Content uploaded by Sophie Vinoy
Author content
All content in this area was uploaded by Sophie Vinoy on Sep 07, 2015
Content may be subject to copyright.
Glycaemic index of cereal products explained
by their content of rapidly and slowly available glucose
Klaus N. Englyst
1
*†, Sophie Vinoy
2
†, Hans N. Englyst
1
and Vincent Lang
2
1
Englyst Carbohydrates – Research & Services Ltd, 2 Venture Road,
Chilworth Science Park, Southampton, Hampshire SO16 7NP, UK
2
Danone Vitapole, Nutrivaleur, route de
´
partementale 128, 91767 Palaiseau cedex, France
(Received 27 March 2002 – Revised 9 October 2002 – Accepted 13 October 2002)
Elucidating the role of carbohydrate quality in human nutrition requires a greater understanding
of how the physico-chemical characteristics of foods relate to their physiological properties.
It was hypothesised that rapidly available glucose (RAG) and slowly available glucose
(SAG), in vitro measures describing the rate of glucose release from foods, are the main deter-
minants of glycaemic index (GI) and insulinaemic index (II) for cereal products. Twenty-three
products (five breakfast cereals, six bakery products and crackers, and twelve biscuits) had their
GI and II values determined, and were characterised by their fat, protein, starch and sugar con-
tents, with the carbohydrate fraction further divided into total fructose, RAG, SAG and resistant
starch. Relationships between these characteristics and GI and II values were investigated by
regression analysis. The cereal products had a range of GI (28–93) and II (61–115) values,
which were positively correlated (r
2
0·22, P, 0·001). The biscuit group, which had the highest
SAG content (8·6 (
SD 3·7) g per portion) due to the presence of ungelatinised starch, was found
to have the lowest GI value (51 (
SD 14)). There was no significant association between GI and
either starch or sugar, while RAG was positively (r
2
0·54, P, 0·001) and SAG was negatively
(r
2
0·63, P, 0·001) correlated with GI. Fat was correlated with GI (r
2
0·52, P, 0·001), and
combined SAG and fat accounted for 73·1 % of the variance in GI, with SAG as the dominant
variable. RAG and protein together contributed equally in accounting for 45·0 % of the variance
in II. In conclusion, the GI and II values of the cereal products investigated can be explained by
the RAG and SAG contents. A high SAG content identifies low-GI foods that are rich in slowly
released carbohydrates for which health benefits have been proposed.
Glycaemic index: Insulinaemic index: Dietary carbohydrate: Carbohydrate quality:
Cereal products
Since its development in 1981, the glycaemic index (GI) has
had a pivotal role in highlighting the variation in physiologi-
cal responses associated with different carbohydrate-
containing foods (Jenkins et al. 1981). This ranking of
foods by the glycaemic responses elicited when equi-carbo-
hydrate portions are consumed has provided a unique, and at
times controversial, perspective on the issue of carbohydrate
quality (Wolever, 1997; Bellisle, 2001). Low-GI diets have
been successfully applied as a dietary therapy in diabetes
mellitus and other conditions exhibiting derangements in
carbohydrate and lipid metabolism (Brand-Miller, 1994).
In these studies, the major dietary alterations were to the
starch-containing foods, with the substitution of slowly
digested low-GI products, such as pasta, wholegrain cereal
and legumes, for rapidly digested high-GI products, such
as bread, breakfast cereals and potatoes.
The relationship between the rate of starch digestion and
GI has been established by investigations of in vitro amylo-
lytic hydrolysis (O’Dea et al. 1981; Jenkins et al. 1982;
Heaton et al. 1988; Bornet et al. 1989; Englyst et al.
1992; Granfeldt et al. 1992). The rate and extent of starch
digestion is influenced by botanical origin as this determines
the amylose:amylopectin ratio and the structural type of the
starch granule (Gallant et al. 1992). The other important
factor is food processing, which determines the extent of
starch gelatinisation, particle size and the integrity of the
* Corresponding author: Dr Klaus N. Englyst, fax +44 23 8076 9654, email Klaus@Englyst.co.uk
† K. N. E. and S. V. are joint first authors.
Abbreviations: BC, breakfast-cereal group; Bi, biscuit group; BP&C, bakery-products and crackers group; GI, glycaemic index; II, insulinaemic index;
RAG, rapidly available glucose; RS, resistant starch; SAG, slowly available glucose.
British Journal of Nutrition (2003), 89, 329–339 DOI: 10.1079/BJN2002786
q Danone Vitapole 2003
plant cell wall (Heaton et al. 1988; Holm et al. 1988;
Colonna et al. 1992; Holt & Brand-Miller, 1994; Heijene
et al. 1995). These physico-chemical variables of starch-
containing foods are difficult to characterise in a quantitat-
ive manner that relates to their likely physiological fate.
Instead, the influence of such physico-chemical character-
istics on the rate and extent of carbohydrate digestion can
be measured, and this can then be used to provide a descrip-
tion of this nutritionally important aspect of the food.
In conjunction with studies on human subjects, we have
developed analytical procedures that characterise dietary
carbohydrates with regard to chemical composition and
likely gastrointestinal fate (Englyst et al. 1992, 1999;
Englyst & Hudson, 1996). The glycaemic carbohydrate
fraction that is available for absorption in the small intestine
is measured as the sum of sugars and starch, excluding
resistant starch (RS). We have divided the glycaemic
glucose fraction (sum of glucose in the glycaemic carbo-
hydrate fraction, but excluding lactose) into rapidly avail-
able glucose (RAG) and slowly available glucose (SAG),
to reflect the likely rate of release and absorption of glucose.
In addition to the rate of carbohydrate digestion, food-
mediated effects on both gastrointestinal events and
post-absorptive metabolism can influence the GI. Gastric
emptying is affected by food particle size (Thomsen et al.
