Content uploaded by Michael Nelson
Author content
All content in this area was uploaded by Michael Nelson on Aug 10, 2016
Content may be subject to copyright.
ORIGINAL ARTICLE
The glycaemic potency of breakfast and cognitive
function in school children
R Micha
1
, PJ Rogers
2
and M Nelson
1,3
1
Nutritional Sciences Research Division, King’s College London , London, UK;
2
Department of Experimental Psychology, University of
Bristol, Bristol, UK and
3
School Food Trust, Sheffield, UK
Objectives: The aim of this study was to assess how the glycaemic potency (blood glucose (BG)-raising potential) of breakfast is
associated with cognitive function (CF) in school children, taking into account important confounders, including iron status,
underlying physiological adaptations and socio-economic status.
Methods: Sixty children aged 11–14 years were selected on the basis of having breakfast habitually. Their breakfast and any
snacks eaten on the morning of the study were recorded. They were categorized into four groups according to the glycaemic
index (GI) and glycaemic load (GL) of the breakfast: low-GI, high-GL; high-GI, high-GL; low-GI, low-GL and high-GI, low-GL
above or below the median for GI ¼61 and GL ¼27. BG levels were measured in finger-prick blood samples immediately before
and immediately after the CF tests.
Results: A low-GI, high-GL breakfast was associated with better performance on a speed of information processing (Po0.01)
and a serial sevens (Po0.001) task 90 min later; a high-GI breakfast with better performance on an immediate word recall task
(Po0.01); and a high-GL breakfast with better performance on a Matrices task (Po0.01).
Conclusions: GI, GL or both were associated with performance on the majority of the CF tests (4 of 7) used. This study describes
the macronutrient composition of breakfast that could have a positive influence on the cognition of school children, proposes
the use of both GI and GL to estimate exposure, and discusses future directions in this area of research.
European Journal of Clinical Nutrition (2010) 64, 948–957; doi:10.1038/ejcn.2010.96; published online 23 June 2010
Keywords: breakfast; glycaemic index; glycaemic load; cognitive function; school children
Introduction
The plethora of studies investigating the effects of breakfast
on cognitive function (CF) support the hypothesis that
skipping breakfast adversely affects cognition, and that
the brain is thus vulnerable to brief fasting (Pollitt
et al., 1981, 1982; Conners and Blouin, 1982; Simeon and
Grantham-McGregor, 1989; Chandler et al., 1995; Vaisman
et al., 1996; Benton and Parker, 1998; Smith et al., 1999;
Wesnes et al., 2003; Mahoney et al., 2005). Despite the
wealth of studies that have been conducted in this area, the
macronutrient composition of breakfast that could selec-
tively facilitate CF after an overnight fast is not well
established. The relationships between breakfast consumption
and CF in school children are not clear, with considerably
conflicting results in prior studies, in part due to differences
in study design and the type of breakfast consumed, as well
as lack of control for potentially important confounders,
including iron status (Halterman et al., 2001; Gordon, 2003),
variation from habitual breakfast eating habits (Lloyd et al.,
1996), timing and selection of CF tests (Dye et al., 2000), the
underlying physiological adaptations (Dolan, 2002), and
socio-economic status (Bradley and Corwyn, 2002).
Research on glucose consumption and CF (Kennedy and
Scholey, 2000; Scholey and Kennedy, 2004) suggests that
the brain may be sensitive to short-term fluctuations of
glucose supply, and thus supports the hypothesis that
glucose content may induce the memory-enhancing effects
of breakfast, by producing metabolic alterations, mainly
changes in circulating glucose levels (glycaemic response)
(Dye et al., 2000; Gibson and Green, 2002; Messier, 2004; Riby,
Received 29 September 2009; revised 26 March 2010; accepted 1 May 2010;
published online 23 June 2010
Correspondence: Dr R Micha, Department of Epidemiology, Harvard School of
Public Health, 677 Huntington Avenue, Kresge Building, Room 913, Boston,
MA 02115, USA.
E-mail: renata_micha@hotmail.com
Contributors: RM performed statistical analysis, data interpretation, selec-
ted cognitive function tests and prepared the manuscript. PR selected and
developed the cognitive function tests, mood scales and task demand
questionnaire. MN supervised the project, statistical analysis, data interpreta-
tion and manuscript preparation. All authors revised the manuscript for
important intellectual content, and approved final manuscript for submission.
European Journal of Clinical Nutrition (2010) 64, 948– 957
&
2010 Macmillan Publishers Limited All rights reserved 0954-3007/10
www.nature.com/ejcn
2004; Hoyland et al., 2008). As pure glucose will be rarely
consumed as part of a normal diet, current research has
focused on varying the carbohydrate (CHO) content of
breakfast meals, and hence their subsequent glycaemic
responses (that is, the potential mediator), to investigate the
effects of breakfast consumption on CF (Benton et al., 2003,
2007; Mahoney et al., 2005; Ingwersen et al., 2007). The effects
of CHO-containing meals on blood glucose (BG) can be
modelled by using the glycaemic index (GI) (Jenkins et al.,
1981) and the glycaemic load (GL) (Salmeron et al., 1997).
When foods with equal amounts of CHO content are
compared, foods with high-GI induce a greater rise and
fall in BG and insulin, leading to lower concentrations of
the body’s two main fuels (BG and fatty acids) in the
immediate post-absorptive period. Thus, GI reflects the
qualities of CHO in relation to digestion and absorption,
but not the amount. Conversely, GL considers both the
amount and type of CHO (that is, GL ¼GI CHO); in fact, GL
predicts BG in an approximately linear manner (that is,
higher the GL, higher the BG response) (Brand-Miller et al.,
2003). Previous studies have considered either GI or GL, but
not both when investigating the effects of different
CHO-containing breakfast meals on CF. It has been hypothe-
sized that a low-GI meal versus a high-GI meal can facilitate
performance by minimizing glycaemia fluctuations, and a
high-GL meal versus a low-GL meal can facilitate performance
by preserving the substrate (that is, higher circulating glucose
levels) for a longer period of time (1–2 h after a meal) (Benton
et al., 2003, 2007; Mahoney et al., 2005; Ingwersen et al.,
2007). Nonetheless, although for example a low-GI, high-
CHO meal and a high-GI, low-CHO meal can produce the
same GL, the metabolic effects produced by the two meals will
be different (Barclay et al., 2005). Thus, GI and GL should be
used in conjunction to best describe the glycaemic potency
(BG-raising potential) of a food or meal (ADA, 2004).
