You are on page 1of 10

European Journal of Clinical Nutrition (2013), 110

& 2013 Macmillan Publishers Limited All rights reserved 0954-3007/13


www.nature.com/ejcn

REVIEW

Is glycaemic index (GI) a valid measure of carbohydrate quality?


TMS Wolever1,2,3,4
Recent criticisms of the glycaemic index (GI) focus on its validity with assertions that GI methodology is not valid, GI values are
inaccurate and imprecise, GI does not predict what foods are healthy and that whole grain and bre are better markers of
carbohydrate quality than GI. None of the critics provide sound reasons for rejecting GI because some of their arguments are based
on agrant errors in understanding and interpretation while others are not supported by current data or are inconsistent with other
nutritional recommendations. This paper addresses current criticisms of GI and outlines reasons why GI is valid: (1) GI methodology
is accurate and precise enough for practical use; (2) GI is a property of foods; and (3) GI is biologically meaningful and relevant to
virtually everyone. Current dietary guidelines recommend increased consumption of whole grains and dietary bre but do not
mention GI. However, this is illogical because the evidence that GI affects health outcomes is at least as good or better than that for
whole grains and bre. GI is a novel concept from a regulatory point of view and a number of problems need to be addressed to
successfully translate GI knowledge into practice. The problems are not insurmountable but no progress can be made until bias and
misunderstanding about GI can be overcome and there is better agreement about what is the actual state of knowledge on GI so
that the real issues can be identied and addressed.
European Journal of Clinical Nutrition advance online publication, 13 February 2013; doi:10.1038/ejcn.2013.27
Keywords: Human nutrition; dietary carbohydrates; nutrition recommendations; public health

The glycaemic index (GI) was proposed over 30 years ago as a


classication of the blood glucose-raising potential of carbohydrate containing foods1 and it has been controversial ever since.
Early critics of GI acknowledged there were differences among
foods but believed they were not clinically important so that in
1997 GI was dismissed as being much ado about (almost)
nothing.2 Since then, however, the force of the criticisms has
escalated to attacks on the validity of the GI with assertions such
as: ...[the] recommended GI methodology is not well standardized
and has several aws....3 and ...[the] GI method does not measure
a meaningful property of a food..4 In 1997, I accused the critics of
GI of ...misinterpreting data, misquoting the literature and
misusing statistics.5 The same is true now.
This paper addresses the question of whether the GI is a valid
marker of carbohydrate quality. The Oxford dictionary6 denes
valid as sound, defensible; well-founded. This subjective
denition suggests that validity, like beauty, is in the eye of the
beholder. I believe that the GI is valid because it fullls the
following conditions: (1) GI methodology is standard, accurate and
precise; (2) GI is a property of foods (that is, it has the same value
in nearly everyone); and (3) GI is biologically meaningful and
inuences outcomes in health and disease in ways that are
relevant to almost everyone. Furthermore, there is good evidence
that GI benecially inuences more and different health outcomes
than other commonly used markers of carbohydrate quality such
as whole grains and bre and, therefore, ought to be part of
nutrition recommendations.
ROLE OF GI IN HEALTH AND DISEASE
Initially, the only clinical application for GI was in the management
of diabetes, and there is now excellent evidence that low-GI diets

improve glycaemic control in diabetes.7,8 However, with the


demonstration in 1997 that low-GI diets were associated with
reduced risk for type 2 diabetes,9,10 a nding which has been
supported by most large subsequent studies,1116 GI graduated
from a narrow role in managing diabetes to a much wider role in
the maintenance of health.
The World Health Organization denes health as a state of
complete physical, mental and social well-being and not merely
the absence of disease or inrmity;17 GI may have relevance in all
these areas of well-being. With respect to physical and mental
well-being, there is evidence that the GI inuences
exercise performance18 and cognitive function.19,20 A role for GI
in social well-being could be proposed from data showing that a
low-GI diet reduces the severity of acne21,22 and inuences
pregnancy outcomes.2328 Regarding the absence of disease or
inrmity there is excellent evidence that low-GI diets reduce the
risk for diabetes29 and assist with weight maintenance.30 In
addition, there is good or emerging evidence that low-GI diets
reduce the risk for cardiovascular disease,31,32 stroke,33,34 metabolic
syndrome,35,36 breast,37 prostate,38 colo-rectal, pancreatic and
other cancers,31 uterine broids,39 depression,40 Parkinsons disease,41 chronic kidney disease,42 eye disease,29 inammation,43 gall
stones,44,45 neural tube defects46 and mortality from inammatory
disease.47 All this suggests that GI inuences physiological
functions in ways that are relevant to virtually everyone.
However, there has been reluctance to incorporate GI into dietary recommendations. Such reluctance would be acceptable if it
was based on valid arguments that were applied consistently for
other nutritional recommendations; but it is not.
Early on major concerns about GI were that it limited food
choice and had questionable clinical utility because differences in

1
Department of Nutritional Sciences, Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada; 2Division of Endocrinology and Metabolism, Department of Medicine,
St Michaels Hospital, Toronto, Ontario, Canada; 3Keenan Research Centre of the Li Ka Shing Knowledge Institute, St Michaels Hospital, Toronto, Ontario, Canada and 4Glycemic
Index Laboratories, Inc., Toronto, Ontario, Canada. Correspondence: Dr TMS Wolever, Department of Nutritional Sciences, Faculty of Medicine, University of Toronto, 150 College
Street, Toronto, Ontario, M5S 3E2, Canada.
E-mail: thomas.wolever@utoronto.ca
Received 11 September 2012; accepted 10 January 2013

Validity of glycaemic index


TMS Wolever

2
glycaemic response between individual foods were not maintained in mixed meals; I refuted these concerns in a commentary
published in 1997.5 However, inuential nutrition experts continue
to assert that GI does not apply in mixed meals3,4851 and cite, as
supporting evidence, the conclusions of studies I have already
argued are not supported by their results. Unfortunately, by
simply continuing to site such studies without pointing out what is
wrong with my arguments,2 progress towards consensus in this
area is not being made. In addition, new concerns are being
raised, namely that: the method for measuring GI is not
standardized3 and is inaccurate and imprecise;3,4,4851 the
international GI tables are fraught with uncertainty and
irreproducibility49 and do not accurately reect what is eaten in
terms of variety, methods, amounts and processing;51 dietary bre
and whole grains are better indicators of carbohydrate quality
than GI;52 and the GI of a food or diet does not predict whether it
is healthy.51 Some of these concerns are reasonable, but many are
not, being based on a misunderstanding about the meaning and
utility of GI.
DEFINITION AND MEANING OF GI
GI is usually dened as the incremental area under the blood
glucose response curve (AUC) elicited by a 50-g available
carbohydrate (avCHO) of a test food expressed as a percentage
of the response elicited by 50 g glucose in the same subject.
However, this is an explanation of the GI concept rather than a
precise description of GI methodology and, perhaps, has been a
source of misunderstanding. Therefore, to be precise, GI is
mathematically dened as follows:
xP
1