1994) and fat content (Gannon et al. 1993), as well as by
viscous fibre, which also limits enzymatic hydrolysis in
the small intestine by restricting access to the food bolus
(Jenkins et al. 1978). Post-absorptive factors that can influ-
ence GI include the identity of the sugar moieties, which
are metabolised differently (Lee & Wolever, 1998), and
the insulinotropic effect of protein, which can increase
the clearance rate of circulating glucose (van Loon et al.
2000). This emphasises the fact that GI values do not rep-
resent a direct measure of carbohydrate absorption from
the small intestine. Rather, the GI values are determined
by the combined effect of all the properties of a food
that influence the rate of influx and removal of glucose
from the circulation. A better understanding of the mechan-
isms involved should provide insight into the concept of
GI, and help to establish whether different types of low-GI
diets are equally beneficial to health.
As cereal grains are the largest contributor to carbo-
hydrate intake, it follows that altering the carbohydrate
quality of cereal products is likely to have the most tangi-
ble effect on this aspect of nutrition. Considerable choice
and flexibility exists for the consumer in their selection
of cereal products based on the grain type, degree of refine-
ment and type of processing (Prochaska et al. 2000).
Unfortunately, with a few notable exceptions, the food-pro-
cessing techniques employed in the manufacture of cereal
products tend to result in the disruption of the food
matrix and the gelatinisation of starch granules, thereby
making them readily digestible and consequently they gen-
erally have high GI values. The challenge is to identify
techniques for cereal processing that result in starch that
is slowly digested, thereby achieving low-GI products.
Previously, we have shown that for a limited number of
predominantly starchy foods, the division between RAG
and SAG has physiological significance with regard to
glycaemic response (Englyst et al. 1999). In the present
paper, we extend our investigations on the correlations
between the chemical and carbohydrate digestibility
characteristics of cereal products and their GI and insuli-
naemic index (II) values. These relationships need to be
established in order to achieve a greater understanding of
the role of carbohydrate quality in nutrition.
Subjects and methods
Test meals
Twenty-three cereal products, selected for investigation on
the basis that they may be interchangeable within a break-
fast meal or snack, were collected from different countries
(Table 1). The products encompass a range of ingredients
and processing techniques that contribute to defining their
physico-chemical properties. The predominant cereal pre-
sent in these products is wheat, with a few exceptions
where maize (cornflakes), rice (Special K) or oats (Alpen)
represent the sole or major cereal component. Apart from
the Alpen muesli, which contained steamed rolled oat ker-
nels (and some dried fruit and nuts), the products were pro-
duced from flour and did not contain dense matrices.
The other products in the breakfast-cereal group (BC)
comprised flakes of extruded cereals, which are low in fat,
with a moderate content of sugar. The bakery-product and
crackers group (BP&C) included two baguette meals and
two brioche-type products where fat and milk are present
during the baking process. The crackers were included
within group BP&C, as these are also produced by baking
in the presence of fat and moisture. The biscuit group (Bi)
was on average moderately high in fat and sugar, and in con-
trast with the other groups, several of the biscuits were
baked under low-moisture conditions, which restricts the
extent to which starch granules are gelatinised.
Glycaemic index determination
The GI values of the selected cereal products were deter-
mined using the previously described standard protocol
(Wolever, 1991; Food and Agriculture Organization/
World Health Organization, 1998). The results were
obtained from a series of seven sets of GI determinations,
each of which included between eleven and fourteen
healthy subjects.
Briefly, subjects who had maintained an overnight fast
were fed the test product in a portion size that was calculated
to contain 50 g carbohydrate (as determined by an initial
analysis of total starch and sugar content). Within the indi-
vidual sets, each subject was given the products being tested
once in a randomised order. In addition, on three separate
occasions each subject also consumed a 50 g glucose sol-
ution as the reference meal (anhydrous glucose (dextrose);
Sigma Chemical Company, St Louis, MO, USA). Blood
samples were taken before and 15, 30, 45, 60, 90 and
120 min after each meal commenced. Plasma glucose con-
centrations were measured in duplicate using an enzymatic
method (Roche Diagnostica, Basle, Switzerland) and insulin
was measured by radioimmunoassay. Incremental areas
under the blood glucose response curves were calculated
