Breakfast

By far the largest number of included publications (fifteen: eight from the USA, three from the UK, one each from France, Chile, Israel and Sweden) examined the effect of breakfast. Eleven of these studies focused on children within the 4–12-year age bracket, whilst the remaining four included a teenage population. Generally the size of the study populations was small (predominantly < 100 participants) and of short duration ( < 1 week; these studies were included due to the high carbohydrate content of the breakfast intervention). All studies examined predominantly healthy, mixed sex populations from mainstream education. Most participants (where recorded) were white Caucasian, with the exception of those participating in three studies⁽Reference Meyers, Sampson, Weitzman, Rogers and Kayne^11, Reference Lieberman, Hunt, Coulson, Clark, Swendseid and Ho^19, Reference Morrell²⁰⁾, who were predominantly black, mixed race and Hispanic, respectively. Ten studies reported socio-economic status (using a variety of scales and indicators); of these, half recruited children from low-income families and half recruited children from middle- to high-income families.

The type of breakfast interventions investigated fell into five categories: (1) the impact of the provision of breakfast clubs⁽Reference Meyers, Sampson, Weitzman, Rogers and Kayne^11, Reference Lieberman, Hunt, Coulson, Clark, Swendseid and Ho^19, Reference Wahlstrom and Begalle^21, Reference Shemilt, Harvey, Shepstone, Swift, Reading, Mugford, Belderson, Norris, Thoburn and Robinson²²⁾; (2) breakfast consumption v. fasting⁽Reference Dickie and Bender^9, Reference Pollitt, Leibel and Greenfield^13, Reference Pollitt, Lewis, Garza and Shulman^23–Reference Mahoney²⁶⁾; (3) low- v. high-protein breakfast consumption⁽Reference Morrell²⁰⁾; (4) low- v. high-energy breakfast consumption⁽Reference Cromer, Tarnowski, Stein, Harton and Thornton^8, Reference Wyon, Abrahamsson, Jartelius and Fletcher¹⁸⁾; (5) habitual breakfast consumption v. a standardised breakfast consumption⁽Reference Michaud, Musse, Nicolas and Mejean^27, Reference Vaisman, Voet, Akivis and Vakil²⁸⁾. Of the studies investigating breakfast clubs, three identified a small but positive impact of breakfast club participation on a selection of educational outcomes, whilst the remaining study found no effect. Four of the six studies investigating breakfast consumption v. fasting identified some improvements (P = 0·05) in problem solving, attention and epidsodic memory after cereal consumption and complex visual display after consuming breakfast, although the remaining two studies were unable to identify any significant differences. The study investigating the protein content of breakfast was unable to demonstrate any significant differences between consumption of high- or low-protein breakfasts. When energy content was examined, one study demonstrated a disadvantage (P = 0·05) of a low-energy breakfast in terms of mood, physical endurance and creative thinking; however, the other study found no significant differences between the two breakfast conditions. When habitutal breakfast conditions were compared against a standardised breakfast provision, both studies demonstrated some improvements to cognitive function after consumption of the standard breakfast. However, although one study⁽Reference Michaud, Musse, Nicolas and Mejean²⁷⁾ demonstrated an increase in memory scale results after consumption of the standard breakfast, this was accompanied by a negative reduction in concentration.

The diverse range of different breakfast interventions and research designs makes it very difficult to draw together the findings presented. The majority of studies (ten out of fifteen), however, were able to demonstrate that the provision of breakfast may have some small benefit to a limited selection of short-term behavioural and cognitive functions. However, this very generalisable overview must be considered within the context of the numerous shortcomings of many of these studies: (1) only three studies considered the impact of habitual diet, which particularly given the very short duration of many of these studies may have had a substantial effect; (2) in general, many of the studies failed to adjust for, or even acknowledge, many of the factors, which like habitual diet, may impact significantly upon the findings presented; (3) there are a huge number of behavioural and cognitive indices that are important in measuring educational attainment – however, most studies selected a limited range of subjective and objective methodologies, many lacking appropriate validation, with very little consistency in methodology between studies; (4) breakfast clubs often have a social dimension to them, for example, some parents may use them primarily because they provide childcare, so it is difficult to dissociate the different benefits; (5) as mentioned previously, many of the studies were of short duration, making it extremely difficult to control for different rates of emotional and neurological maturation, as well as being unable to quantify sustainability and longer-term benefits. Given these shortcomings, it was unsurprising that many of the studies gained very low quality-assurance scores, with eight achieving a SIGN grading of under ten (out of a possible twenty-seven) and an EPPI weighting of low.