GI

Figure 1. Incremental AUC and estimates of GI in 31 subjects (9


normal, 12 hyperinsulinaemic and 10 with type 2 diabetes) elicited
by 50 g avCHO portions of five different carbohydrate foods (instant
potato, white bread, spaghetti, barley and sucrose); values are
meanss.d. for the five test meals in each subject. AUC differs
significantly among subjects but GI does not. Inset: the amounts of
variance in AUC and GI due to the different foods, the different
subjects (between subjects) and day-to-day variation (within
subjects) from repeated-measures analysis of variance. Expressing
results as GI reduced variation between subjects by 95%. Data
derived from Lan-Pidhainy and Wolever56 Figure adapted from
Wolever.57

subjects, that is, the individual values of 100Fx /Gx , reect day-today variation in glycaemic responses, not true between-subject
variation in GI. The value for GI obtained in an individual is not
that subjects GI, it is the estimate of the food GI obtained in that
subject. Since within-individual variation is large, the GI of a food is
the mean value in at least 10 subjects.

100Fx /Gx

1

where Fx is the incremental AUC in subject x elicited by 50 g


avCHO from the test food and Gx is mean the AUC in subject x
elicited by 50 g glucose tested on two or three separate occasions.
The GI is the mean of these values in n subjects; the current
internationally accepted GI method53 stipulates that nX10. Thus,
any so-called GI value not based on Equation (1) with nX10 is not
a valid GI.
Several key implications of the denition of GI as given in
Equation (1) are GI is a property of foods that is measured in
human subjects; GI is not a glycaemic response; and GI is a
property of high carbohydrate foods tested alone (not as part of a
mixed meal).
GI is a property of foods
GI was intended to indicate the extent to which the available
carbohydrate (avCHO) in a food raises blood glucose relative to an
equal weight of glucose. It was reasoned that, by expressing the
results as a percentage of oral glucose, any differences between
subjects would be normalized; that is, the GI of a food would have
the same value in nearly everyone. There now is a good deal of
evidence to support this. When each of 12 subjects with diabetes
tested bread, rice and spaghetti 4 times, the mean glycaemic
responses varied among subjects over a 4.3-fold range (Po0.001)
but mean GI estimates were similar among subjects.54 There was
no signicant effect of age, sex, body mass index or ethnicity on
the mean GI of 3 foods tested in 77 non-diabetic subjects selected
to differ by these characteristics.55 We recently showed that the
mean GI of ve carbohydrate foods (sucrose, instant potato, bread,
rice and barley) did not differ signicantly among normal,
hyperinsulinaemic and type 2 diabetic subjects56 (Figure 1). If
there is no signicant between-individual variation, then it follows
that GI is a property of the food whose value is similar in all
subjects. Thus, variation of GI estimates obtained in individual
European Journal of Clinical Nutrition (2013) 1 10

GI is not a glycaemic response


A common mistake is to discuss GI as if it were equivalent
to a glycaemic response. For instance one expert51 states that:
Eating a food as part of a mixed dish or meal changes the
GI. For example, the addition of nuts can drop the GI of a
food by as much as half. This is wrong. The glycaemic response of
the meal is reduced by the addition of nuts, not the GI
of the bread. In fact, the relative glycaemic responses of foods
are roughly maintained when other nutrients are added
to them. For example, 37.5 g fat reduced the glycaemic
response elicited by both potato and lentils by 4050%;58 thus,
the response of lentils was about 4050% less than that of
potatoes either with or without fat. Similar results were found by
Bornet et al.59 and Henry et al.60 Thus, GI is one determinant of
glycaemic response; others include the amounts of carbohydrate,
fat and protein consumed.
If eating a food in the context of a mixed meal altered its GI,
then the GI of individual foods would not explain the glycaemic
responses of mixed meals. However, the glycaemic responses
elicited by meals of equivalent carbohydrate, fat and protein
contents but containing different carbohydrate foods is directly
proportional to the meal GI calculated from the GI of the
individual foods it contains5,61 (Figure 2). Moreover, the glycaemic
responses elicited by typical breakfast meals containing unequal
amounts of energy (220450 Kcal), fat (018 g), protein (018 g)
and carbohydrate (1679 g) and differing in GI (35100) were
almost entirely explained by the amount of carbohydrate and
calculated meal GI,62 which together accounted for nearly 90% of
the variation in mean AUC (Figure 3), at least for meals within this
range of nutrient composition. Thus, eating a food as part of a
mixed meal affects the glycaemic response, but does not alter the
foods GI. The degree to which the glycaemic response of a food is
altered in the presence of other foods depends on the amount
and source (GI) of carbohydrate and the amounts and types of fat
and protein added.
& 2013 Macmillan Publishers Limited

Validity of glycaemic index


TMS Wolever

3
Does calculated meal GI predict measured meal GI?
In 1984, we suggested63 that the GI (GImeal) of a meal containing n
foods be calculated as follows:
xP
1

GImeal

Figure 2. Relationship between calculated meal GI and the mean


incremental AUC elicited by five different meals in subjects with
type 2 diabetes (n 6). Each meal contained 70 g avCHO, 2634 g
protein and 1214 g fat.61 Points represent mean AUC; the solid line
is the regression line (and 95% confidence interval) and the dotdash line is the regression forced through the origin (not
significantly different from the regression line). The correlation
coefficient, r 0.987 (P 0.002).

Figure 3. Relationship between the mean incremental AUC elicited


by 14 different meals in normal subjects (Toronto, Canada, 8 meals,
n 10; Sydney, Australia, 6 meals, n 16) and 1.5  avHCO 1.4 
GI  46, where avCHO is the meal avCHO content and GI is meal GI
calculated using Equation (2). The meals varied in energy (220
450 Kcal), protein (018 g), fat (018 g), avCHO (1679 g) and GI (35
100).62 The b coefficients and constant used to calculate the values
on the x axis were derived from multiple regression analysis and the
points represent mean AUC. The solid line is the regression line
(r2 0.882, Po0.001 for both avCHO and GI). Data from Wolever
et al.62 Adapted from Wolever.57

Table 1.