To address this important question and elucidate the
conflicting results of previous studies, we investigated
whether, after an overnight fast, breakfasts differing in
both their GI and GL are associated with differences in CF.
We focused on adolescents aged 11–14, as previous work has
focused on younger children (Wesnes et al., 2003; Mahoney
et al., 2005; Ingwersen et al., 2007). We took into account
the confounders listed above and used CF tests reported to be
sensitive to variations in the glucose supply. It could be
argued that a low-GI meal would minimize glycaemia
fluctuations and facilitate performance for longer after
breakfast consumption compared with a high-GI meal, and
that a high-GL meal would potentiate the glycaemic potency
of the meal. We therefore hypothesized that 90–120 min
after breakfast (a) a low-GI, high-GL breakfast would be
associated with the best CF, a high-GI, low-GL would be
associated with the lowest CF, and the other
two breakfast types would be associated with intermediate
levels of CF; and (b) the high-GL meals would be associated
with higher BG levels compared with the low-GL meals
(Ludwig, 2002).
Methods
Study design
Sixty pupils (24 boys, 36 girls) aged 11–14 years from two
schools in South London participated; all of them were in
good health and free from learning disabilities. Children
who said that they never had breakfast were excluded. The
study was approved by the King’s College’s Research Ethics
Committee.
Questionnaires
A screening questionnaire was filled in by the parents/carers
to determine socio-economic group (Government Statistical
Service, 1990), and to exclude children for medical reasons
(that is, anaemia or other blood disorders, food allergies,
diabetes or glucose intolerance, other acute or chronic
illnesses/diseases, colour blindness, severe learning
disabilities and mood disorders). The type of breakfast
typically consumed and the usual breakfast eating habits
were also recorded.
A self-rating mood questionnaire was developed from
the Profile of Mood States bipolar form (POMS-BI) (Lorr
and McNair, 1984) and the short form of the Activation-
Deactivation Adjective Checklist. It was modified from
previous research (Rogers et al., 1995); 22 terms were used
to assess mood, energy levels, hunger and thirst. Responses
were made on integer scales from zero (not at all) to four
(extremely) (see Appendix A for the list of 22 mood states
assessed).
In a self-reported task demand questionnaire, participants
rated how difficult, effortful and tiring they found the tests
to be, using the same rating scale.
Interview
Participants were interviewed about their eating habits, physical
activity, current health status, medication/supplements, sleeping
patterns, and menstrual status, including food and drink
consumed at home on the morning of the CF testing (‘break-
fast’); anything else eaten or drunk on their way to school or
since arriving at school (‘snack’); and the last meal they had the
night before their appointment (‘dinner’). A photographic atlas
of food portion sizes (Nelson et al., 1997) was used to quantify
the amounts of food and drink consumed. The pupils were
weighed in their school uniform, after removing their blazer and
shoes (Salter scale) and measured for height (Chasmores Ltd
portable stadiometer).
Blood measurements
Duplicate finger-prick blood samples were obtained before
(mean±s.e.: 105.1±3.1 min after breakfast) and after the CF
tests (149.2±3.2 min after breakfast). Haemoglobin (Hb) was
measured using a ‘HemoCue’ 201+ analyzer (coefficient of
variation (CV) ¼1.3%) (HemoCue Ltd, Lake Forest, CA,
Glycaemic potency of breakfast and cognition
R Micha et al
949
European Journal of Clinical Nutrition
USA). Blood glucose was measured using a plasma-calibrated
Glucose Freestyle Mini meter (CVo5%) (Abbott Labora-
tories, Maidenhead, UK) (ADA, 1996; DH, 2005).
CF tests
CF tests were those used in previous studies to detect
differences in CF induced by glucose administration (Dono-
hoe and Benton, 1999; Kennedy and Scholey, 2000; Sunram-
Lea et al., 2001). The tests were administered in the same
order: word generation task (1), immediate word recall (2),
Stroop task (3), matrices (4), number search task (5), serial
sevens (6) and delayed word recall (7) (see Appendix A for a
detailed description of the CF tests).
Procedure
Each subject was seen once between 0945 and 1015 hours,
90–120 min after the start of usual breakfast, as stated on the
parents’ screening questionnaire. If a participant had eaten
a ‘snack’ as well, they were seen 90 min after the start of the
snack if it contained 410 g CHO.
The order of procedures was as follows: anthropometric
measurements; first finger-prick blood sample; mood scales
(before); CF tests; mood scales (after); second finger-prick
blood sample; task demand questions; interview.
Computation of GI and GL, and classification of pupils
Microdiet (Downlee Systems Ltd, High Peak, UK) was used
for nutrient analysis. The GI value of the individual foods
that comprised ‘breakfast’, ‘snacks’ and ‘dinner’ was derived
from published sources (Foster-Powell et al., 2002). The GI of
the composite meals was calculated as the sum of the
weighted GI values of the foods comprising the meal
(Wolever and Jenkins, 1986). The GL of a food was calculated
as follows (Salmeron et al., 1997):
GL ¼½GICHO=portion ðgÞ =100
The GL of a meal was calculated as the sum of the GL values
of the individual foods. The GI and GL values included
snacks contributing 10 g or more of CHO. Participants were
classified using a 2 2GIGL grid for breakfast: above or
below the median for GI ¼61 and for GL ¼27.
