

Methodological issues

Breakfast intervention

Breakfast manipulations were often not matched for energy or volume and no mention is made of palatability, which could affect cognitive, affective and behavioural responses. Studies varied in the amount of each breakfast consumed, rendering conditions unmatched in terms of energy content when the variable of interest was carbohydrate quality or GL. In this respect studies were generally poorly conducted. Few firm conclusions can be made from studies of school breakfast programmes which do not tend to record or report the nature or quantity of the breakfast. Habitual breakfast studies rely on food diaries which are subject to the same bias of under- and over-reporting in children as in adults⁽⁴⁸⁾.

Often, choice of breakfast intervention was not driven by a priori hypotheses about the mechanisms by which breakfast could have an impact on cognitive performance. School breakfast programme evaluations are bound by the constraints of the programme. Other studies may be guided by the commercial interests of the study sponsor. However, recent studies have attempted to evaluate whether food characteristics such as GI and GL are related to the effects of breakfast on cognitive function^{(39, 41)}. Ingwersen et al. ⁽³⁹⁾ found advantageous performance after a low-GI breakfast whereas Benton et al. ⁽⁴¹⁾ report benefits after low-GL breakfasts. Moreover, Benton et al. ⁽⁴¹⁾ calculated the GL of the breakfasts used in Ingwersen et al.'s⁽³⁹⁾ study and report that the low-GI breakfast was also low GL. Hence it is difficult to attribute the effects observed to GI or GL. There is also no agreement as to the relative importance of GI v. GL in terms of which best predicts health outcomes in adults⁽⁴⁹⁾. Importantly, these studies have not profiled the metabolic response to breakfasts varying in GI or GL in children and the assumption is that the responses are the same as in adults.

Cognitive performance testing

Some issues relate to the appropriateness of cognitive testing employed by the studies to date in terms of their difficulty level, style and cognitive domain under study. Moreover, the suitability of some of the tests employed for children of different ages, developmental stages and intellectual level is often not considered. The cognitive tests employed in experimental studies were fairly limited, and these were not necessarily selected for their sensitivity to nutrient intervention or change over time. Some studies used global neuropsychological tests, more usually employed for diagnostic purposes. Across studies, tests were not readily comparable and accuracy and error rates were not provided by all studies. Although ecologically valid, end-of-year school performance may not provide the most sensitive indicator of the effect of a school breakfast programme. Many studies do not control for other factors which are likely to influence school grade, including home environment, parental involvement, school system and quality.

Performance on many of the cognitive tasks is evaluated in terms of accuracy scores. Little consideration is given to motivation and effort including the ability to sustain performance over time. However, breakfast consumption might facilitate motivation and reduce the underlying ‘maintenance costs’ of sustained performance. Sustaining concentration and retaining information are cognitive processes of key importance for scholastic achievement. The cost of maintaining these processes may vary between children and this may be a partial explanation why positive effects of breakfast are most easily identifiable in nutritionally at-risk children. Thus, future assessment of cognitive performance should include measures of motivation, such as number of trials attempted or frustration tolerance, in addition to accuracy.

Cognitive performance testing: design

Studies in adults suggest that glucoregulation following macronutrient manipulation, rather than absolute levels of blood glucose, may be most important for cognition^{(50, 51)}. Better glucoregulation has been associated with superior short-term and delayed verbal memory⁽⁵²⁾. Smaller blood glucose excursions, rather than sharp fluctuations, reflect lighter metabolic stress and may be better for cognition. If glucoregulatory processes moderate the relationship between food and cognition in children, then repeated-measures designs are necessary. Participants are likely to demonstrate inter-individual variability in glucoregulation, and independent-groups designs will not control for the potential error that this introduces. This issue is of particular importance in studies of children because of their relatively faster metabolism⁽¹³⁾.

Cognitive performance testing: analysis

In some of the studies reviewed, the statistical analysis was inappropriate to the design of the study. Often more powerful analyses could have been performed. Some studies failed to describe the statistical procedures and no or limited critical values were reported, precluding meta-analyses of these data. The age of the study had some bearing on the complexity of the statistical analysis performed, with older studies tending towards simpler, non-parametric statistical procedures, possibly due to the lack of available computing power.

Mechanisms

Mechanisms for the facilitation of cognitive performance by breakfast are not well established⁽⁸⁾. There are likely to be different mechanisms of action responsible for short- and long-term effects of breakfast consumption. Moreover, these purported mechanisms may not necessarily be the same in adults as in young children, whose brain metabolic requirements are relatively much greater than those of the adult brain. In addition, metabolic processes of adolescents' brains begin to resemble those of adults during puberty but do not reach adult levels until 16–18 years⁽¹³⁾.