GIx Cx

2

Cmeal

where GIx is the GI of the xth food, Cx is the amount of avCHO in


the xth food and Cmeal is the amount of avCHO in the meal.
Calculated meal GI is proportional to meal glycaemic response
when comparing meals of equivalent nutrient content.5,61
Recently, however, investigators have begun to measure the GI
of mixed meals using a method analogous to that used to
measure the GI of foods, except that subjects consume a portion
of the mixed meal containing 50 g avCHO.6466 The conclusion of
these studies is that the calculated meal GI is an imprecise
estimate of the measured meal GI. However, this is not
unexpected because calculated meal GI only takes into account
the source and amount of avCHO in the meal, whereas the
measured meal GI is also determined by the amounts of fat and
protein it contains. Based on doseresponse studies, we
performed67,68 the effect on glycaemic responses of adding fat
(corn oil) and protein (soy or whey) to carbohydrate (oral glucose
solution) can be estimated. Using this information, it can be
shown that differences in the fat and protein contents of the
mixed meals fed by Dodd et al.65 largely account for the difference
between measured and calculated meal GI.69 Table 1 shows, using
data from Hatonen et al.,66 how to adjust calculated meal GI for
differences in avCHO, fat and protein; after accounting for these
differences, adjusted meal GI is closely related to measured meal
GI (Figure 4).
Thus, the results of studies showing that calculated meal GI
does not predict measured meal GI suggest that it is inappropriate
to measure the GI of mixed meals. It is proper to measure the GI of
high carbohydrate foods fed alone and to use the resulting values
to calculate a meal GI. The relative glycaemic response of a meal is
determined by its calculated meal GI and the amounts of avCHO,
fat, and protein it contains.
GI is a property of high carbohydrate foods
It is common to ascribe low carbohydrate foods a GI value of
0.7173 However, this is inappropriate because GI is a property of
high carbohydrate foods; also, it is meaningless. Conceptually, the
GI of a 0-g carbohydrate food would be the glycaemic response
elicited by a portion of food containing 0 g avCHO (AUC 0)
expressed as a % of that elicited by 0 g glucose (AUC 0); that is,
GI 100  0/0. This equation can be rearranged to: 0  GI 0,

Adjustment of calculated meal GI for differences in fat, protein and carbohydrate content of mixed meals

Test meal

Potato (P)
P Oil (O)
P Chicken (C)
P Salad (S)
POCS
P O C S RB

CHO (g)

50.0
50.0
50.0
53.7
53.8
53.7

Fat (g)

3.6
33.6
10.4
3.7
40.5
44.6

Pro (g)

4.7
4.7
34.6
5.9
35.8
37.5

Meal GIa

108
108
108
103
103
96

Adjustment for
CHOb

Fatc

Prod

1.00
1.00
1.00
1.04
1.04
1.04

1.00
0.91
0.98
1.00
0.89
0.88

1.00
1.00
0.57
0.98
0.55
0.52

Overall Adje

Adjusted meal GIf

0.99
0.91
0.55
1.01
0.51
0.48

107
98
60
104
52
46

Abbreviations: AUC, area under the blood glucose response curve; CHO, available carbohydrate; GI, glycaemic index; Pro, protein; RB, rye bread. aGImeal is as
given by Hatonen et al.66 and presumably calculated via Equation (2). bAdjustment for CHO 1.49  (1  e  0.0222 g), where g grams of CHO.70 cAdjustment
for fat 1  ((0.29  (M  P))/100) where M meal fat and P potato meal fat (3.6 g). The factor 0.29 is the mean % reduction in AUC/g fat from Moghaddam
et al.67 and Lan-Pidhainy and Wolever.68 dAdjustment for protein 1  ((1.45  (M  P))/100) where M meal protein and P potato meal protein (4.7 g). The
factor 1.45 is the mean % reduction in AUC/g protein fromMoghaddam et al.67 and Lan-Pidhainy and Wolever.68 eOverall adjustment aC  aF  aP, where aC,
aF and aP represent the adjustments for CHO, fat and protein, respectively. fAdjusted meal GI meal GI  overall adjustment.

& 2013 Macmillan Publishers Limited

European Journal of Clinical Nutrition (2013) 1 10

Validity of glycaemic index


TMS Wolever

Figure 4. Relationship between calculated meal GI and the mean


incremental AUC elicited by the six different meals shown in Table 1
in healthy subjects (n 11).66 Meals contained 5054 g avCHO, 5
38 g protein and 445 g fat. Open symbols represent the meal GI
calculated using Equation (2) (not adjusted for differences in meal
composition); closed symbols represent the calculated meal GI
adjusted for the differences in the carbohydrate, fat and protein
content of the meals using the method shown in Table 1. The solid
line is the regression line (and 95% confidence interval) for the
closed symbols and the dot-dash line is the regression forced
through the origin (not significantly different from the regression
line). The correlation coefficient, r 0.912 (P 0.011).

and, thus, GI could be any value since any number multiplied by 0


equals 0. Ascribing a GI value of 0 to a food containing no
carbohydrate does not affect the results of diet GI calculations, but
it is an important misconception when used to argue that GI does
not predict the nutritional quality of a food. A prominent
nutritionist used pork rind as an example of a low-GI food
choice (GI 0) that may not improve the diet.51 While it may have
been a reductio ad absurdum, it really is an absurd example
because pork rind is not a food for which GI is a relevant
nutritional property. Of course GI does not predict the nutritional
quality of a foodit was never intended to do so! In 1981, we
indicated the purpose of GI was ... to supplement tables based
solely on chemical analysis.1 No single nutritional attribute of a
food can indicate its overall nutritional quality, but this does not
stop recommendations being made to limit fat, sodium and sugar
intakes.
GI METHODOLOGY
Is there a standard GI method?
In 2009, a scientist in the Bureau of Nutritional Sciences, Health
Canada, published a paper whose purpose was yto specically
and critically analyse the GI methodology.....3 The paper discusses
some of the methodological variations that inuence the results of
glycaemic response testing such as the different ways of
calculating AUC and different blood sampling schedules. His
conclusion that ...recommended GI methodology is not well
standardized and has several aws....3 demonstrates a remarkable
lack of care and precision and is without substance because the
author does not indicate what specic method is used to
determine GI. Standard GI methods have been described: an
FAO/WHO method for determining the GI of foods was published
in 1998;74 a far more detailed review of GI methodology than3 was
published by ILSI Europe in 2005;75 an Australian Standard was
gazetted in 200776 and the method for determining the GI of
foods was evaluated in international inter-laboratory studies in
2003 and 2008.77,78 The description of GI methodology is similar in
all these documents, some of which were cited, but not in the
context of providing an authoritative description of GI
methodology. Since then an ofcial International Standards
European Journal of Clinical Nutrition (2013) 1 10