Sugars

Six of the included studies⁽Reference Wolraich, Lindgren, Stumbo, Stegink, Appelbaum and Kiritsy^17, Reference Wolraich, Milich, Stumbo and Schultz^29–Reference Wender and Solanto³³⁾ (five from the USA and one from Canada) investigated the effect of sugar intake on learning and behavourial outcomes in school-aged children. All studies were carried out between 1985 and 1994, and examined a population of pre-teens (3–12 years) with the exception of one study⁽Reference Saravis, Schachar, Zlotkin, Leiter and Anderson³²⁾ that included teenagers. Every study examined the effect of sucrose against aspartame or aspartame and saccharin in an RCT cross-over design. The period of exposure to the sugar or sweetener substitutes was short in most cases (1–2 d) with the exception of two studies⁽Reference Wolraich, Lindgren, Stumbo, Stegink, Appelbaum and Kiritsy^17, Reference Rosen, Booth, Bender, McGrath, Sorrell and Drabman³¹⁾ which examined each exposure over 1 week.

Only one study examined the effects of sugar in an entirely healthy population of children. The remaining studies were investigated in a population of chidren with symptoms of attention deficit hyperactivity disorder (ADHD), either on their own (two studies) or alongside control healthy children (three studies). The two small studies (population n 16 in each) carried out in an all-male ‘ADHD’ population found no significant difference between the sugar and aspartame exposure on any of the objective or subjective outcomes measured. Of the remaining studies, two identified small statisticially (P = 0·05) (but often not clinically) significant adverse effects on behavioural outcomes in the sucrose group, whilst another could only identify a disadvantage in the ‘ADHD’ subgroup. The only long-term diet that combined the sugar or sweeteners into habitual diet rather than a drink was unable to identify any difference between the sucrose and sweetener in primary-aged children. This was, incidentally, the only study also to examine any blood biochemistry. However, a small but significant improvement was seen in a small number of the subjective behaviour scores for the infants, although this finding was inconsistent with a significant decrease in the score for one of the objective performance measures in the same group.

Whilst all but one study gained reasonable quality-assurance scores (for both EPPI and SIGN), only two studies calculated power. As with the breakfast studies, very little attention was paid to important factors, particularly physical activity and habitual diet, both of which could significantly impact upon the metabolic responses to sugar consumption. Given the small number of studies included and diversity of the populations examined, it becomes difficult to derive any conclusive outcome from these studies. However, the data presented do suggest that short-term exposure to sucrose has no dramatic detrimental effects on educational and behavioural outcomes in school-aged children, when compared with commonly consumed artificial sweeteners.

Fish oil supplementation

A total of five included studies⁽Reference Richardson and Montgomery^14, Reference Voigt, Llorente, Jensen, Fraley, Berretta and Heird^34–Reference Hirayama, Hamazaki and Terasawa³⁷⁾ examined the effect of fish oil supplementation on learning and behavioural outcomes. All studies were carried out during the last 5 years and used a placebo-controlled RCT design in a population aged between 5 and 13 years, with symptoms of neurodevelopmental disorders (dyspraxia and ADHD). The study population sizes were small, ranging from forty to 117 particpants, all of which included both boys and girls although boys predominated, probably due to the prevalence of male ADHD diagnoses. Studies were carried out for periods between 2 and 4 months and took place in the USA (two studies), UK (two studies) and Japan (one study). The two UK and one of the USA studies used fish oil, the other USA study used an algae-derived oil, and these were taken as capsules, whilst the Japanese study incorporated the test fish oil into pre-prepared food and drink. The fatty acid composition selected for each study differed, with four out of the five studies using a mixture rich in DHA, whilst a recent study⁽Reference Richardson and Montgomery¹⁴⁾ used a composition rich in EPA, which also contained γ-linolenic acid (GLA).