Mechanisms: physiological

While some studies suggest that glucose ingestion facilitates cognitive performance^(53–56), others report no direct relationship between performance and glucose levels following breakfast consumption^{(57, 58)}. Rather than a direct effect of glucose on cognitive performance, the mechanism of action may involve one or many correlates of blood glucose. Glucose ingestion gives rise to changes in levels of acetylcholine, insulin, serotonin, glutamate and cortisol, all of which can affect cognitive function^(58–61). It is possible that any or a combination of these central and peripheral changes are involved in the impact of breakfast on performance.

Therefore, we are currently lacking a strong theoretical or evidence base to relate specific neurochemical or physiological activity to specific cognitive functions. These potential biomarkers for cognitive function are more difficult to measure than the actual function itself. This is particularly pertinent in children in whom invasive measures are ethically and practically difficult. Some physiological measures of autonomic nervous system activity such as heart rate and electroencephalogram which have been shown to vary in relation to cognitive demand⁽⁶²⁾ may be more easily applied in studies of children. Appropriate methods to track biomarkers during cognitive activity are required in order to elucidate the mechanisms by which breakfast may affect performance.

The use of physiological biomarkers relies on the assumption that peripheral measures reflect central activity. Indeed, peripheral glucose and its metabolites are well regulated in healthy individuals to maintain homeostasis and elicit appropriately rapid postprandial responses. Peripheral glucose regulation and metabolism in children may not be the same as in adults⁽⁶³⁾.

In the long term, breakfast consumption may lead to beneficial physiological changes in nutrient status. Thus positive effects of breakfast on cognitive performance may be the product of better nutritional profiles rather than transient changes in blood parameters. Achieving better nutrient and vitamin status or rectifying deficiencies, as demonstrated by some studies in the developing world, may be responsible for the effects seen on performance.

Behavioural mechanisms

While physiological explanations for the effects of breakfast on cognitive performance are appealing, behavioural mechanisms may play an important role in the cognitive response to breakfast. In the short term, breakfast consumption may function to heighten subjective feelings of alertness and motivation to concentrate and learn. This could occur because of the learned association between breakfast consumption and feelings of wellbeing, or the reduction of hunger (for example, Wesnes et al. ⁽³³⁾, Dye & Blundell⁽⁹⁾ and Gibson⁽⁵⁸⁾). Unfortunately, few studies incorporate measures of both appetite and mood alongside objective cognitive measures.

In the long term, studies indicate that school breakfast programmes increase scholastic performance. This outcome could be a direct effect of the repeated consumption of breakfast and the development of a learned association of this with wellbeing or ability to concentrate. Alternatively, positive effects could be a result of the improved nutritional state which is known to result from regular breakfast consumption^{(1, 64)}. The effects seen in response to school breakfast programmes could simply be explained as an artifact of increased school attendance, motivated to attend by the provision of the breakfast^{(23, 29, 46)}.

Conclusions

Overall, evidence suggests that breakfast consumption has generally positive effects on cognitive performance in comparison with breakfast omission. This effect appears to be pervasive in both acute studies and longer-term school breakfast programmes. However, the apparent beneficial effects of school breakfast programmes may be linked to increased attendance and reduced absenteeism, and effects of such provision in older children are not known. In addition, breakfast effects are more easily demonstrable in nutritionally vulnerable children.

It is difficult to recommend an optimal breakfast for cognitive function based on the currently available research. One study has suggested that solid breakfasts may be advantageous over liquid breakfasts⁽³³⁾, perhaps due to differing rates of gastric emptying. In addition, some studies indicate that low-GI or low-GL breakfasts may confer benefits but it is hard to differentiate between these two indicators of glycaemic response in the few studies that have examined these food characteristics.

The majority of studies have concentrated on the measurement of memory and attention performance, with less examination of tasks that engage other cognitive domains, processes and aptitudes. There have been few examinations of problem solving and psychomotor skill, for example. However, from the studies reviewed, it is difficult to conclude which specific cognitive domains are most sensitive to nutritional manipulations at breakfast time, although there is most abundant support for effects on memory.

Recommendations for future work

The present review has highlighted problems in this research area which could be addressed by the following recommendations for future work.

Future studies should test focused hypotheses, based on a small number of theoretically selected breakfast conditions, presented according to carefully counterbalanced repeated-measures designs with large samples which yield sufficient power. Breakfasts must be matched for energy but should differ sufficiently in key features, for example, macronutrient composition, style or glycaemic response, in order to experimentally investigate potential mechanisms of action. Tasks that span a wide range of cognitive domains with demonstrated sensitivity to nutritional manipulations should be used so that null findings can be ascribed to true lack of effect rather than test insensitivity. In acute studies, effects of breakfast may be more pronounced some time after consumption when the metabolic challenge of food consumption has subsided. Late-morning testing rather than immediate post-breakfast testing merits further exploration. Enduring and meaningful effects are likely to result from chronic interventions of at least 12 weeks with appropriate timing of the tests of cognitive and/or scholastic performance. Lastly, the present review has highlighted the need for greater examination of the effect of breakfast in adolescents.