Organization (ISO) method for determining the GI of foods has


been published.53
However, having an ofcial method does not mean that it is
used correctly. Common problems include feeding food portions
containing 50 g total (instead of available) carbohydrate, testing
the reference food only once in each subject, incorrect blood
sampling schedule, incorrect AUC calculation, having no10
subjects, not excluding outlying values, using an imprecise
method to measure glucose and having mean coefcient of
variation (CV) of within-individual variation of AUC 430%. Thus,
not every value that calls itself GI is a valid GI. This is a problem
about which I am very concerned and will have to be addressed
through a programme of laboratory accreditation; but it is not a
problem with the method, it reects a lack of knowledge
translation. Appropriate regulations about the use of GI could
help to reduce errors through education and enforced use of
correct methodology.
Is GI methodology accurate and precise?
Concerns about the accuracy and precision and GI are often
raised, but it is not clear how accurate or precise GI values ought
to be. In addition, some of the estimates of the precision of GI in
the literature are grossly incorrect. Four lines of evidence have
been used to suggest that GI values are inaccurate and
imprecise: (1) high variation of GI values in the international GI
tables;49 (2) the many factors in foods that affect GI;4951 (3) high
day-to-day variation of glycaemic responses within and between
subjects;48,50,51 and (4) high variation of GI values reported in an
interlaboratory study.4,50,51
Variation of GI values in the international GI tables. International
tables of GI values were published in 1995,79 200280 and 2008.81 It
has been pointed out that the GI values of the same food vary over
such a large range that it is not possible to say that the GI of a food
is predictable.49 It is not clear what point is being made, but one can
hardly argue that there is not a large amount of variation of the GI
values in those tablesthere is! The most variable food is potato
whose reported GI in the 2008 tables varies from 23 to 118. Such
variation is of concern because these tables are used to develop
databases of the GI of foods used in prospective studies. For
example, what GI value should be given to porridge oats when there
are 22 GI values in the 2002 tables ranging from 40 to 80 with a
mean of 58 and s.d. of 10? The mean or median of the existing
values would most likely be used, but this would only be correct if
the variation in published GI values was due to random error. The
magnitude of random variation in GI values in the GI tables can be
estimated from the results of two interlaboratory studies,77,78 both of
which suggest that the s.d. of GI values is about 9. By comparing the
s.d. of the GI values of 62 types of foods in the 2002 international
GI tables with an s.d. of 9 using the F ratio to estimate the
probability of heterogeneity,82 I estimated it was likely that the
variation of GI values in the international tables was greater than
random error for 45 of the 62 types of foods (73%). While some of
this variation may be due to methodological errors, the major
implication is that much of the variation of GI values in the
International GI tables represents real differences among foods
within the same category (for example, potato), which are not
related to their chemical composition as given on the nutrition label.
The many factors in foods that affect GI. The GI values of foods
may vary because of differences in variety, processing, growing
conditions and so forth, and it is difcult to use the GI tables to
determine the correct GI value of a specic food. This is a problem
because the GI values of foods in nutrient databases may be
inaccurate and, thus, the conclusions of prospective trials showing
that low diet GI is associated with reduced risk of chronic disease
may be unreliable. Indeed, a group in Europe has shown that
& 2013 Macmillan Publishers Limited

Validity of glycaemic index


TMS Wolever

5
calculated diet GI values may differ substantially depending on
who created the GI database, because different people ascribe
different GI values to the same food.83 However, evidence for the
clinical utility of GI comes not only from epidemiological studies,
but also from clinical trials when the investigators altered the GI of
the diets fed to subjects. In the latter cases, the probability of
misclassifying the GI values of foods is reduced, especially when
the investigators have been involved in GI testing and, therefore,
are familiar with the GI of their local foods.
The variability in GI due to cooking, processing and variety
presents some problems, but also is an opportunity to reduce diet
GI without major changes in the nature of the diet. The challenge
is how to provide reliable data on the GI of specic foods to
consumers and health professionals.
High day-to-day variation of glycaemic responses within and
between subjects
Several authors are concerned that GI values vary tremendously
among and within subjects48,50,51 due to, among other things, the
degree of mastication, the quantity of food ingested at any one
time, the time period of eating, the frequency of eating, the
accompanying foods and foods eaten at prior meals and on prior
days. However, most of these comments are about glycaemic
responses, not GI. Subjects do not have GI values, foods have GI
values that do not differ signicantly among subjects (Figure 1).
Thus, the variation in the estimated GI among individual subjects
reects day-to-day variation (which is substantial), not true
differences between subjects. When measuring GI, the quantity
of food ingested, the time period and frequency of eating and the
accompanying foods are all controlled for and, thus, do not affect
the resulting value.
Nevertheless, within-individual variation does inuence the
accuracy and precision of measured GI values, and for this reason
GI methodology has been designed to minimize these effects. We
showed in 1985 that within-individual variation of glycaemic
responses, expressed as CV ( 100  s.d./mean) was greater in
people without diabetes, 25%, than those with type 2 diabetes,
15%, but greatest of all in those with type 1 diabetes, 30%.84 Since
the GI is a ratio of two independently variable values, high withinindividual variation skews the mean of such ratios and likely
explains why GI values were somewhat higher in people with type
1 diabetes than those with type 2 diabetes.85 Mathematical
modelling suggested that reducing variation by using the mean
AUC of three tests of the reference food as the denominator in GI
calculations instead of only one reduced the mean and s.d. of the
resulting GI values;86 later, we showed this actually did occur.75
This is why the denominator in the GI calculation must be the
mean of X2 tests of the reference food in each subject.53,75
Further renements of the method to improve accuracy and
precision include the requirement that the mean CV of the
reference food should be o30%78 and that the mean of two
fasting blood samples be used when calculating the AUC.87
High variation of GI values reported in an interlaboratory study.
We reported the s.d. of GI values in n 68 subjects in an
interlaboratory study; for the 5 foods tested they varied from 25 to
38.77 An eminent food chemist used these s.d. values to calculate
the expected range of GI values as being the mean2.8 s.d.
(99.5% condence interval); thus, for potato, with a GI
(means.d.) of 8533, he concluded that the GI value could be
anywhere from  7 to 176.4 Others,51 including Health
Canada,88 used the same approach to support the conclusion
that the GI is too variable for practical use.
However, this is unequivocally wrong. The s.d. of 33 refers to the
variation of the estimates in the individual subjects, not the
variation of food GI values (which are the mean of the estimates in
at least 10 subjects). The s.d. of food GI values is about 9 not 33.
& 2013 Macmillan Publishers Limited

Figure 5. Incremental AUC elicited by portions of white bread


containing 25 or 50 g avCHO tested repeatedly by 23 normal
subjects. Each subject tested the 25-g avCHO portion 37 (median 5)
times and the 50-g avCHO portion 310 (median 6) times in the
morning after 1014 h overnight fasts. Data from Wolever et al.89