using the trapezoid rule, with only the area above the
K. N. Englyst et al.330
Table 1. Nutritional composition and physiological characteristics of twenty-three cereal products‡
Nutritional composition (g per portion size)‡
Total Total Glycaemic
GI§ II§
Product Manufacturer Origin Portion size (g) Fat Protein NSP Starch Sugar fructose FSG glucose glucose RS RAG SAG Mean
SEM Mean SEM
Breakfast-cereals group
Chocapic Nestle
´
France 60 2·6 4·2 1·3 25·4 26·2 12·2 14·0 39·4 38·5 0·9 37·9 0·6 84 9 86 7
Cornflakes Kellogg’s France 58 0·4 3·0 0·6 47·2 5·3 2·7 2·6 49·8 47·5 2·3 45·7 1·8 93 16 69 7
Energy Mix Quaker France 58 1·3 5·2 2·3 38·5 10·8 5·0 5·8 44·3 42·4 1·8 40·6 1·9 80 7 83 6
Special K Kellogg’s France 64 0·9 9·0 1·3 38·7 11·0 5·5 5·5 44·2 43·2 1·0 41·3 1·9 84 12 97 20
Alpen original Weetabix UK 80 5·6 9·5 4·2 32·9 18·2 9·4 8·8 41·7 42·0 0·0 38·8 3·2 55 10 76 8
Bakery-products and
crackers group
French baguette with
butter + jam
Bakery France 90 7·9 7·9 1·5 38·3 13·5 6·7 6·8 45·1 43·8 1·3 42·6 1·2 62 7 77 16
French baguette
with chocolate spread
Bakery France 80 8·8 5·0 1·6 39·5 9·5 4·6 4·8 44·3 43·3 1·0 41·7 1·6 72 8 81 15
Pain au lait Pasquier France 94 11·6 7·5 1·3 39·2 11·5 6·0 5·5 44·6 42·9 1·7 39·5 3·4 63 10 74 11
Brioche Bakery France 101 12·4 10·1 1·5 45·9 4·5 2·6 1·9 46·5 44·7 1·7 41·1 3·6 70 18 115 19
Cracker Triunpho Brazil 75 9·9 7·0 1·9 49·9 0·7 0·2 0·6 50·5 49·2 1·3 44·1 5·1 64 11 106 16
High-Ca cracker Jacob’s Malaysia 72 13·0 5·5 1·4 41·3 6·7 3·1 3·6 44·9 43·7 1·2 36·8 6·9 52 8 86 21
Biscuit group
Barquette abricot LU France 65 1·8 3·1 0·8 18·9 33·8 15·7 18·1 37·1 35·4 1·7 32·8 2·6 71 6 77 9
Oro Saiwa Italy 62 6·0 5·3 1·5 39·6 12·5 6·1 6·4 46·1 44·7 1·4 40·6 4·1 61 9 87 21
Prince – gou
ˆ
t chocolat LU France 69 13·2 3·7 1·6 26·8 22·0 10·5 11·5 38·3 37·3 1·0 33·0 4·3 53 5 77 6
Ve
´
ritable petit beurre LU France 67 7·7 5·3 1·2 36·4 15·1 7·6 7·6 43·9 42·2 1·7 38·0 4·2 51 8 72 12
Vitasnella frollini Saiwa Italy 61 5·9 4·6 0·9 35·9 12·0 5·7 6·4 42·3 41·4 0·9 32·3 9·0 59 10 90 17
Petit brun extra LU France 62 7·7 4·3 1·3 34·1 15·4 7·4 7·9 42·0 40·4 1·6 31·7 8·7 77 12 75 10
P’tit de
´
jeuner chocolat LU France 73 13·0 5·5 2·2 30·6 19·5 9·6 9·9 40·5 40·1 0·4 28·7 11·4 42 5 75 5
The
´
LU France 64 9·3 4·5 1·1 32·2 16·3 8·1 8·2 40·4 39·4 1·0 29·3 10·1 41 7 72 13
P’tit dejeuner miel et pe
´
pites LU France 71 12·8 4·0 2·3 27·0 19·4 9·1 10·2 37·2 36·9 0·3 27·3 9·6 45 5 66 6
Gran’Dia – banana com mel Danone Brazil 65 10·1 4·2 1·4 33·4 14·9 7·3 7·6 41·0 39·9 1·1 26·1 13·8 28 5 61 9
Gran’Dia chocolate com
cinque cereals
Danone Brazil 69 11·6 4·6 1·9 29·7 18·5 9·1 9·4 39·1 38·2 0·8 26·2 11·9 39 8 61 10
Principe megamanana vanilla LU Spain 69 12·1 4·6 1·8 28·9 19·2 9·4 9·8 38·7 38·2 0·5 27·3 10·9 45 6 80 16
Group means{ Mean
SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD
Breakfast cereals (n 5) 2·2* 2·1 6·2 2·9 1·9 1·4 36·5 8·1 14·3 8·1 7·0 3·8 7·3 4·3 43·9 3·9 42·7 3·2 1·2 0·9 40·9 3·0 1·9 0·9 78·9 14·1 82·2 10·5
Bakery products
and crackers (n 6)
10·6 2·0 7·2 1·8 1·5 0·2 42·3 4·6 7·7 4·7 3·9 2·4 3·9 2·3 46·0 2·3 44·6 2·3 1·4 0·3 41·0 2·5 3·6 2·2 63·7 7·0 89·8 17·0
Biscuits (n 12) 9·3† 3·6 4·5* 0·7 1·5 0·5 31·1* 5·5 18·2* 5·8 8·8* 2·6 9·4* 3·2 40·6* 2·7 39·5* 2·5 1·0 0·5 31·1*† 4·6 8·4*† 3·7 51·0*† 14·0 73·7 8·8
All (n 23) 8·1 4·3 5·5 2·0 1·6 0·7 35·2 7·4 14·6 7·3 7·1 3·4 7·5 3·9 42·7 3·7 41·5 3·4 1·2 0·5 35·8 6·2 5·7 4·0 60·4 16·6 79·8 13·2
FSG, free sugar glucose; RS, resistant starch; RAG, rapidly available glucose; SAG, slowly available glucose; GI, glycaemic index; II, insulinaemic index.
Mean values were significantly different from those of the bakery-products and crackers group: *P, 0·05.
Mean values were significantly different from those of the breakfast-cereals group: †P, 0·05.
‡ The nutritional composition of the cereal products is expressed per portion size used in the physiological studies. The carbohydrate component was characterised according to a number of different chemical and digestibility
characteristics used to investigate which best described the variance in the physiological responses.
§ GI and II values were determined according to standardised methodology based on test meals of portion sizes calculated to provide 50 g carbohydrate. A solution of glucose was used as the reference meal, to which the
responses to the other test meals were related (hence without units).
{ Mean values and standard deviations for comparison between product groups.
Glycaemic indices of cereal products 331
baseline being included. For each subject, a GI value for the
test food was calculated as the incremental area under the
blood glucose response curve for the test food, expressed
as a percentage of the average incremental area under the
blood glucose response curve for the three glucose
reference test meals. The GI value of a product was calcu-
lated as the average of the GI values for that product
obtained from the individual subjects. Calculation of II
values for the test foods followed the principles utilised in
the GI calculations.