Statistical analysis
To establish whether the four GIGL groups as classified by
the median for GI and GL were well matched, we performed
one-way analysis of variance (ANOVA) to assess the mean
differences in the descriptive characteristics (age, height,
weight, body mass index), the BG and Hb levels, and the
GI, GL and macronutrient composition of the breakfast
between the four GI GL groups; Bonferroni post hoc tests
were further performed to assess the statistically signifi-
cant differences in the macronutrient composition of the
breakfasts corresponding to the GI GL classification. Two-way
ANOVA was used to assess differences in the CF test scores
between the four GIGL groups, using GI and GL as main
factors, and gender and socio-economic group as potential
confounders. Further potential confounders were explored by
performing correlation and multiple regression analyses,
including the order of administration of the tests; ‘having a
snack’; descriptive characteristics (age, height, weight, body
mass index); finger-prick blood measurements (Hb, BG); total
energy, protein, fat and CHO content of the breakfast eaten;
the GI and GL of the dinner the night before (52±1.6 and
46±3.9, respectively); hours of sleep (8.3±0.2); time between
breakfast and the first CF test (113±3 min); time between
waking up and the first CF test (149±4min); mood scores
before the CF tests; and exercise on the day and the day
before. The significant potential confounders (in addition to
gender and socio-economic group) identified, including age,
height, weight, body mass index, Hb levels, BG levels, ‘happy’
mood score before the CF tests, and time between breakfast
and the first CF test, were all initially included as covariates in
the two-way ANOVA. The ANOVA model for each CF test was
then further refined, removing the non-significant interac-
tions first (starting with the non-significant interaction with
the highest P-value), then removing the non-significant
factors and covariates, until all of the non-significant
interactions, factors and covariates had been removed from
the model. Correlation analyses were further performed
between the mood scales ‘after’ and the task demand scores
and CF; either measure could not be considered as a predictor
of performance on the CF tests as it was recorded on
completion of the testing. Nonetheless, mood scales ‘after’
and task demand scores were separately considered to
uncover potential associations between these measures and
cognitive performance. All P-values were two-tailed
(alpha ¼0.05). All analyses were performed using SPSS 15.0.
Results
Table 1 presents the descriptive characteristics of the sample.
The four GI GL groups were well matched, with no
significant differences in age, height, weight, Hb, body mass
index or BG. BG was higher in the high-GL group before the
tests compared with the low-GL group (105.6±2.1 versus
99.4±1.7 mg/dl, P¼0.025), and the decrease in BG
(2.6±2.6 mg/dl) in the high-GL group was statistically
significantly different from the rise in BG (4.7±2.3 mg/dl) in
the low-GL group (P¼0.04), suggesting that BG levels were
returning to baseline for the high-GL meals and recovering
from baseline for the low-GL meals. These two findings
support the original expectation that between 90 and
120 min after breakfast, high-GL meals would be associated
with higher BG levels.
Table 2 presents the macronutrient composition of break-
fast corresponding to the GI GL classification. The four
breakfast meals differed in their energy content, and %
Glycaemic potency of breakfast and cognition
R Micha et al
950
European Journal of Clinical Nutrition
energy from fat and CHO; % energy from protein was not
statistically different between the breakfast meals. Although
the two low-GL meals (high-GI and low-GI) had similar
energy content, the low-GI, high-GL meal had 32% higher
energy content than the high-GI, high-GL breakfast. Post-hoc
tests showed that this difference between the two high-GL
meals was not statistically significant, and that it was evident
between the low-GI, low-GL and the low-GI, high-GL
breakfast (P¼0.003); % energy from fat and CHO differed
between the low-GI, low-GL and high-GI, high-GL meal
(Po0.05 for both). Table 3 shows examples of the breakfasts
corresponding to the GI and GL classifications.
Table 1 Descriptive characteristics, GI–GL values and finger-prick blood measures in 60 children participating in the study, in all children and in the four
GI and GL groups
Breakfast meals
High-GL Low-GL P
a
All children Low-GI High-GI Low-GI High-GI
n(females/males) 60 (36/24) 11 (3/8) 19 (13/6) 19 (12/7) 11 (8/3)
Mean s.e. Mean s.e. Mean s.e. Mean s.e. Mean s.e.
Age (years) 13.0 0.1 13.2 0.1 12.8 0.1 12.9 0.1 12.8 0.2 0.294
Height (cm) 159.5 1.1 157.9 2.9 157.6 1.9 157.4 1.0 161.1 1.6 0.199
Weight (kg) 53.0 2.0 49.9 2.6 50.0 2.8 50.6 1.5 54.6 3.9 0.651
BMI (kg/m
2
) 20.7 0.6 20.0 0.8 19.9 0.8 20.3 0.5 21.0 1.4 0.895
GI
b
57 2 53 1 68 1 58 2 64 1 o0.001
GL
b
27 2 43 3 44 4 31 2 23 1 o0.001
Hb (g/l)
c
125.9 2.1 127.1 6.5 126.6 1.6 126.8 1.5 124.2 3.0 0.828
BG before test battery (mg/dl)
c
100.7 1.6 104.7 3.5 106.2 2.7 102.5 1.4 98.2 2.9 0.155
BG after test battery (mg/dl)
c
100.9 2.0 100.8 2.8 104.3 3.6 103.5 1.8 100.7 4.8 0.686
Difference in BG (mg/dl)
c
(after test-before test) 0.2 2.1 3.9 3.2 1.9 3.8 1.0 1.8 2.5 3.8 0.193
Abbreviations: BG, body glucose; BMI, body mass index; GI, glycaemic index; GL, glycaemic load; Hb, haemoglobin.
a
One-way ANOVA.
b
The GI and GL values corresponding to breakfast or breakfast plus snack.
c
Finger-prick blood measurements of Hb and BG.
Table 2 Macronutrient composition of breakfast in 60 children participating in the study, in all children and in the four GI and GL groups
Breakfast meals
High-GL Low-GL P
a
Females Males All children Low-GI High-GI Low-GI High-GI
(n ¼36)(n¼24)(n¼60)(n¼11)(n¼19)(n¼19)(n¼11)
Mean s.e. Mean s.e. Mean s.e. Mean s.e. Mean s.e. Mean s.e. Mean s.e.