The food chemist boasted that chemical analysis of food was


much more precise and presumably more meaningful than GI
because The inability of the GI method to differentiate between
foods on eating occasions leads to the conclusion that the food
itself is a minor contributor to a given GI measurement, and
therefore that GI method does not measure a meaningful property
of a food..4 However, varying the amount of food consumed also
performs poorly in terms of predicting an individuals glycaemic
response. We measured the glycaemic responses elicited by 25
and 50 g carbohydrate portions of white bread in 23 subjects each
of whom tested the 25 g portion 37 times and the 50 g portion
310 times.89 There was high variation of the glycaemic responses
both between and within subjects (Figure 5). The means.d. of
the AUCs after the 25-g CHO loads was 5724% of the respective
subjects mean AUC after 50 g CHO. Applying the same criteria the
food chemist used to judge GI (mean2.8 s.d.), the response
elicited by 25 g CHO could be anywhere from  9 to 123% of
that after 50 g CHO, a range similar to that suggested for GI of
potato. Applying the same logic he used to criticize GI leads to the
rather ludicrous conclusion that: The inability of the method used
to measure the amount of food consumed to differentiate
between food portion sizes on individual eating occasions leads
to the conclusion that the amount of food itself is a minor
contributor to a given glycaemic response, and therefore food
portion size does not indicate a meaningful food property.
Of course this argument fails because it is based on the false
premise that it is meaningful to predict glycaemic responses on
individual eating occasions. Except for people with type 1 diabetes
who need to work out the amount of insulin to give prior to a
meal, being able to predict the glycaemic response on any one
eating occasion is unimportant for most people. Indeed, it is
very difcult to predict individual glycaemic responses because, as
we have seen, they vary considerably from day-to-day. There
is no food measure that can predict an individual response
any better than GI; to reject GI on the grounds of high day-to-day
variation is to reject any method of assessing the glycaemic
impact of foods.
Is GI methodology precise enough for clinical utility?
The most common way GI used clinically is to classify foods as
being low- (GIp55), medium- (GI 5669) or high-GI (GIX70)9092
with advice to use low-GI foods more often.93,94 Thus, a clinically
important issue is whether GI methodology can distinguish
between low-GI and high-GI foods. How certain is it that a lowGI food is not really high-GI? Specically, if a foods GI is p55, then
what is the probability that its true GI is o70? The probability is
equal to the proportion of the area under the normal curve having
mean X and s.d. 9 which is o70, where X is the GI of the food and
9 is the between-lab s.d. of GI values from previous interlaboratory
studies.77,78 If X 55, then the area 0.94 (probability of 94%); if
European Journal of Clinical Nutrition (2013) 1 10

Validity of glycaemic index


TMS Wolever

6
X 54, then the probability is 95%, if X 50, then the probability is
98% and if X 48, then there is 499% probability that the real GI
is o70. This suggests that GI methodology can distinguish
between low-GI and high-GI foods with reasonable certainty.
However, it should be noted that neither of the studies used to
estimate the s.d. of GI77,78 complied with the current ISO
method,53 which includes new procedures (nX10 subjects, 2
fasting blood samples and exclusion of outlying values) and
internal validity checks (analytical CVp3.6% and within-subject
CVo30%) intended to improve accuracy and precision. Thus, the
performance of the current GI method may be better than that of
the previous studies. Further studies are needed to assess the
performance of the current ISO method, and to determine if
changes can be made to further improve accuracy and precision.
IS THERE A BETTER INDICATOR OF CARBOHYDRATE QUALITY
THAN GI?
There is a wealth of evidence that the GI of dietary carbohydrate
has, or may have, a signicant impact on clinically relevant outcomes in health and disease.716,1847 However, it is often considered that the GI has limitations as an indicator of carbohydrate
quality compared with other measures such as dietary bre or
whole grains.
Limitations of GI
In a debate about the clinical utility of GI at the 2010 Experimental
Biology meeting,95 it was suggested that GI had limitations as an
indicator of carbohydrate quality because many high GI foods
such as whole grains and starchy vegetables are linked to positive
health outcomes, while many low-GI foods, such as sugar
sweetened beverages (SSB) and fructose, are linked to negative
health outcomes. Are these statements accurate?

median (49 g/d) and 95th percentile (87 g/day) of fructose intake
in the United States,100 suggesting that the harmful effects of
fructose are linked to overconsumption rather than a harmful
effect of fructose per se. Recent systematic reviews and metaanalyses conclude that iso-caloric substitution of fructose for other
carbohydrates does not cause weight gain99 or raise serum uric
acid,101 signicantly improves glycaemic control in diabetes102
and signicantly reduces mean arterial blood pressure.103 However, intakes greater than 60 g104 or 100 g105 per day may increase
serum triglycerides. I am not suggesting that fructose is healthy, or
ought to be used as a food ingredient to create low-GI foods.
However, foods naturally rich in fructose (for example, fruits) may
have health benets linked to their low-GI106 and moderate
intakes of added fructose p36 g/day improve glycaemic control
with no adverse cardiometabolic effects.107 Taken together it
seems, therefore, that either avoidance or overconsumption of
fructose may have negative health outcomes.
Is whole grain a better indicator of carbohydrate quality than GI?
The evidence for the health benets of whole grains comes from
epidemiological studies showing strong associations between
high intake of whole grains and reduced risk for cardiovascular
disease,108 diabetes109 and obesity.110,111 Recent large clinical
trials in subjects without diabetes showed that increased wholegrain consumption reduced systolic blood pressure,112 but none
showed any signicant effect of whole grains on blood glucose,
insulin sensitivity and/or insulin secretion, inammatory markers
or body weight.112114 One study found that whole grains reduced
LDL cholesterol, but another found the opposite (Figure 6). On the
other hand, trials of low-GI diets of similar magnitude in subjects
without diabetes showed signicant benecial effects of a low-GI
diet on blood lipids,115,116 inammatory markers,43 body weight30
and possibly insulin sensitivity.115 When the results of the four

Do whole-grain foods and starchy vegetables have a high GI?


In the 2002 GI tables,80 there are 102 values for wheat and rye
breads of which 35 are whole grain or whole meal. The mean GI
for the 35 whole-grain breads, 63.3, is equivalent to that of the 67
other breads, 63.1, and the distribution of low- (p55), medium
(56-69) and high-GI (X70) values for the whole-grain breads, 23,
46 and 31%, respectively, is similar to that for the other breads, 21,
34 and 45% (P 0.40). Similarly, for the 46 GI values for mashed,
sweet or whole potatoes, 20% are low-GI, 28% medium-GI and
52% high-GI. Finally, of the 20 GI values for soft drinks, sports
drinks and sucrose, 10% are low-GI, 60% medium-GI and 30%
high-GI. Thus, it is incorrect to assert that whole grains and root
vegetables have a high GI and SSB have a low-GI. In reality, the GI
of all of these foods may vary widely depending upon their
variety, composition and processing.
Does consumption of fructose, a low-GI sugar, have negative
health outcomes?
It is commonly held that fructose, a low-GI sugar, has negative
health outcomes. The evidence for this comes partly from early
prospective studies showing that the consumption of SSB is linked
to increased risk for weight gain in adolescents.96 However, a
meta-analysis of prospective trials showed no association between
body mass index and SSB consumption in children and
adolescents,97 and a recent systematic review and meta-analysis
concluded that the evidence from clinical trials did not
demonstrate conclusively a causal relationship between SSB
consumption and weight gain.98 The evidence that fructose
per se is harmful to humans comes from clinical trials of fructose
overfeeding in which 182 g/day (154201 g/day) (median
(interquartile range)) of fructose was fed, representing 37.5%
(31.343.8%) of energy.99 These amounts are way above the
European Journal of Clinical Nutrition (2013) 1 10