Dietary carbohydrate analysis
Carbohydrate analysis of the products was performed using
previously described procedures (Englyst et al. 1994,
2000). For the measurement of RAG, SAG and starch frac-
tions, samples were minced using the specified procedure
to simulate buccal mastication. Portions of the samples
containing approximately 500 mg carbohydrate were
weighed into 50 ml centrifuge tubes together with an
internal standard. After an initial treatment with pepsin,
the samples were incubated with a mixture of hydrolytic
enzymes under controlled conditions of pH, temperature,
viscosity and mechanical mixing. Subsamples were taken
from the incubation mixture exactly 20 and 120 min after
the commencement of the hydrolysis. RAG and SAG
values were calculated as the glucose released at 20 min
and between 20 and 120 min respectively. Any starch
remaining in the main incubation tube was dispersed and
hydrolysed, with the increase in released glucose calcu-
lated as the RS fraction. Free sugar glucose and fructose
(including that derived from sucrose) were determined
after extraction procedures and incubation with invertase.
Five models describing different chemical and digest-
ibility properties of the carbohydrate components were
constructed from the analytical fractions. These were
developed in a progressive fashion, with each successive
carbohydrate model based on the previous one: model 1,
starch and sugar; model 2, starch, with the sugar fraction
divided into total fructose and free sugar glucose com-
ponents; model 3, fructose and the total glucose fraction
calculated as the sum of free sugar glucose and glucose
from starch; model 4, fructose, with the total glucose frac-
tion divided into glycaemic glucose and RS; model 5, fruc-
tose and RS, with the glycaemic glucose divided into RAG
and SAG.
To investigate the relationships between the physico-
chemical characteristics of foods and their in vitro digest-
ibility profiles, three products from each group had their
extent of starch gelatinisation determined by scanning
differential calorimetry (Biliaderis et al. 1980). The results
were expressed as the starch gelatinisation index (extent to
which starch is gelatinised, with 100 representing total
gelatinisation).
Statistical analysis
Student’s t test was used for comparisons between cereal
product groups; a P value , 0·05 was considered significant.
The relationships between nutrients and with the GI and II
were investigated by univariate correlations. The total
variance in GI and II that each of the carbohydrate models
1– 5 could explain was investigated by ANOVA. The var-
iance explained by fat, protein and NSP was also investi-
gated. From these findings, simplified models were
developed that best described the variance in GI and II
values of the cereal products by their chemical and digest-
ibility characteristics. Statistical analysis was performed
with SPSS (version 9; SPSS Inc., Chicago, IL, USA).
Results
Characteristics of the cereal products
Table 1 shows the nutrient compositions of the twenty-
three foods that were investigated in the physiological
studies, expressed per portion size, and with the corre-
sponding GI and II values. There were significant differ-
ences between the three groups of products in their
chemical and digestibility characteristics. There was less
fat in BC than the other groups, and less protein and
starch and more sugar in Bi than in BP&C. The differences
between groups extended to the detailed carbohydrate frac-
tions, where Bi contained less RAG and more SAG than
BC, and Bi contained more total fructose than BP&C. Bi
had significantly lower GI values than BP&C and BC.
There was no significant difference in II values between
groups. There was a positive correlation between GI and
II values (Fig. 1), although II values were significantly
higher (P, 0·001) than GI values.
The starch gelatinisation indices for the subset of pro-
ducts in the BC group were: Energy Mix 99, Chocapic
100, Special K 100. For the BP&C group, the values
were: French baguette with butter + jam 98, French bagu-
ette with chocolate 98, pain au lait 100. For the Bi group
Fig. 1. Relationships between the glycaemic index (GI) and insuli-
naemic index (II) values of twenty-three cereal products, determined
from test meals calculated to contain 50 g carbohydrate with a glu-
cose solution as the reference meal. (W), Breakfast cereals; (A),
bakery products and crackers; (X), biscuits. For details of
subjects, cereal products and procedures, see Table 1 and p. 330.
II ¼ 57·5 £ 0·368 £ GI; r
2
¼ 0·215; P, 0·05.
K. N. Englyst et al.332
the values were: P’tit de
´
jeuner miel et pe
´
pites 56, Principe
megamanana vanilla 50, Ve
´
ritable petit beurre 40. Fig. 2
shows that the Bi group had the lowest mean GI value,
with the highest mean SAG content and the lowest mean
starch gelatinisation index compared with the BC and
BP&C groups.
Relationships between nutrients
In order to understand the statistical models that explain the
variance in GI and II values, it is essential to establish the
relationships that exist between the different carbohydrate
fractions and the other nutritional components. Due to the
Fig. 2. (A) glycaemic index (GI), (B) starch gelatinisation index (SGI) and (C) slowly
available glucose (SAG) content of subsets of three products from each of three cereal
groups. BC, breakfast-cereals group; BP&C, bakery-products and crackers group; Bi,
biscuit group. Values are means with standard deviations shown by vertical bars. For
details of subjects, cereal products and procedures, see Table 1 and p. 330. Mean
values were significantly different from those of the Bi group: *P, 0·05, †P¼ 0·08.
Glycaemic indices of cereal products 333
design of the present study, with the carbohydrate content of
the test meals fixed at 50 g carbohydrate, there is necessarily
an interaction between those carbohydrate fractions
described in the five models where a reciprocal relation
with total carbohydrate exists, e.g. starch:sugar (model 1),
total glucose:fructose (model 3). In addition, because the
models build progressively on one another, a number of
other correlations between carbohydrate fractions could be
explained by their close links with other associated
analytical fractions in different models, e.g. sugars:free
sugar glucose:fructose (models 1 and 2), starch:total
glucose:glycaemic glucose:RAG (models 1– 5).
A summary of the significant univariate correlations
between nutrients is shown in Table 2, which focuses on
the RAG fraction in order to demonstrate its relative associ-
ations with other carbohydrate components. The positive
correlation between starch and RAG in these products can
be explained by the high proportion of the starch that is
included in the RAG fraction (mean value 79·1 (range
55·3– 94·3) %) compared with the SAG (mean value 17·5
(range 2·2– 41·2) %) or RS (mean value 3·4 (range 0·0–
8·9) %) fractions. The total glucose and glycaemic glucose
fractions, which combine the glucose components from
starch and sugar, have strong positive correlations with
RAG, reflecting the fact that an average of 83·5 % total glu-
cose in the products is included in the RAG fraction.