Energy (kcal) 307.8 25.2 393.6 42.7 342.2 23.2 502.2 63.2 378.7 39.8 272.0 32.8 240.3 20.3 o0.001
Energy (kJ) 1288.0 105.2 1647.0 178.6 1431.6 97.2 2101.1 264.4 1584.3 166.6 1138.0 137.0 1005.4 84.8
Breakfast contribution
to EAR (%)
b
16.7 1.4 17.7 1.9 17.1 1.1 24.2 3.4 18.9 1.7 13.9 1.7 12.5 1.2 0.002
Protein (g) 10.8 1.1 14.6 2.1 12.3 1.1 17.9 2.9 12.6 2.2 10.2 1.6 9.7 1.0 0.053
Protein (% energy) 13.9 0.8 14.5 0.7 14.2 0.6 13.8 1.1 13.0 1.0 14.4 1.2 16.1 0.9 0.302
Fat (g) 10.0 1.3 11.2 2.2 10.5 1.2 14.0 3.5 9.3 1.6 11.3 2.5 7.7 1.8 0.376
Fat (% energy) 27.9 2.4 24.1 2.7 26.3 1.8 22.1 3.0 20.8 1.7 34.5 4.1 26.2 3.9 0.011
Carbohydrate (g) 46.6 3.6 62.5 6.2 52.9 3.4 81.2 6.0 65.2 5.5 34.5 3.7 35.4 1.9 o0.001
Carbohydrate (% energy) 58.2 2.4 61.5 3.0 59.5 1.9 64.1 3.1 66.2 1.9 51.1 4.3 57.7 3.7 0.006
Abbreviations: GI, glycaemic index; GL, glycaemic load.
The term ‘breakfast’ refers to either the ‘breakfast’ eaten by the children at their homes or to ‘breakfast’ and ‘snack’ combined, where appropriate (see section
Methods in the text).
a
One-way ANOVA.
b
EAR, Estimated Average Requirements (FAO/WHO, 1998).
Glycaemic potency of breakfast and cognition
R Micha et al
951
European Journal of Clinical Nutrition
The reported mood before the CF tests was considered
as a potential predictor of performance on the CF tests (see
Supplementary Table for average mood scores before the CF
tests among the four GI and GL groups). In general, before
the tests ‘positive’ feelings such as ‘friendly’, ‘happy’,
‘relaxed’ and ‘calm’ were significantly correlated with lower
scores on a number of CF tests, whereas other mood states,
including hunger and thirst, were not associated with
CF; the most consistent finding was that the more ‘happy’
a child the worse their performance (data not shown).
After the tests ‘negative’ feelings such as ‘drowsy’, ‘tired’ and
‘sluggish’ were generally associated with higher CF scores.
Correlation analyses between self-reported task demand and
CF showed that higher perceived difficulty was signi-
ficantly associated with lower scores on all CF tests
performed; perceived effort and tiredness were not signi-
ficantly associated with CF scores (data not shown). On the
basis of students’ scores, serial sevens (3.0±0.1) and speed
of information processing (2.7±0.1) were the two most
difficult tasks (Friedman test, Po0.001).
Table 4 shows mean performance scores across the
four GI GL groups. Among several potential confounders
explored, the order of administration of the tasks, ‘having
a snack’, the macronutrient composition of the breakfast,
the GI and GL of the dinner the night before, hours of
sleep, time between waking up and the first CF test,
mood states before the testing (with the exception of
feeling ‘happy’), and exercise on the morning of the
testing and the night before were unrelated to the CF
tasks. Only the significantly associated confounders were
included as covariates in the two-way ANOVA between
the four GI GL groups and CF test scores. The results of
the two-way ANOVA taking GI, GL and all relevant covari-
ates into account are summarized in Table 5. Performance
Table 3 Examples of breakfast meals corresponding to the GI and GL classification
Breakfast meals
High-GL Low-GL
Low-GI High-GI Low-GI High-GI
Special K/Muesli/Fruit and Fiber
Semi-skimmed milk
Orange juice
Sugar
Corn flakes/Coco pops/Rice
Krispies/Cheerios
Semi-skimmed milk
Sugar
Orange juice
Special K/Shredded wheat
Semi-skimmed milk
Sugar
Cheerios/Crunchy nut corn flakes
Semi-skimmed milk/Chocolate milk
Porridge
Orange juice
Whole-meal bread
Jam
Butter
Abbreviations: GI, Glycaemic index; GL, Glycaemic load.
Table 4 Average cognitive function test scores among the four GI and GL groups
Breakfast meals
High-GL Low-GL
Low-GI High-GI Low-GI High-GI
(n ¼11)(n¼19)(n¼19)(n¼11)
Mean s.e. Mean s.e. Mean s.e. Mean s.e.
Cognitive function test
Word generation task ‘correct’ (no. of words) 20.2 1.9 18.4 1.3 16.8 0.9 16.9 1.7
Memory recall immediate ‘correct’ (no. of words) 7.5 0.9 8.2 0.6 7.6 0.5 8.6 0.5
Stroop task (time of completion in sec) 71.3 2.9 80.6 3.5 80.4 2.9 76.8 4.9
Matrices (no. of matrices) 13.2 0.4 12.6 0.4 12.5 0.4 11.5 0.7
Speed of information processing ‘correct’ (no. of hits) 13.9 1.1 13.0 0.7 12.1 0.6 9.8 0.9
Serials sevens ‘correct’ (no. of subtractions) 29.3 5.3 15.1 2.1 15.6 2.2 18.0 2.7
Memory recall delayed ‘correct’ (no. of words) 6.5 1.2 6.2 0.7 5.8 0.6 6.6 0.5
Abbreviations: GI, Glycaemic index; GL, Glycaemic load.