Whilst all studies gained medium to high quality ratings, the outcome assessment measures and results varied considerably. Only two of the studies measured blood biochemistry alongside objective testing and subjective parental and teacher observations. However, despite concurrent increases in blood concentrations of the long-chain n-3 fatty acid DHA in the treatment groups, one study⁽Reference Voigt, Llorente, Jensen, Fraley, Berretta and Heird³⁴⁾ found no significant differences in the behavioural and educational outcomes between the treatment (345 mg DHA/d) and control groups. One study⁽Reference Stevens, Zhang, Peck, Kuczek, Grevstad, Mahon, Zentall, Arnold and Burgess³⁶⁾ using a dose of 480 mg DHA, 80 mg EPA, 96 mg GLA and 40 mg arachidonic acid (AA) per d detected a small improvement in just two out of sixteen subjective parental and teacher observations after secondary intention-to-treat analysis, although blood fatty acid concentrations were shown to correlate significantly with further parent- (P = 0·05) and teacher-rated (P = 0·03) behaviours. Of the remaining three studies, one⁽Reference Richardson and Puri³⁵⁾ showed a small statistically significant (P = 0·05) improvement in three out of fourteen subjective parental behaviour scores in the fish oil plus other fatty acids (186 mg EPA, 480 mg DHA, 96 mg GLA and 42 mg AA per d) supplemented group. One study⁽Reference Hirayama, Hamazaki and Terasawa³⁷⁾ found no significant difference between treatment (DHA 3·6 g/week or 514 mg/d) and placebo groups, apart from a small significant improvement in continuous performance (P = 0·001) and visual short-term memory (P = 0·02) in the control group only, over 2 months. However, they later reported⁽Reference Hamazaki and Hirayama³⁸⁾ a significant (P = 0·001) reduction in aggression in the DHA treatment group. A recent study⁽Reference Richardson and Montgomery¹⁴⁾ was the only study to report consistent significant (P = 0·05) improvements in both objective and subjective behavourial and educational outcomes assessed in the EPA-rich treatment group (558 mg EPA, 174 mg DHA, 60 mg GLA per d).

There remain too many inconsistencies between the studies included in the present review to reliably inform any conclusion, raising several areas of concern, including: (1) most studies were carried out in children with varying degrees of neurodevelopmental disorders, which may raise questions over the applicability of these data to mainstream children without any developmental complications; (2) the trials were carried out in different populations, likely to have differences in baseline n-3 long-chain fatty acid status; (3) there was a lack of biochemical data – whilst the extraction of blood samples from child populations raises many ethical implications, further metabolic and neurophysiological analyses are required in order to place these findings into any mechanistic context and identify differences in baseline status between different populations; (4) the dosage used in these studies are inconsistent and unlikely to be achieved through dietary means in the UK; (5) whilst the fish oils consumed in the aforementioned studies appeared to be well tolerated, there is currently no national UK dietary reference value and it is essential that optimal dose, duration and fatty acid compositions be established.

Since the date of the search for studies for the present review, several new RCT studies have been reported at scientific meetings and have been published. At the International Society for the Study of Fatty Acids and Lipids (ISSFAL) 7th Congress in July 2006 two research groups reported on the effects of EPA+DHA supplementation in primary schoolchildren (6–10 years). One study⁽Reference van Klinken, Wilson, Lukito and Osendarp³⁹⁾ in South Australian and Indonesian children found no consistent effects on cognition in well- and marginally nourished children in a total of 780 children over 12 months at a dose of 88 mg DHA/d and 22 mg EPA/d. The other study⁽Reference Smuts, Dalton, Witthuhn and Wolmarans⁴⁰⁾ in 355 South African children over 6 months at a dose of 274 mg n-3 long-chain-PUFA/d found not only improved verbal learning, memory and spelling ability of subjects, but also lessening of the number of days they were absent from school. The third study, now published⁽Reference Sinn and Bryan⁴¹⁾, reported the effects of an EPA (558 mg), DHA (174 mg) and GLA (60 mg) daily supplement in 132 Australian children aged 7–12 years with ADHD over 15 weeks. They found no significant treatment effects on the Conners Teacher Rating Scales but significant effects (between P = 0·05 and P = 0·01 on the various scores) on the Conners Parent Rating Scales. Another recently published RCT⁽Reference Birch, Garfield, Castaneda, Hughbanks-Wheaton, Uauy and Hoffman⁴²⁾ was a follow-up study at 4 years of age of breast-fed infants and infants supplemented with and without DHA and AA formula. DHA- and AA-supplemented infants had similar visual acuity and intelligence quotient (IQ) to that of breast-fed infants whilst children who were fed formula containing no DHA and AA had significantly poor visual acuity and verbal IQ at 4 years of age.