In model 5, RAG was correlated with total fructose, SAG
and RS, with the SAG fraction demonstrating the strongest
relationship. No significant relationships were observed
between SAG and either starch or sugar, indicating that
the SAG content of the products must be determined by
the food-processing technique. Whilst the positive corre-
lation between RS and starch is to be expected, there was
a negative correlation between RS and NSP. There was no
significant correlation between NSP and starch contents,
and this finding remained even after the exclusion of
Alpen, with its exceptionally high NSP value (in part due
to its fruit and nut components), from the analysis.
The correlation of fat with RAG and SAG in these
products is of particular interest. To some extent, this
relationship reflects the different characteristics of the pro-
duct groups, with BC having significantly lower fat and
SAG contents than Bi, and with BP&C intermediate in
SAG but relatively high in fat content. Even so, positive
correlations between fat and SAG were apparent within
each group, but did not reach significance due to the
small sample sizes. RAG was also positively correlated
with protein content.
Relationships with the glycaemic and insulinaemic indices
In order to explain the variance in GI and II values, their
correlation with the chemical and digestibility character-
istics of the cereal products was investigated. The relation-
ships between the carbohydrate fractions and the GI values
are shown in Fig. 3, together with the total variance
explained by each of the models 1– 5. Neither starch nor
sugar in model 1 were significantly correlated with GI, as
was the case for free sugar glucose (model 2) and fructose
(models 2 – 5). The positive correlations of GI with total
glucose (model 3) and RS (models 4 and 5) can be
explained by the analytical association that exists between
these carbohydrate fractions and the RAG fraction
(Table 2). GI was strongly correlated with both RAG and
SAG contents in model 5.
The variance in GI that could be explained by the differ-
ent carbohydrate models ranged from 33·8 % for model 1 to
68·8 % for model 5 (Fig. 3). The combined effect of fat, pro-
tein and NSP was to describe 59·1 % of the variance in GI,
but of these, only the fat content was significant (r
2
0·52,
P, 0·01). The simplified model that best described the
variance in GI combined the strongest variables, SAG and
fat, and accounted for 73·1 % (Fig. 4). The partial corre-
lations of SAG and fat with GI within this model suggest
that SAG is the dominant factor in comparison with fat.
The relationship between the carbohydrate fractions and
the II values is summarised in Fig. 3, together with the var-
iance in II explained by each of the models 1 –5. Total
fructose exhibited the strongest negative correlation and,
through association, explained the correlations observed
with sugars and free sugar glucose. Positive correlations
were observed for total starch, total glucose and glycaemic
glucose, all of which can be related to their association
with RAG, which had the strongest positive correlation.
A negative correlation between II and SAG was observed.
The variance in II that could be explained by the differ-
ent carbohydrate models ranged from 27·6 % for model 3
to 41·1 % for model 5 (Fig. 3). Fat, protein and NSP
together accounted for 46·4 % of the variance, the majority
of which was explained by the positive correlation with
protein (r
2
0·33, P, 0·01). The simplified model that
best described the variance in II combined RAG and pro-
tein, and accounted for 45·0 %, (Fig. 4), with RAG and pro-
tein having similar partial correlations within this model.
Discussion
The present study has investigated the relationship between
the chemical and digestibility characteristics of a selection
of cereal products and their physiological properties
(GI and II). Various models were investigated, which
Table 2. Selected relationships between the nutritional components
of the cereal products investigated†
(Univariate correlation coefficients)
Protein Starch RS RAG SAG
NSP 0·443* 2 0·620**
Fat 2 0·465* 0·578**
Protein 0·462* 0·458*
Starch 0·469* 0·705**
Sugar 2 0·567**
Fructose 2 0·573**
FSG 2 0·558**
Total
glucose
0·567** 0·816** 2 0·427*
Glycaemic
glucose
0·803** 2 0·443*
RS 0·513* 2 0·419*
RAG 2 0·866**
RS, resistant starch; RAG, rapidly available glucose; SAG, slowly available
glucose; FSG, free sugar glucose.
*P, 0·05, **P, 0·01.
† For details of products and procedures, see Table 1 and p. 330.
K. N. Englyst et al.334
characterised the carbohydrate fraction progressively from
solely chemical attributes in model 1 (starch and sugars) to
a profile of the rate and extent of release in model 5 (fruc-
tose, RAG, SAG, and RS). The division of the glycaemic
glucose fraction into RAG and SAG fractions (model 5)
was found to describe 68·8 % of the variance in GI, com-
pared with 33·8 % explained by the division into starch
and sugars.
The twenty-three products investigated encompass a
range of ingredients and processing techniques that con-
tribute to defining their physico-chemical properties.
Although starch is the main constituent of the products
investigated in the present study, they also had a range
of sugar, fat, protein and NSP contents. Each of these
components has been found to influence glycaemic
response, with much of this evidence based on observing
the effects of addition of the component of interest to
carbohydrate meals (Jenkins et al. 1978; Gannon et al.
1993; van Loon et al. 2000). The few studies that have
specifically investigated the relationship between
nutrient composition and GI value can be difficult to inter-
pret as they incorporate a varied range of food groups.
For example, Trout et al. (1993) identified a negative
relationship between protein content and GI, but this dis-
appeared when the legume group, with its high-protein
content and low GI value, was considered separately
from the rest of the products. The present study was limited
to the investigation of processed cereal products, thereby
eliminating any interference associated with a wider
range of food groups. This focus on cereal products did
not restrict the range of GI and II values, but rather illus-
trated that the type of cereal product consumed can have
a marked influence on physiological responses.