Glycaemic potency of breakfast and cognition
R Micha et al
952
European Journal of Clinical Nutrition
on the word generation, Stroop and delayed word recall
task was not significantly associated with either GI or
GL. The remaining four tasks, immediate word recall,
matrices, speed of information processing and serial
sevens, were significantly associated with either GI, GL
or both. High-GI meals were associated with better
performance on the immediate word recall task; low-GI
meals with better performance on the speed of information
processing and serial sevens task; and high-GL meals with
better performance on the matrices, speed of information
processing and serial sevens task. Overall, findings were
similar for boys and girls (that is, gender was not
a statistically significant associated covariate), with the
exception of the serial sevens task, where boys performed
better than girls.
Discussion
The main finding of our study is that four of seven CF tests
administered were associated with either GI or GL, or both.
The method of classifying participants by both GI and GL
has proved useful for investigating the effects of the
glycaemic potency of breakfast on CF. Specifically, high-GI
was associated with better immediate recall (short-term
memory), high-GL with better matrices performance (in-
ductive reasoning), and low-GI and high-GL with better
speed of information processing (vigilance, sustained atten-
tion) and serial sevens performance (vigilance, working
memory). We hypothesized that a low-GI, high-GL breakfast
would be associated with the best CF. Our findings
are therefore consistent with the hypothesis for a low-GI,
Table 5 Cognitive function performance and statistically significantly associated factors
Cognitive function test Factor associated Positive (þ)or negative ()
association
PAdjusted-R
2
(%)
Word generation task Length of the interval between waking up and the first CF test () 0.041
Overall model: F(1,58) ¼4.203 0.041 5.1
Immediate word recall High-GI ( þ) 0.045
Feeling ‘happy’ before the testing () 0.003
Overall model: F(2,57) ¼5.752 0.005 13.9
Stroop task Age of child ( þ)o0.001
BMI () 0.049
Overall model: F(2,57) ¼10.013 o0.001 23.4
Matrices High-GL ( þ) 0.012
Being a boy (þ) 0.072
GL gender interaction (in the low-GL groups girls did better than boys,
whereas in the high-GL groups boys did better than girls)
(þ) 0.001
Overall model: F(3,56) ¼5.574 0.002 18.9
Speed of information
processing
Low-GI (þ) 0.031
High-GL (þ) 0.001
GI gender interaction (girls performed better in the low-GI than
in the high-GI group)
(þ) 0.046
Being taller ( þ) 0.003
Overall model: F(5,54) ¼5.026 0.001 25.4
Serial sevens Low-GI ( þ) 0.009
High-GL (þ) 0.055
Higher SEG ( þ) 0.029
Being a boy (þ)o0.001
Being taller ( þ) 0.016
GI GL (in the high-GL group performance was better for the
low-GI compared with the high-GI)
(þ) 0.010
GI SEG (performance better in lower SEG) ( þ) 0.033
GI GL SEG (performance better in lower SEG) ( þ) 0.035
Overall model F(8,51) ¼6.670 o0.001 46.5
Delayed word recall Height ( þ) 0.028
Feeling ‘happy’ before testing () 0.017
Overall model: F(2,57) ¼6.644 0.003 16.1
Abbreviations: BMI, body mass index; GI, glycaemic index; GL, glycaemic load; SEG, socio-economic group.
Two-way ANOVA was performed to assess differences in each of the CF test scores between the four GI GL groups, using GI and GL as the main factors, and
gender, SEG, age, height, weight, BMI, Hb levels, BG levels, ‘happy’ mood score before the CF tests and time between breakfast and the first CF test as covariates.
The order of administration of the tasks, ‘having a snack’, the macronutrient composition of the breakfast, the GI and GL of the dinner the night before, hours of
sleep, time between waking up and the first CF test, mood states before the testing (with the exception of feeling ‘happy’) and exercise on the morning of the testing
and the night before were unrelated to the CF tasks in correlation and regression analyses, and were thus not included as potential confounders. The ANOVA model
for each CF test was then further refined, removing the non-significant interactions first (starting with the non-significant interaction with the highest P-value),
then removing the non-significant factors and covariates, until all of the non-significant interactions, factors and covariates had been removed from the model.
The findings in table show results for the two-way ANOVA taking both main factors and relevant covariates into account.
Glycaemic potency of breakfast and cognition
R Micha et al
953
European Journal of Clinical Nutrition
high-GL effect for the speed of information and serial sevens
task, and for a high-GL but not a GI effect for the matrices. It
is not consistent with the hypothesis for a low-GI effect for
the immediate word recall.
The low-GI, high-GL breakfast was associated with better
scores on the two CF tasks (serial sevens and speed of
information processing) that were reported by the partici-
pants to be the most difficult (that is, mentally demanding).
This finding is consistent with previous studies (Owens et al.,
1997; Donohoe and Benton, 1999; Scholey et al., 2001;
Sunram-Lea et al., 2001) showing that in order for an effect of
glucose on cognition to be observed, the tasks need to be
sufficiently mentally demanding. Although it is difficult to
quantify cognitive demand, the duration of the task, its
complexity and time pressure probably all contributed to the
ratings of cognitive demand obtained in this study. Thus,
it could be argued that a low-GI, high-GL breakfast could
selectively facilitate mentally demanding CF tasks. None-
theless, GI or GL were also shown to be associated
with performance on the immediate word recall (high-GI)
and matrices (high-GL) tasks, which were not ranked by
the participants among the most difficult tasks. Therefore,
it appears that GI might be differentially associated with
different cognitive domains; low-GI was associated with
improved performance on two vigilance tasks (how quickly
information is being processed), but worse performance on
a memory task. The reason for this difference remains
unclear, although potential unmeasured interactions
between gluco-regulatory processes, arousal and subsequent
cortisol secretion could partially account for that. Lastly, the
order of administration of the tests was not significantly
associated with CF performance, suggesting that testing
fatigue (that is, the later tests only being affected) was not
driving the observed associations.