Vitamin and mineral supplementation

Two relatively recent studies⁽Reference Southon, Wright, Finglas, Bailey, Loughridge and Walker^16, Reference Schoenthaler, Bier, Young, Nichols and Jansenns⁴³⁾ examined the effect of low-dose multivitamin and mineral supplementation on measures of IQ, assessed by the established Wechsler Intelligence Scale for Children revised (WISC-R) test. Both studies took place over several months, were well conducted and gained high quality-assurance assessments. The larger and more recent of the two studies⁽Reference Schoenthaler, Bier, Young, Nichols and Jansenns⁴³⁾ utilised an RCT design in a USA population comprised of predominantly Hispanic working-class pre-teens (6–12 years). The authors reported a moderate, but statistically significant, average increase in the non-verbal IQ (P = 0·038) of children from the active supplement group. However, further matched pairing analyses indicated that this may have been accounted for by a substanital net IQ increase from just a small sub-sample of these children. Analysis of blood biochemistry and/or habitiual diet would have assisted the interpretation of these findings. The other smaller, slightly older, case–control study⁽Reference Southon, Wright, Finglas, Bailey, Loughridge and Walker¹⁶⁾, carried out in a British population of teenagers (13–14 years), reported no significant effect of supplementation on verbal and non-verbal IQ, although this study may have been under-powered. Interestingly, however, blood biochemistry of the participants demonstrated a significant association between plasma ascorbic acid and non-verbal IQ of boys (P < 0·001) and whole-blood glutathione peroxidase activity and non-verbal and verbal IQ of both sexes (P = 0·05). This highlights the importance of examining baseline nutritional status before the functional significance of vitamin and mineral supplementation can be determined.

Although well conducted, these two studies alone provide insufficient evidence as to the effects of low-dose vitamin and mineral supplementation on the IQ score of schoolchildren from the developed world. As both of these studies indicate that particular subgroups of children may be more ‘at risk’, it is imperative that future studies determine baseline and habitual nutritional status. As there may also be potential sex and age differences, it remains important to assess mixed sex populations powered to represent different age groups.

Other

One study⁽Reference Chavez and Martinez⁴⁴⁾ examined the effect of a supplemented diet of ‘good food’ v. no supplementation in a very small group of deprived Mexican schoolchildren during their first year of elementary school. Whilst significant improvements were observed in the examination results and behaviour of those children who had received the supplemented diet, this study lacked detail, quality and rigour, making it difficult to draw any reliable conclusions.

Comparison of UK and non-UK data

Only six RCT or case–control trial studies were included from the UK (three breakfast studies⁽Reference Dickie and Bender^9, Reference Shemilt, Harvey, Shepstone, Swift, Reading, Mugford, Belderson, Norris, Thoburn and Robinson^22, Reference Wesnes, Pincock, Richardson, Helm and Hails²⁵⁾, two fish oil studies⁽Reference Richardson and Montgomery^14, Reference Richardson and Puri³⁵⁾ and one multivitamin study⁽Reference Southon, Wright, Finglas, Bailey, Loughridge and Walker¹⁶⁾), all of which were carried out during this decade, with the exception of one study⁽Reference Dickie and Bender⁹⁾. As there were not sufficient UK studies, the search strategy was expanded to include other countries categorised as ‘highly developed’ according to the World Bank criterion. It is, however, important to acknowledge the potential dietary, economic, social, educational and cultural differences that may exist between countries. Indeed, the different measures of educational attainment between the countries made it difficult to draw parallels across the datasets. The present review highlighted the requirement for further high-quality, longer-term research to examine the effect of diet on the educational attainment of UK schoolchildren.

Quality-assurance issues

In general the quality scores for the studies investigating sugar, fish oils and multivitamin and mineral supplementation were higher than those investigating breakfast. It is, however, important to recognise the difficulties of delivering a complex intervention such as breakfast within the rigorous framework of an RCT. There were, however, several quality-assurance issues arising generally across the studies included in the present review.

1. (1) Factors that could affect the external validity of the studies included in this systematic review.

   Whilst nearly all studies took account of age and sex, ten studies accounted for body weight, six recorded ethnicity and seven adjusted for disease status (although the majority of these studies were in children with ADHD-like behaviours). Only five studies considered habitual nutritional and socio-economic status (four of these also accounted for parental education) whilst just two studies considered family status. As discussed previously, habitual dietary intake may have a significant effect, particularly in the studies of short duration, where the body is physiologically adapted to override any short-term stresses. Although logistically difficult to measure in some circumstances, family status may also impact upon measures of educational attainment, through a complex interplay of socio-cultural modifying effects. Whilst not within the scope of the present review, family status may also be linked to other early life factors such as low birth weight, weaning, breast-feeding, nutrition in utero and incidence of illness, which have also been shown to be implicated in later educational attainment⁽Reference Gregory, Lowe, Bates, Prentice, Jackson, Smithers, Wenlock and Farron^4, Reference Grantham-McGregor, Walker and Chang⁴⁵⁾.