For the products investigated, the determinants of the
digestibility profile of the carbohydrate fraction are the
starch:sugar ratio used, the cereal type and the degree of
food processing. The production of the breakfast cereals
and bakery products investigated involves heating in the
presence of moisture, which results in the gelatinisation
of starch and its consequent rapid digestion. In contrast,
several of the biscuits investigated were produced by
baking under very-low-moisture conditions, which
reduces the extent of starch gelatinisation and results in
partially intact starch granules that are less susceptible to
Fig. 3. Relationships between (A) glycaemic index and (B) insulinaemic index of twenty-three cereal products and their carbohydrate compo-
sition expressed as five models describing different chemical and digestibility characteristics. FSG, free sugar glucose; RS, resistant starch;
RAG, rapidly available glucose; SAG, slowly available glucose. For details of subjects, cereal products and procedures, see Table 1 and
p. 330. The univariate correlation coefficients (r ) for the individual carbohydrate fractions are provided alongside their labels *P , 0·05,
**P , 0·01. The variance explained by each full model is derived from multiple covariate regression analysis (r
2
values).
Glycaemic indices of cereal products 335
Fig. 4. Simplified models of the variance in glycaemic index (GI; (A), (B), (C)) and insulinaemic index (II; (D), (E), (F)) of twenty-three cereal
products, explained by their chemical and carbohydrate digestibility characteristics. SAG, slowly available glucose; RAG, rapidly available glu-
cose. (A), GI model with SAG + fat; GI ¼ 85·8 2 2·335 £ SAG þ 1·493 £ fat; r
2
¼ 0·731; (B), SAG (within GI model), r
2
¼ 0·445; (C), fat (within
GI model), r
2
¼ 0·272; (D), II model with RAG + protein; II ¼ 35·8 þ 0·821 £ RAG 2 2·624 £ protein; r
2
¼ 0·450; (E), RAG (within II model),
r
2
¼ 0·178; (F), protein (within II model), r
2
¼ 0·184. For details of subjects, cereal products and procedures, see Table 1 and p. 330.
K. N. Englyst et al.336
the action of amylolytic enzymes (Bornet et al. 1989). This
relationship between a low extent of starch gelatinisation
and high SAG values was confirmed in the present study
by the low starch gelatinisation index for the Bi group
compared with the BP&C group (Fig. 2).
Several of the relationships between nutrients identified
in the present study are of interest, both in characterising
the overall physico-chemical profile of the products, and
in explaining observed correlations with GI and II values.
For instance, the positive correlation between RS and GI
is difficult to explain on its own, as the RS fraction is
not absorbed in the small intestine and therefore cannot
elicit a glycaemic response. Indeed, a negative relationship
between RS content and GI has been reported, though the
high RS products in that study contained 30 –50 % starch
as RS (Bjo
¨
rck et al. 2000), compared with an average
RS content of 3·5 (range 0·0– 9·0) % starch in the present
study. Our present finding of a positive correlation between
RS and GI is explained by the positive relationship that
exists between RS and RAG in these products. This is typi-
cal of products where starch has been gelatinised, with the
majority being readily digestible, except for the small
amount of retrograded starch that resists digestion.
In the present study, the inter-relationship between RAG
and SAG makes it difficult to identify which of these fac-
tors explains the strong correlations that they have with GI.
However, we have previously shown that it is RAG, by
virtue of its rapid digestion and absorption in the small
intestine, that is responsible for the postprandial rise in
blood glucose concentrations (Englyst et al. 1999). It
follows that SAG exerts its reductive effect on GI values
by replacing RAG in the test meal. An increase in the pro-
portion of total fructose in a test meal will also result in a
reduction in RAG and explains why foods high in fructose
or sucrose have been reported to have only moderate GI
values (Brand-Miller et al. 1995). In the present study, a
correlation between fructose and GI was not observed,
despite the fact that it replaced part of the RAG fraction.
Consequently, this is likely to explain why the RAG frac-
tion described only 55 % of the variance in GI compared
with the 63 % explained by the SAG fraction. Despite its
low GI value, high intakes of fructose are generally not
recommended due to concerns about adverse effects on
lipid profiles (Daly et al. 1997). It seems probable, there-
fore, that the greatest health benefits of low-GI diets will
accompany those that contain carbohydrates that are
slowly digested and absorbed.
The 51·5 % variance in GI described by fat content is not
fully explained by the association between fat and SAG,
which together described 73 % of the variance in GI
(Fig. 4). Numerous studies have demonstrated that fat
can lower the glycaemic response to foods (Welch et al.
1987; Collier et al. 1988; Gannon et al. 1993). However,
a recent study found that while the addition of 40 g fat to
a 75 g carbohydrate meal of pasta delayed the appearance
of exogenous glucose in blood, this effect was not apparent
with the addition of 15 g fat to the meal (Normand et al.
2001). The relatively small range of intake (mean value
8·1 (range 0·4–13·2) g) in the present study would indicate
that an effect of fat on gastric emptying is unlikely.
Another possibility is that starch –lipid complexes may
have formed during the processing of some of the cereal
products, and this could restrict enzymatic hydrolysis
(Biliaderis, 1991; Crowe et al. 2000). However, amy-
lose– lipid complexes are developed with non-esterified
fatty acids or monoacylglycerol, which are present in
very small quantities in the cereal products investigated.
It is also possible that the fat component of the meals trig-
gered an incretin response with an insulinotropic effect
(Welch et al. 1987; Collier et al. 1988), but if this was
the case, the effect was either inconsistent or masked by
other factors, as no significant correlation between fat
and II was observed. The finding that the two high-fat
baguette meals had moderately high GI values lends
further support to the suggestion that fat per se only has
a minor effect on GI in the present study.
Several studies have included measures of II values,
which are of interest due to the role of insulin in glucose
homeostasis and its regulatory effects in lipid metabolism.