Only recently has GI or GL been used as a tool to assess the
effects of CHO-containing foods or meals on CF. In 71 young
female adults, a low-GI breakfast cereal predicted improved
performance only on the difficult abstract words of a word
recall task 150 and 210 min after breakfast, but not earlier
(Benton et al., 2003). Mahoney et al. (2005) administered
oatmeal (low-GI), ready-to-eat cereal (high-GI) or no break-
fast to school children aged 9–11 (15 males, 15 females) and
6–8 (15 males, 15 females). Spatial memory and short-term
working memory (in girls only) in both age groups, and
auditory attention in 6–8 year olds, were improved 1 h after
the low-GI breakfast only. Visual recall memory, visual
attention and verbal memory (recall) were not affected.
Benton et al. (2003) and Mahoney et al. (2005) considered
GI (but not GL) and maintained the same energy and
macronutrient content for the two breakfasts. Ingwersen
et al. (2007) found that in 64 children (26 boys, 38 girls)
aged 6–11, there was less decline in accuracy of attention
and recall memory (visual and verbal) 2 h after a low-GI
breakfast as compared with a high-GI breakfast. Working
memory, speed of memory and vigilance were not affected.
The latter study did not control for GL; furthermore, the
macronutrient composition of the meals was different
(high-GI cereal: 133 kcal, low-GI cereal: 98 kcal; 36%
difference in energy alone). To test if GI alone is having an
effect, the macronutrient composition should be similar and
only the CHO-source varied. In the study by Benton et al.
(2007), which included 9 boys and 10 girls aged 6–7, no
associations were observed between GL and any of the CF
domains assessed, that is memory (visual recall, spatial
memory) and the ability to sustain attention. This study
did not control for GI, and the macronutrient composition
of the meals was different. Among 189 adult females,
high-GL predicted improved reaction times and vigilance,
only in people with better gluco-regulation; verbal memory
was unaffected (Nabb and Benton, 2006). Nonetheless, this
study was not designed to investigate the effects of GI or GL,
and it was unbalanced.
Across all of these studies (including this study), there are
inconsistencies regarding all cognitive domains, including
declarative verbal memory and memory recall. One would
expect that these measures would be consistently affected by
GI and/or GL, as glucose administration has been shown to
improve both domains (Owens et al., 1997; Donohoe and
Benton, 1999; Scholey et al., 2001; Sunram-Lea et al., 2001).
Nonetheless, a CHO-rich breakfast may not produce the
same effects on CF and mood as glucose, as it will be
different in terms of absorption rates, gastric emptying,
metabolic effects and secondary hormonal responses.
Furthermore, the tasks selected between studies are not
identical; therefore, lack of sensitivity of the task, rather than
lack of an effect, could account for the inconsistencies. Most
importantly though, none of these studies investigated
the effects of both GI and GL, and with the exception of
two (Benton et al., 2003; Mahoney et al., 2005), none
of the remaining controlled for GL when investigating GI
(Ingwersen et al., 2007), and vice versa (Nabb and Benton,
2006; Benton et al., 2007).
Specifically, there should be a distinction between a low
glycaemic response as determined by both GI and GL
(the recommended approach) and a low glycaemic response
as determined solely by GI or GL. The literature to-date
generally predicts that a low glycaemic response is beneficial,
but it does not distinguish between a high, intermediate and
a truly low glycaemic response (that is, the lowest among the
meals compared when both GI and GL are taken into
account). This lack of distinction can be attributed to none
of the studies in this field assessing the impact of both GI
and GL. Therefore, when a low glycaemic response is
suggested as beneficial, this should be interpreted with
caution, as results for the same test can differ according to
whether GI (Ingwersen et al., 2007) or GL (Nabb and Benton,
2006) is measured. Future studies in this field should
investigate whether low-GI and high-GI meals selectively
facilitate different cognitive domains (for example, low-GI
facilitating vigilance, high-GI facilitating memory or vice
versa, as there are no firm findings supporting either one) by
controlling for GL, and including a selection of tests that
Glycaemic potency of breakfast and cognition
R Micha et al
954
European Journal of Clinical Nutrition
assess both mnemonic processes and vigilance. High-GL
seems to be more consistently associated with improved
performance (present findings; Nabb and Benton, 2006).
Also, other studies have not reported the possible impact of
mood as a confounder. The finding that pupils who were
feeling ‘happy’ did less well might be explained because they
were not motivated/aroused enough by the testing or the
mental load, and hence did not perform as well.
There are important limitations that need to be consid-
ered. First, the cross-sectional design means the observed
associations may not be causal, leaving open the possibility
that the relationships observed might be explained by
unmeasured confounders. Second, the GI and GL values
are based on what the participants reported having eaten
for breakfast; this age group is prone to under-reporting
(NDNS, 2000). Any resulting misclassification may some-
what underestimate the observed associations.
Third, the low-GI, high-GL breakfast had the highest
energy content in comparison with the other three meals. It
would have been desirable that within the same GL groups,
not only the energy content but also the macronutrient
composition to be similar for the GI meals. (It would, of
course, differ between the GL groups). Although this
difference in self-selected meals was not statistically signi-
ficant, it could have promoted higher BG levels and, thus
(assuming the hypothesis is correct), better performance
90–120 min later. While this was not evident from the
present findings, the effect of the energy, fat and
protein content per se cannot be strictly differentiated from
the effect of the glycaemic potency. A further extension
of this problem relates to some differences in the macro-
nutrient composition of the meals. Nonetheless, the
differences in composition do not explain differences in
CF as consistently as the observed combined effects of
GL and GI in the two CF measures affected, the high-GL
and the low-GI groups performing better (in accordance with
the hypothesis).
Another limitation is that any snacks that contained
o10 g of available CHO were not included in the total
breakfast meal. These snacks were excluded on the basis that
they would not change BG levels, and to provide consistent
classification of breakfast GI and GL in relation to the timing
of the CF tests. Nonetheless, it should be noted that
introduction of other macronutrients into the stomach
might, to a small degree, alter the absorption and digestion
profile of the breakfast meals, thus potentially altering
glycaemic, insulinaemic and other hormonal responses
that in turn might have an effect on CF test scores. However,
‘having a snack’ did not appear as a significant factor
predicting CF test outcomes.