Figure 6. Forest plots of the effect of whole grains (left) and low-GI
diets (right) on body weight, systolic (SBP) and diastolic (DBP) blood
pressure, c-reactive protein (CRP) and LDL cholesterol in subjects
without diabetes. Values are standardized means with 95%
confidence intervals. The numbers of subjects on whole grain and
low-GI diets are indicated; numbers on the control diets are not
included, but are roughly similar. Data from references 43,112116
& 2013 Macmillan Publishers Limited

Validity of glycaemic index


TMS Wolever

Figure 7. Percentage of studies with statistically significant reductions in fasting glucose (FBG), HbA1c (A1c), urinary glucose excretion
(UGE) and serum cholesterol (Chol) upon treatment of diabetes with
purified viscous or non-viscous dietary fibres. Adapted from Wolever
et al.121 The numbers on top of the bars represent the number of
studies with a significant effect over the total number of studies.

Figure 8. Meta-analysis of the effect of dietary fibre on glycaemic


control in diabetes. Values are changes in HbA1c with 95% confidence intervals. Adapted from Post et al123 by separating studies
using soluble fibre (solid symbols) from those using insoluble fibre
(open symbols).

low-GI studies are combined, there is a signicant reduction in LDL


cholesterol and C-reactive protein and a strong trend towards
weight reduction (Figure 6).
Thus, whole grains and low-GI foods have benecial effects on
different disease risk factors and tend to complement each other.
Low-GI diets have at least as many, if not more, statistically
signicant effects than whole grain-enriched diets; this does not
support the hypothesis that whole grain is a better marker of
carbohydrate quality than GI.

I summarized the results of over 30 studies on the effect of various


puried bres on glucose and lipid control in diabetes.120 A
signicantly greater proportion of studies demonstrated signicant improvements with viscous than with non-viscous bres
(Figure 7). This is consistent with the results of more recent studies
showing that foods high in soluble bre reduce serum cholesterol
more than those high in insoluble bre122 and that the ability
of oat b-glucan to reduce cholesterol depends on its viscosity.121
A recent systematic review and meta-analysis concluded that high
bre diets improve glycaemic control in diabetes.123 Although the
authors did not differentiate between bre types, there is a clear
trend towards a larger impact of soluble bres (Figure 8). Thus,
rather than bre being a better marker of carbohydrate quality
than GI, it is the GI of the high bre food, or the ability of the bre
to reduce glycaemic responses, which predicts its health effects.
CONCLUSIONS
Recent criticisms of the GI focus on its lack of validity with
assertions that GI methodology is non-standard, awed, inaccurate and imprecise, that GI values do not represent a valid
property of foods, that GI does not predict what foods are healthy
and that whole grain and bre are better markers of carbohydrate
quality. However, some of these conclusions are based on gross
errors in understanding and interpretation and others are not
supported by current data or are inconsistent with other
nutritional recommendations. GI is a valid marker of carbohydrate
quality because: (1) GI methodology is accurate and precise
enough for practical use; (2) GI is a property of the food; and (3) GI
is biologically meaningful and inuences outcomes in health and
disease in ways that are relevant to almost everyone. It is illogical
for current dietary guidelines to recommend increased consumption of whole grains and dietary bre but not to mention GI
because the evidence that GI affects health outcomes is at least as
good as or better than that for whole grains and bre. GI is a novel
concept from a regulatory point of view and a number of
problems need to be addressed to successfully translate GI
knowledge into practice. These problems are not insurmountable
but no progress can be made until bias and misunderstanding
about GI is overcome and there is better agreement about what is
the actual state of knowledge on GI so that the real issues can be
identied and addressed.
CONFLICT OF INTEREST
I have an academic conict of interest, in that I am co-inventor of the GI concept and
want to see it accepted, however, neither my academic position nor salary depends
on the success of the concept. I receive remuneration as President, Medical Director
and Scientist of Glycemic Index Laboratories, Inc., a contract research organization; I
receive remuneration as President and part owner of Glycaemic Index Testing, Inc.,
which provides services to GI Labs. I receive royalties as co-author of a number of
popular books on GI under the general title of The Glucose Revolution, and consulting
fees from Tamasek Polytechnic, Singapore for advice related to GI research. In the last
3 years, I have received payment as a member of the scientic advisory board of
McCain Foods, Inc. Except for the preceding, I have no stocks or shares in any
company that may gain or lose nancially through publication (with the possible
exception of companies included in mutual funds about which I am unaware and
have no control over) and I have no ownership position in any patents or patent
applications whose value may be affected by publication.

ACKNOWLEDGEMENTS
Is dietary bre a better marker of carbohydrate quality than GI?
It has been recognized for over 30 years that different types of
bres have different health effects with viscous bres tending to
reduce postprandial glucose and serum cholesterol and nonviscous bres to bulk the stool. The ability of puried bres to
reduce postprandial glucose is directly related to viscosity
and the amount solubilized in the small intestine.117119 In 1992,
& 2013 Macmillan Publishers Limited

The authors research cited in this article was supported by grants to the author and
scholarship support for the authors graduate students from the Canadian Institutes
for Health Research (CIHR Operating Grant MOP-79382), the Natural Sciences and
Engineering Research Council of Canada (NSERC), Glycaemic Index Testing, Inc., and
Glycemic Laboratories, Inc., and by grants to Glycemic Index Laboratories, Inc., from
Agriculture and Agri-Food Canada, Ag West Bio, Inc., CreaNutrition AG, The Swedish
Governmental Agency for Innovations Systems, Sydney University GI Research
Services and ILSI Europe. Collaborators in the interlaboratory GI studies cited here

European Journal of Clinical Nutrition (2013) 1 10

Validity of glycaemic index


TMS Wolever

8
also received funding from the multiple sources indicated on the respective publications.