In addition, the large insulin demand associated with high-
GI diets has been proposed to be involved in the aetiology
of diabetes (Salmeron et al. 1997; Wolever, 2000). The
present study confirms the relationship between GI and II
values for starchy foods, although the correlation is not
as strong as reported previously (Bjo
¨
rck et al. 2000). It is
possible that this rather weak association, and the obser-
vation that II values were higher than GI values, could
be explained by the combined insulinotropic effects of pro-
tein, fat and possibly of other undetermined properties of
the foods. This is supported by the investigation into the
glycaemic and insulinaemic responses to 1000 kJ portions
of a range of foods (Holt et al. 1997). Holt et al. (1997)
could only explain 23 % of the variance in the insulin
score by the glycaemic score of the foods, and only a
further 10 % could be accounted for by the macronutrient
composition. Of the carbohydrate fractions investigated
in the present study, RAG demonstrated the strongest cor-
relation with II, but still only explained 32 % of the var-
iance. The observed positive correlation between protein
and II is in agreement with previous findings (Kabadi,
1991; Trout et al. 1993; Brand-Miller et al. 1995), although
the association between RAG and protein may in part
explain this relationship. The model incorporating RAG
and protein accounted for 45 % variance in II, considerably
less than could be identified for GI.
The present study has demonstrated that RAG and SAG,
describing the rate of carbohydrate release from foods, are
the carbohydrate fractions that best describe the variability
in physiological attributes of the cereal products investi-
gated. Although there appeared to be independent effects
of fat on GI and of protein on II, these relationships may
in part be explained by their correlations with RAG and
SAG. The effect of protein and fat should not be comple-
tely dismissed, and indeed their presence in the foods
investigated is probably responsible for some of the
variance observed in GI and especially II values. This
acts to highlight further the consistency of the effect of
RAG and SAG on the physiological responses, in spite
of the complexity of factors that can have an influence. It
is apparent from the present study that food processing is
the major determinant of the state of gelatinisation of
starch and the SAG content of cereal products. The Bi
Glycaemic indices of cereal products 337
group had the highest mean SAG content, although this
group also exhibited the greatest range in the proportion
of SAG in the starch fraction, reflecting the diversity in
processing methods used in biscuit production. This
demonstrates that limiting the extent of starch gelatinisa-
tion, as occurs with some types of biscuit manufacture, rep-
resents a feasible method by which to lower the GI values
of cereal products.
In conclusion, the present paper has shown that carbo-
hydrate identity and food processing largely determine
the variation in GI values of cereal products, and that
this is adequately reflected by the classification scheme
describing the rate of carbohydrate release from foods.
Specifically, the SAG measurement allows the identifi-
cation of those low-GI foods containing carbohydrates
that are slowly digested and absorbed, for which health
benefits are likely to be associated. We suggest that the
proposed classification scheme would be valuable in the
further elucidation of the mechanisms by which carbo-
hydrate quality can influence health.
Acknowledgements
We thank Marinos Elia and Steve Wootton from the Insti-
tute of Human Nutrition, University of Southampton, for
their support and advice. This work was supported by
Danone Vitapole.
References
Bellisle F (2001) Glycaemic Index and Health: the Quality of the
Evidence. Montrouge, France: John Libbey Eurotext.
Biliaderis CG, Maurice TJ & Vose JR (1980) Starch gelatinisation
phenomena studied by scanning differential calorimetry. Jour-
nal of Food Science 45, 1669–1675.
Biliaderis CG (1991) The structure and interactions of starch with
food constituents. Canadian Journal of Physiology and Phar-
macology 69, 60 –78.
Bjo
¨
rck I, Liljeberg H & Ostman E (2000) Low glycaemic-index
foods. British Journal of Nutrition 83, Suppl., S149–S155.
Bornet FRJ, Fontvielle AM, Rizkalla S, Colonna P, Blayo A,
Mercier C & Slama G (1989) Insulin and glycaemic responses
in healthy humans to native starches processed in different
ways: correlation with in vitro a-amylase hydrolysis. American
Journal of Clinical Nutrition 50, 315– 323.
Brand-Miller JC (1994) Importance of glycaemic index in dia-
betes. American Journal of Clinical Nutrition 59, Suppl.,
747S–752S.
Brand-Miller J, Pang E & Broomhead L (1995) The glycaemic
index of foods containing sugars: comparison of foods with
naturally occurring v. refined sugars. British Journal of Nutri-
tion 73, 613 –623.
Collier GR, Greenberg GR, Wolever TMS & Jenkins DJA (1988)
The acute effect of fat on insulin secretion. Journal of Clinical
Endocrinology and Metabolism 66, 323 –326.
Colonna P, Leloup V & Buleon A (1992) Limiting factors of
starch hydrolysis. European Journal of Clinical Nutrition 46,
Suppl., S17– S32.
Crowe TC, Seliman SA & Copeland L (2000) Inhibition of enzy-
matic digestion of amylose by free fatty acids in vitro contri-
butes to resistant starch formation. Journal of Nutrition 130,
2006–2008.
Daly ME, Vale C, Walker M, Alberti KGMM & Mathers JC
(1997) Dietary carbohydrates and insulin sensitivity: A
review of the evidence and clinical implications. American
Journal of Clinical Nutrition 66, 1072– 1085.
Englyst HN & Hudson GJ (1996) The classification and measure-
ment of dietary carbohydrates. Food Chemistry 57, 15– 21.
Englyst HN, Kingman SM & Cummings JH (1992) Classifi-
cations and measurement of nutritionally important starch frac-
tions. European Journal of Clinical Nutrition 46, Suppl.,
S33–S50.
Englyst HN, Quigley ME & Hudson GJ (1994) Determination of
dietary fibre as non-starch polysaccharides with gas-liquid
chromatographic, high-performance liquid chromatographic or
spectrophotometric measurement of constituent sugars. Analyst
119, 1497– 1509.
Englyst KN, Englyst HN, Hudson GJ, Cole TJ & Cummings JH
(1999) Rapidly available glucose in foods: an in vitro measure-
ment that reflects the glycaemic response. American Journal of
Clinical Nutrition 69, 448– 454.