This study, like other studies in the field (Benton et al.,
2003; Mahoney et al., 2005; Nabb and Benton, 2006; Gibson,
2007; Ingwersen et al., 2007), did not have a baseline
measure of performance, as the interest was in short-term
differences of high-GI or low-GI and high-GL or low-GL
meals on CF, and not on whether there is an improvement or
decline in overall CF as a result of a meal. In all these studies,
the GI calculations were based on published values
(Foster-Powell et al., 2002), which may have introduced
error in the estimation of the exposure. Any resulting
misclassification would be likely to attenuate the significant
associations observed here rather than reveal associations,
which do not exist.
Despite these limitations this study suggests that the GI
and GL of breakfast may affect performance in specific
cognitive domains and under real-life conditions, and
that performance in the classroom might be better after a
low-GI, high-GL breakfast. This study is the first of its kind
to consider both GI and GL when assessing the effects
of breakfast on CF in teenage school children, and to show
that both are associated with specific cognitive outcomes.
The fact that performance in only four of the seven tests
administered was associated with differences in GI and GL
does not undermine the overall findings. There was never
the expectation that all of the CF tests would be affected.
Indeed, theory predicts that differences in gluco-regulatory
processes and cortisol secretion under stress (that is, arousal)
may differentially affect performance after administration of
meals differing in their GI and/or GL (Gibson, 2007); to what
extent remains to be elucidated.
Conflict of interest
The authors declare no conflict of interest.
Acknowledgements
All authors revised the manuscript for important intellectual
content, and approved final manuscript for submission. We
are grateful to Ms Julia Forbes and Ms Kathryn Lowes, who
helped in carrying out fieldwork and data coding, and to all
the enthusiastic volunteers who participated in this trial.
Financial support was obtained from the Harokopeio Uni-
versity PhD scholarship.
References
ADA (1996). Consensus statement on clinical practice recommen-
dation: self-monitoring of blood glucose. Diabetes Care 19, S62–S66.
ADA (2004). Dietary carbohydrate (amount and type) in the
prevention and management of diabetes. A statement by the
American Diabetes Association. Diabetes Care 27, 2266–2271.
Barclay A, Brand-Miller J, Wolever TM (2005). Glycemic index,
glycemic load, and glycemic response are not the same. Diabetes
Care 28, 1839.
Benton D, Maconie A, Williams C (2007). The influence of the
glycaemic load of breakfast on the behaviour of children in school.
Physiol Behav 92, 717–724.
Benton D, Owens D, Perales CG (1994). Blood glucose influences
memory and attention in young adults. Neuropsychobiology 32,
595–607.
Benton D, Parker PY (1998). Breakfast, blood glucose, and cognition.
Am J Clin Nutr 67, S772–S778.
Glycaemic potency of breakfast and cognition
R Micha et al
955
European Journal of Clinical Nutrition
Benton D, Ruffin MP, Lassel T, Nabb S, Messaoudi M, Vinoy S et al.
(2003). The delivery rate of dietary carbohydrates affects cognitive
performance in both rats and humans. Psychopharmacology (Berl)
166, 86–90.
Bradley RH, Corwyn RF (2002). Socioeconomic status and child
development. Annu Rev Psychol 53, 371–399.
Brand-Miller JC, Thomas M, Swan V, Ahmad ZI, Petocz P, Colagiuri S
(2003). Physiological validation of the concept of glycemic load in
lean young adults. J Nutr 133, 2728–2732.
Chandler AM, Walker SP, Connolly K, Grantham-McGregor SM
(1995). School breakfast improves verbal fluency in under-
nourished Jamaican children. J Nutr 125, 894–900.
Conners CK, Blouin AG (1982). Nutritional effects on behavior of
children. J Psychiatr Res 17, 193–201.
Craft S, Murphy C, Wemstrom J (1994). Glucose effects on complex
memory and nonmemory tasks: the influence of age, sex, and
glucoregulatory response. Phychobiology 22, 95–105.
DH (2005). Department of Health; Devise Evaluation Services; Blood
Glucose Monitoring: Know Your Systems. http://www.mhra.gov.uk
Dolan RJ (2002). Emotion, cognition, and behavior. Science 298,
1191–1194.
Donohoe RT, Benton D (1999). Cognitive functioning is susceptible to
the level of blood glucose. Psychopharmacology (Berl) 145, 378–385.
Dye L, Lluch A, Blundell JE (2000). Macronutrients and mental
performance. Nutrition 16, 1021–1034.
Elliot C, Smith P, McCulloch K (1996). British Ability Scales (BAS II),
2nd ed. Windsor, Berkshire.
FAO/WHO (1998). FAO Food and Nutrition Paper No 66. Carbo-
hydrates in Human Nutrition. Report of a Joint FAO/WHO Expert
Consultation. FAO Rome.
Foster-Powell K, Holt S, Brand-Miller J (2002). International table of
glycemic index and glycemic load values. Am J Clin Nutr 76,5–56.
Gibson EL, Green MW (2002). Nutritional influences on cognitive
function: mechanisms of susceptibility. Nutr Res Rev 15, 169–206.
Gibson EL (2007). Carbohydrates and mental function: feeding or
impeding the brain? Nutr Bull 32, 71–83.
Gordon N (2003). Iron deficiency and the intellect. Brain Dev 25, 3–8.
Government Statistical Service (1990). Standard Occupational
Classification; Volumes 1, 2 & 3, Office for National Statistics.
The Stationery Office: London.
Halterman JS, Kaczorowski JM, Aligne CA, Auinger P, Szilagyi PG (2001).
Iron deficiency and cognitive achievement among school-aged children
and adolescents in the United States. Pediatrics 107, 1381–1386.
Hayman M (1942). Two minute clinical tests for measurement of
intellectual impairment in psychiatric disorders. Arch Neurol
Psychiatry 47, 454–464.
Heatherley SV, Hayward RC, Seers HE, Rogers PJ (2005). Cognitive
and psychomotor performance, mood, and pressor effects of
caffeine after 4, 6 and 8 h caffeine abstinence. Psychopharmacology
(Berl) 178, 461–470.