REFERENCES
1 Jenkins DJA, Wolever TMS, Taylor RH, Barker HM, Fielden H, Baldwin JM et al.
Glycemic index of foods: a physiological basis for carbohydrate exchange.
Am J Clin Nutr 1981; 34: 362366.
2 Coulston AM, Reaven GM. Much ado about (almost) nothing. Diabetes Care 1997;
20: 241243.
3 Aziz A. The glycemic index: methodological aspects related to the interpretation
of health effects and to regulatory labeling. J AOAC Int 2009; 92: 879887.
4 DeVries JW. Glycemic index: the analytical perspective. Cereal Foods World 2007;
52: 4549.
5 Wolever TMS. The glycemic index: ogging a dead horse? Diabetes Care 1997;
20: 452456.
6 Sykes JB (ed) The Concise Oxford Dictionary. 7th Edition (Clarendon Press: Oxford,
UK, p 1186, 1982.
7 Brand-Miller J, Hayne S, Petocz P, Colagiuri S. Low-glycemic index diets in the
management of diabetes: a meta-analysis of randomized controlled trials.
Diabetes Care 2003; 26: 22612267.
8 Jenkins DJA, Kendall CWC, McKeown-Eyssen G, Josse RG, Silverberg J, Booth GL
et al. Effect of a low-glycemic index or a high-cereal ber diet on type 2 diabetes.
A randomized trial. JAMA 2008; 300: 27422753.
9 Salmeron J, Ascherio A, Rimm EB, Colditz GA, Spiegelman D, Jenkins DJ et al.
Dietary ber, glycemic load, and risk of NIDDM in men. Diabetes Care 1997; 20:
545550.
10 Salmeron J, Manson JE, Stampfer MJ, Colditz GA, Wing AL, Willett WC. Dietary
ber, glycemic load, and risk of non-insulin-dependent diabetes mellitus in
women. JAMA 1997; 277: 472477.
11 Meyer KA, Kushi LH, Jacobs Jr DR, Slavin J, Sellers TA, Folsom AR. Carbohydrates,
dietary ber, and incident type 2 diabetes in older women. Am J Clin Nutr 2000;
71: 921930.
12 Schulze MB, Liu S, Rimm EB, Manson JE, Willett WC, Hu FB. Glycemic index,
glycemic load, and dietary ber intake and incidence of type 2 diabetes in
younger and middle-aged women. Am J Clin Nutr 2004; 80: 348356.
13 Hodge AM, English DR, ODea K, Giles GG. Glycemic index and dietary ber and
the risk of type 2 diabetes. Diabetes Care 2004; 27: 27012706.
14 Zhang C, Liu S, Solomon CG, Hu FB. Dietary ber intake, dietary glycemic load,
and the risk for gestational diabetes mellitus. Diabetes Care 2006; 29: 22232230.
15 Villegas R, Liu S, Gao Y-T, Yang G, Li H, Zheng W et al. Prospective study of
dietary carbohydrates, glycemic index, glycemic load, and incidence of
type 2 diabetes mellitus in middle-aged Chinese women. Arch Int Med 2007; 167:
23102316.
16 Krishnan S, Rosenberg L, Singer M, Hu FB, Djousse L, Cupples A et al. Glycemic
index, glycemic load, and cereal ber intake and risk of type 2 diabetes in US
black women. Arch Int Med 2007; 167: 23042309.
17 World Health Organization. Denition of health. URL https://apps.who.int/
aboutwho/en/denition.htmlaccessed 9 December 2011.
18 OReilly J, Wong SHS, Chen Y. Glycaemic index, glycaemic load, and exercise
performance. Sport Med 2010; 40: 2739.
19 Benton D, Rufn M-P, Lassel T, Nabb S, Messaoudi M, Vinoy S et al. The delivery
rate of dietary carbohydrates affects cognitive performance in both rats and
humans. Psychopharmacology (Berl) 2003; 166: 8690.
20 Micha R, Rogers PJ, Nelson M. The glycaemic potency of breakfast and cognitive
function in school children. Eur J Clin Nutr 2010; 64: 948957.
21 Smith RN, Mann NJ, Braue A, Makelainen H, Varigos GA. A low-glycemic-load diet
improves symptoms in acne vulgaris patients: a randomized controlled trial.
Am J Clin Nutr 2007; 86: 107115.
22 Berra B, Rizzo AM. Glycemic index, glycemic load, wellness and beauty: the state
of the art. Clin Dermatol 2009; 27: 230235.
23 Clapp III JF. Maternal carbohydrate intake and pregnancy outcome. Proc Nutr Soc
2002; 61: 4550.
24 Clapp III JF, Lopez B. Low- versus high-glycemic index diets in women: effects on
caloric requirements, substrate utilization, and insulin sensitivity. Metabol Synd
Rel Disord 2007; 5: 231242.
25 Moses RG, Luebcke M, Davis WS, Coleman KJ, Tapsell LC, Petocz P et al. Effect of
a low-glycemic-index diet during pregnancy on obstetric outcomes. Am J Clin
Nutr 2006; 84: 807812.
26 Rhodes ET, Pawlak DB, Takoudes TC, Ebbeling CB, Feldman HA, Lovesky MM et al.
Effects of a low-glycemic load diet in overweight and obese pregnant women: a
pilot randomized controlled trial. Am J Clin Nutr 2010; 92: 13061315.
27 Chavarro JE, Rich-Edwards JW, Rosner BA, Willet WC. A prospective study of
dietary carbohydrate quality in relation to risk of ovulatory infertility. Eur J Clin
Nutr 2009; 63: 7886.