Englyst KN, Hudson GJ & Englyst HN (2000) Starch analysis in
food. In Encyclopaedia of Analytical Chemistry, pp.
4246–4262 [RA Meyers, editor]. Chichester, Sussex: John
Wiley & Sons.
Food and Agriculture Organization/World Health Organization
(1998) Expert Consultation: Carbohydrates in Human Nutri-
tion, Food and Agriculture Organization Food and Nutrition
Paper no. 66. Geneva: FAO/WHO.
Gannon MC, Nuttall FQ, Westpal SA & Seaquist ER (1993) The
effect of fat and carbohydrate on plasma glucose, insulin
C-peptide and triglycerides in normal male subjects. Journal
of American College of Nutrition 12, 36 – 41.
Gallant DJ, Bouchet B, Buleon A & Perez S (1992) Physical
characteristics of starch granules and susceptibility to enzy-
matic degradation. European Journal of Clinical Nutrition
46, Suppl., S3 – S16.
Granfeldt Y, Bjo
¨
rck AI, Drews A & Tovar J (1992) An in vitro
procedure based on chewing to predict metabolic response to
starch in cereal and legume products. European Journal of
Clinical Nutrition 46, 649– 660.
Heaton KW, Marcus SN, Emmett PM & Bolton CH (1988) Particle
size of wheat, maize, and oat test meals: effects on plasma glucose
and insulin responses and on the rate of starch digestion in vitro.
American Journal of Clinical Nutrition 47, 675–682.
Heijene MLA, van Amelsvoort JMM & Westrate JA (1995) Inter-
action between physical structure and amylose:amylopectin
ratio of foods on postprandial glucose and insulin responses
in healthy subjects. European Journal of Clinical Nutrition
49, 446– 457.
Holm J, Lundquist I, Bjo
¨
rck I, Eilasson AC & Asp NG (1988)
Degree of starch gelatinization, digestion rate of starch in
vitro, and metabolic response in rats. American Journal of
Clinical Nutrition 47, 1010– 1016.
Holt SHA & Brand-Miller JC (1994) Particle size, satiety and the
glycaemic response. European Journal of Clinical Nutrition 48,
496–502.
Holt SHA, Brand-Miller JC & Petocz P (1997) An insulin index
of foods: the insulin demand generated by 1000 kJ portions of
common foods. American Journal of Clinical Nutrition 66,
1264–1276.
Jenkins DJA, Ghafari H, Wolever TMS, Taylor RH, Jenkins AL,
Barker HM, Fielden H & Bowling AC (1982) Relationship
between rate of digestion of foods and post-prandial glycaemia.
Diabetologia 22, 450 –455.
Jenkins DJA, Wolever TMS, Leeds AR, Gassule MA, Dilawari
JB, Goff DV, Metz GL & Alberti KGMM (1978) Dietary
fibres, fibre analogues and glucose tolerance: importance of
viscosity. British Medical Journal 1, 1392– 1394.
K. N. Englyst et al.338
Jenkins DJA, Wolever TMS & Taylor RH (1981) Glycaemic
index of foods: a physiological basis for carbohydrate exchange.
American Journal of Clinical Nutrition 134, 362–366.
Kabadi UM (1991) Dose-kinetics of pancreatic a- and b-cell
responses to a protein meal in normal subjects. Metabolism
40, 236– 240.
Lee BM & Wolever TMS (1998) Effect of glucose, sucrose and
fructose on plasma glucose and insulin responses in normal
humans: comparison with white bread. European Journal of
Clinical Nutrition 52, 924– 928.
Normand S, Khalfallah Y, Louche-Pelissier C, Pachiaudi C,
Antoine J-M, Blanc S, Desage M, Riou JP & Laville M
(2001) Influence of dietary fat on postprandial glucose metab-
olism (exogenous and endogenous) using intrinsically
13
C-
enriched durum wheat. British Journal of Nutrition 86, 3 –11.
O’Dea K, Snow P & Nestel P (1981) Rate of starch hydrolysis in
vitro as a predictor of metabolic responses to complex carbo-
hydrate in vivo. American Journal of Clinical Nutrition 34,
1991–1993.
Prochaska LJ, Nguyen XT, Donat N & Piekutowski WV (2000)
Effect of food processing on the thermodynamic and nutritive
value of foods: literature and database survey. Medical Hypo-
theses 54, 254 –262.
Salmeron J, Manson JE, Stampfer MJ, Colditz GA, Wing AL &
Willett WC (1997) Dietary fiber, glycaemic load and risk of
non-insulin-dependent diabetes mellitus in women. Journal of
American Medical Association 277, 472 –477.
Thomsen C, Rasmussen OW, Christiansen C, Andreasen F, Poul-
sen PL & Hermansen K (1994) The glycaemic index of spa-
ghetti and gastric emptying in non-insulin-dependent
diabetic patients. European Journal of Clinical Nutrition 48,
776–780.
Trout DL, Behall KM & Osilesi O (1993) Prediction of glycaemic
index for starchy foods. American Journal of Clinical Nutrition
58, 873– 878.
van Loon LJC, Saris WHM, Verhagen H & Wagemakers AJM
(2000) Plasma insulin responses of different amino acid or
protein mixtures with carbohydrate. American Journal of Clini-
cal Nutrition 72, 96– 105.
Welch I McL, Bruce C, Hill SE & Read NW (1987) Duodenal
and ileal lipid suppresses postprandial blood glucose and insu-
lin responses in man: possible implications for the management
of diabetes mellitus. Clinical Science 72, 209 –216.
Wolever TMS (1991) The glycaemic index: Methodology and
clinical implications. American Journal of Clinical Nutrition
54, 846– 854.
Wolever TMS (1997) The glycaemic index: flogging a dead
horse? Diabetes Care 20, 452 –456.
Wolever TMS (2000) Dietary carbohydrates and insulin action in
humans. British Journal of Nutrition 83, Suppl., S97–S102.
Glycaemic indices of cereal products 339