Hoyland A, Lawton CL, Dye L (2008). Acute effects of macronutrient
manipulations on cognitive test performance in healthy young
adults: a systematic research review. Neurosci Biobehav Rev 32, 72–85.
Ingwersen J, Defeyter MA, Kennedy DO, Wesnes KA, Scholey AB
(2007). A low glycaemic index breakfast cereal preferentially
prevents children’s cognitive performance from declining
throughout the morning. Appetite 49, 240–244.
Jenkins DJ, Wolever TM, Taylor R, Barker H, Fielden H, Baldwin J
et al. (1981). Glycemic Index of foods: a physiological basis for
carbohydrate exchange. Am J Clin Nutr 34, 362–366.
Kennedy DO, Scholey AB (2000). Glucose administration, heart rate
and cognitive performance: effects of increasing mental effort.
Psychopharmacology (Berl) 149, 63–71.
Lloyd HM, Rogers PJ, Hedderley DI, Walker AF (1996). Acute effects
on mood and cognitive performance of breakfasts differing in fat
and carbohydrate content. Appetite 27, 151–164.
Lorr M, McNair D (1984). Profile of Mood States Bi-polar Form.
Educational & Industrial Testing Service: San Diego, CA.
Ludwig DS (2002). The glycemic index: physiological mechanisms
relating to obesity, diabetes, and cardiovascular disease. JAMA 287,
2414–2423.
Mahoney CR, Taylor HA, Kanarek RB, Samuel P (2005). Effect of
breakfast composition on cognitive processes in elementary
school children. Physiol Behav 85, 635–645.
Manly T, Anderson V, Nimmo-Smith I, Turner A, Watson P, Robertson
IH (2001). The differential assessment of children’s attention: the
Test of Everyday Attention for Children (TEA-Ch), normative
sample and ADHD performance. J Child Psychol Psychiatry 42,
1065–1081.
Messier C (2004). Glucose improvement of memory: a review. Eur J
Pharmacol 490, 33–57.
Nabb S, Benton D (2006). The influence on cognition of the
interaction between the macro-nutrient content of breakfast and
glucose tolerance. Physiol Behav 87, 16–23.
NDNS (2000). Food Standards Agency. National Diet and Nutrition Survey:
Young People Aged 4 to 18 Years. The Stationery Office: London.
Nelson M, Atkins M, Meyer J (1997). Food Portion Sizes, a User’s Guide
to Photographic Atlas. MAFF Publications: London.
Owens DS, Parker PY, Benton D (1997). Blood glucose and subjective
energy following cognitive demand. Physiol Behav 62, 471–478.
Pollitt E, Leibel RL, Greenfield D (1981). Brief fasting, stress, and
cognition in children. Am J Clin Nutr 34, 1526–1533.
Pollitt E, Lewis NL, Garza C, Shulman RJ (1982). Fasting and
cognitive function. J Psychiatr Res 17, 169–174.
Riby LM (2004). The impact of age and task domain on cognitive
performance: a meta-analytic review of the glucose facilitation
effect. Brain Impair 5, 145–165.
Rogers PJ, Richardson NJ, Elliman NA (1995). Overnight caffeine
abstinence and negative reinforcement of preference for caffeine-
containing drinks. Psychopharmacology (Berl) 120, 457–462.
Salmeron J, Manson J, Stampfer M, Colditz G, Wing A, Willet W
(1997). Dietary fiber, glycemic load, and risk of non-insulin-
dependent diabetes mellitus in women. JAMA 277, 472–477.
Scholey AB, Harper S, Kennedy DO (2001). Cognitive demand and
blood glucose. Physiol Behav 73, 585–592.
Scholey AB, Kennedy DO (2004). Cognitive and physiological
effects of an ‘energy drink’: an evaluation of the whole drink
and of glucose, caffeine and herbal flavouring fractions 112.
Psychopharmacology 176, 320–330.
Simeon DT, Grantham-McGregor S (1989). Effects of missing break-
fast on the cognitive functions of school children of differing
nutritional status. Am J Clin Nutr 49, 646–653.
Smith AP, Clark R, Gallagher J (1999). Breakfast cereal and caffeinated
coffee: effects on working memory, attention, mood, and
cardiovascular function. Physiol Behav 67,9–17.
Stroop JR (1935). The basis of Ligon’s theory. Am J Psychol 47,499–504.
Sunram-Lea SI, Foster JK, Durlach P, Perez C (2001). Glucose facilitation
of cognitive performance in healthy young adults: examination of
the influence of fast-duration, time of day and pre-consumption
plasma glucose levels. Psychopharmacology (Berl) 157, 46–54.
Taylor LA, Rachman SJ (1988). The effects of blood sugar level
changes on cognitive function, affective state, and somatic
symptoms. J Behav Med 11, 279–291.
Vaisman N, Voet H, Akivis A, Vakil E (1996). Effect of breakfast
timing on the cognitive functions of elementary school students.
Arch Pediatr Adolesc Med 150, 1089–1092.
Wesnes KA, Pincock C, Richardson D, Helm G, Hails S (2003).
Breakfast reduces declines in attention and memory over the
morning in schoolchildren. Appetite 41, 329–331.
Wolever TM, Jenkins DJ (1986). The use of the glycemic index in
predicting the blood glucose response to mixed meals. Am J Clin
Nutr 43, 167–172.
Supplementary Information accompanies the paper on European Journal of Clinical Nutrition website (http://www.nature.com/ejcn)
Glycaemic potency of breakfast and cognition
R Micha et al
956
European Journal of Clinical Nutrition
Appendix A
List of 22 mood states assessed
1. Friendly
2. Nervous
3. Drowsy
4. Happy
5. Calm
6. Uncertain
7. Sad
8. Energetic
9. Muddled
10. Relaxed
11. Dissatisfied
Glycaemic potency of breakfast and cognition
R Micha et al
957
European Journal of Clinical Nutrition