   Only one study accounted for smoking status, whilst no studies reported alcohol intake, illegal drug misuse or physical activity levels. Whilst it is unlikely in very young children that smoking, alcohol and drug misuse are prevalent, it would be prudent to control for these measures, particularly in older children, given the potential influence on educational and behavioural measures⁽Reference Gregory, Lowe, Bates, Prentice, Jackson, Smithers, Wenlock and Farron⁴⁾. Physical activity levels have also been shown to positively impact upon cognitive performance in school-aged children⁽Reference Tomporowski^46–Reference Hillman, Castelli and Buck⁴⁸⁾ and must consequently be considered as having a potentially important effect in all age groups. Similarly, no studies adjusted for locality and context, although it is accepted that these factors are highly complex and may be very difficult to control for particularly within ‘free-living’ situations. However, if studies are carried out under tightly controlled conditions such as a research laboratory, it raises questions over the potential impact of stresses that may be encountered within this artificial environment and the applicability of the findings to ‘real-life’ situations.

2. (2) Maturation and development.

   Although the majority of studies focused on the pre-teen population, only two studies⁽Reference Cromer, Tarnowski, Stein, Harton and Thornton^8, Reference Lopez, de Andraca, Perales, Heresi, Castillo and Colombo²⁴⁾ formally assessed pubertal development. The mean age at entry to puberty in UK females is 11·2 years⁽Reference Tanner⁴⁹⁾, with the current definitions of normality for entry into puberty being aged 8 years or more in females and 9 years or more in males. Therefore a significant number of studies would have included pre-pubertal and pubertal children. It is now acknowledged that brain and cognitive development progresses through puberty⁽Reference Blakemore and Choudhury⁵⁰⁾ and studies assessing cognitive function should, if possible, take account of this (particularly in short-term studies).

3. (3) Measurement of educational outcomes.

   There was little consistency in methodological assessments of educational attainment (including behaviour). The included papers gave rise to a vast array of different objective and subjective measures of behavioural and educational outcomes, often lacking in appropriate validation. This variety of assessment tools may have, at least in part, been accountable to international differences; however, it may also reflect the vast range of indices that make up an educational achievement, i.e. the complex interaction of mood, motivation, knowledge, application and capability, to name but a few.

4. (4) Power, concealment, compliance, randomisation and blinding.

   Statistical power was calculated in just five of the twenty-nine included studies. In addition to this, the degree to which concealment and compliance was reported varied dramatically, which made interpretation and comparisons between studies difficult.

Future research

The current evidence base examining the effect of diet and dietary change on educational attainment (including behaviour) remains inconclusive; it is therefore imperative that future research is undertaken. More research is required within the UK to investigate the effects of supplementation (for example, vitamins and fish oils), specific nutrients (for example, sugar) and complex meals (for example, consumption and nutritional composition of breakfasts and school meals). However, the limitation of existing studies should inform future research, in order to improve the quality and depth of evidence:

1. (1) Additional longer-term or -duration studies are required to assess long-term impact, sustainability and overcome the possibility of short-term mechanistic adaptations and differences in the rate of pubertal maturation and neurological development.

2. (2) More studies are required within older (teenage) young individuals.

3. (3) Biochemical monitoring is important, not only to establish baseline levels of nutrients under investigation, but also to assess compliance and support the understanding of mechanisms. However, withdrawal of blood samples from children requires careful ethical consideration and other alternatives such as finger-prick blood, hair and cheek-cell analyses should be considered where possible.

4. (4) Studies should be statistically powered and provide clear details regarding randomisation, blinding, concealment and compliance, where an RCT structure can be easily applied (for example, for supplementation studies). In free-living situations (for example, breakfast), that may not easily fit into the rigid confines of the RCT, extremely well-designed cohort studies should be considered as a possible alternative.

5. (5) Studies should measure and adjust for all potential counfounders, where feasibly possible.

6. (6) Authors should acknowlege any limitations of their research and use this to place their findings into context.

7. (7) There should be a broad range of investigations that include children in care, with special needs and from different ethnic minority and socio-economic groups, in order to address populations that may have different needs to the ‘mainstream’.

8. (8) Authors should base research on existing mechanistic evidence and where possible use this to place their findings into context.

9. (9) We suggest that some consideration be given to the development and use of universal validated measures of educational attainment which would assist comparisons between studies and allow for appropriate meta analyses. However, although this problem is acknowledged by experts in the field, we understand that it will be difficult to achieve a consensus.