European Journal of Clinical Nutrition (2013) 1 10

28 Marsh KA, Steinbeck KS, Atkinson FS, Petocz P, Brand-Miller JC. Effect of a low
glycemic index compared with a conventional healthy diet on polycystic ovary
syndrome. Am J Clin Nutr 2010; 92: 8392.
29 Chiu CJ, Liu S, Willett WC, Wolever TM, Brand-Miller JC, Barclay AW et al.
Informing food choices and health outcomes by use of the dietary glycemic
index. Nutr Rev 2011; 69: 231242.
30 Larsen TM, Dalskov SM, van Baak M, Jebb SA, Papadaki A, Pfeiffer AF et al. Diets
with high or low protein content and glycemic index for weight-loss maintenance. N Engl J Med 2010; 363: 21022113.
31 Barclay AW, Petocz P, McMillan-Price J, Flood VM, Prvan T, Mitchell P et al.
Glycemic load and chronic disease risk-a meta-analysis of observational studies.
Am J Clin Nutr 2008; 87: 627637.
32 Mursu J, Virtanen JK, Rissanen TH, Tuomainen T-P, Mykanen I, Laukkanen JA et al.
Glycemic index, glycemic load, and the risk of acute myocardial infarction in
Finnish men: the Kuopio Ishcaemic Heart Disease Risk Factor Study. Nutr Metab
Cardiovasc Dis 2011; 21: 144149.
33 Oh K, Hu FB, Cho E, Rexrode KM, Stampfer MJ, Manson JE et al. Carbohydrate
intake, glycemic index, glycemic load, and dietary ber in relation to risk of
stroke in women. Am J Epidemiol 2005; 161: 161169.
34 Oba S, Nagata C, Nakamura K, Fujii K, Kawachi T, Takatsuka N et al. Dietary
glycemic index, glycemic load, and intake of carbohydrate and rice in relation to
risk of mortality from stroke and its subtypes in Japanese men and women.
Metabolism 2010; 59: 15741582.
35 McKeown NM, Meigs JB, Liu S, Saltzman E, Wilson PWF, Jacques PF. Carbohydrate nutrition, insulin resistance, and the prevalence of the metabolic syndrome
in the Framingham Offspring Cohort. Diabetes Care 2004; 27: 538546.
36 Silva FM, Steemburgo T, de Mello VD, Tonding SF, Gross JL, Azevedo MJ. High
dietary glycemic index and low ber content are associated with metabolic
syndrome in patients with type 2 diabetes. J Am Coll Nutr 2011; 30: 141148.
37 Dong JY, Qin LQ. Dietary glycemic index, glycemic load, and risk of breast
cancer: meta-analysis of prospective cohort studies. Breast Cancer Res Treat 2011;
126: 287294.
38 Shikany JM, Flood AP, Kitahara CM, Hsing AW, Mayer TE, Willcox BJ et al. Dietary
carbohydrate, glycemic index, glycemic load, and risk of prostate cancer in the
Prosate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO) cohort.
Cancer Causes Cont 2011; 22: 9951002.
39 Radin RG, Palmer JR, Rodenberg L, Kumanyika SK, Wise LA. Dietary glycemic
index and load in relation to risk of uterine leiomyomata in the Black
WomensHealth Study. Am J Clin Nutr 2010; 91: 12811288.
40 Mwamburi DM, Liebson E, Folstein M, Bungay K, Tucker KL, Qiu WQ. Depression
and glycemic intake in the homebound elderly. J Affect Dis 2011; 132: 9498.
41 Murakami K, Myake Y, Sasaki S, Tanaka K, Fukushima W, Kiyohara C et al. Dietary
glycemic index is inversely associated with the risk of Parkinsons disease: a casecontrol study in Japan. Nutrition 2010; 26: 515521.
42 Gopinath B, Harris DC, Flood VM, Burlutsky G, Brand-Miller JC, Mitchell P. Carbohydrate nutrition is associated with the 5-year incidence of chronic kidney
disease. J Nutr 2011; 141: 433439.
43 Gogebakan O, Kohl A, Osterhoff MA, van Baak MA, Jebb SA, Papadaki A et al.
Effects of weight loss and long-term weight maintenance with diets varying in
protein content and glycemic index on cardiovascular risk factors. The diet,
obesity and genes (DiOGenes) study: a randomized controlled trial. Circulation
2011; 124: 28292838.
44 Tsai CJ, Leitzmann MF, Willett WC, Giovanucci EL. Dietary carbohydrates and
glycaemic load and the incidence of symptomatic gall stone disease in men. Gut
2005; 54: 823828.
45 Tsai CJ, Leitzmann MF, Willett WC, Giovanucci EL. Glycemic load, glycemic index,
and carbohydrate intake in relation to risk of cholecystectomy in women. Gastroenterology 2005; 129: 105112.
46 Yazdy MM, Liu S, Mitchell AA, Werler MM. Maternal dietary glycemic intake and
the risk of neural tube defects. Am J Epidemiol 2010; 171: 407414.
47 Buyken AE, Flood V, Empson M, Rochtchina B, Barclay AW, Brand-Miller J et al.
Carbohydrate nutrition and inammatory disease mortality in older adults. Am J
Clin Nutr 2010; 92: 634643.
48 Franz MJ. In defense of the American Diabetes Associations recommendations
on the glycemic index. Nutr Today 1999; 34: 7881.
49 Pi-Sunyer FX. Glycemic index and disease. Am J Clin Nutr 2002; 76: 290S298S.
50 Anderson GH. The glycemic index: clinical and public health signicance.
Carbohydrate News (publication of the Canadian Sugar Institute) 2008, URL
http://www.sugar.ca/english/healthprofessionals/carboIssue13.cfm (accessed 9
December 2011).
51 Jones JM. Glycemic index: the state of the science, part 1: the measure and its
variability. Nutr Today 2012; 47: 207213.
52 Slavin J. Basic qualities of carbohydrates and GI/GL concept. In: Oral session at
EB2010, Anaheim, CA; Controversy: Are we ready to adapt glycemic index as an
eating guide?.

& 2013 Macmillan Publishers Limited

Validity of glycaemic index


TMS Wolever

9
53 International Standards Organization. ISO 26642:2010. Food products - Determination of the glycaemic index (GI) and recommendation for food classication.
Published 2010-09-23. (available at,
http://www.iso.org/iso/catalogue_
detail.htm?csnumber=43633 accessed 4 February 2013.
54 Wolever TMS, Jenkins DJA, Vuksan V, Josse RG, Wong GS, Jenkins AL. Glycemic
index of foods in individual subjects. Diabetes Care 1990; 13: 126132.
55 Wolever TMS, Jenkins AL, Vuksan V, Campbell J. The glycaemic index values of
foods containing fructose are affected by metabolic differences between subjects. Eur J Clin Nutr 2009; 63: 11061114.
56 Lan-Pidhainy X, Wolever TMS. Are the glycemic and insulinemic index values of
carbohydrate foods similar in healthy control, hyperinsulinemic and type 2
diabetic patients? Eur J Clin Nutr 2011; 65: 727734.
57 Wolever TMS. Glycemic index: the measure and its variability: the true state of
the science. Nutr Today 2012; 47: 214221.
58 Collier G, McLean A, ODea K. Effect of co-ingestion of fat on the metabolic
responses to slowly and rapidly absorbed carbohydrates. Diabetologia 1984; 26: 5054.
59 Bornet FRJ, Costagliola D, Rizkalla SW, Blayo A, Fontvieille A-M, Haardt M-J et al.
Insulinemic and glycemic indexes of six starch-rich foods taken alone and in a
mixed meal by type 2 diabetics. Am J Clin Nutr 1987; 45: 588595.
60 Henry CJK, Lightowler HJ, Kendall FL, Storey M. The impact of the addition of
toppings/llings on the glycaemic response to commonly consumed carbohydrate foods. Eur J Clin Nutr 2006; 60: 763769.
61 Collier GR, Wolever TMS, Wong GS, Josse RG. Prediction of glycemic response to
mixed meals in non-insulin dependent diabetic subjects. Am J Clin Nutr 1986; 44:
349352.
62 Wolever TMS, Yang M, Zeng XY, Atkinson F, Brand-Miller JC. Food glycemic
index, as given in GI tables, is a signicant determinant of glycemic responses
elicited by composite breakfasts. Am J Clin Nutr 2006; 83: 13061312.
63 Wolever TMS, Jenkins DJA. The use of the glycemic index in predicting the blood
glucose response to mixed meals. Am J Clin Nutr 1986; 43: 167172.
64 Flint A, Mller BK, Raben A, Pedersen D, Tetens I, Holst JJ et al. The use of
glycaemic index tables to predict glycaemic index of composite breakfast meals.
Br J Nutr 2004; 91: 979989.
65 Dodd H, Williams S, Brown R, Venn B. Calculating meal glycemic index by using
measured and published food values compared with directly measured meal
glycemic index. Am J Clin Nutr 2011; 94: 992996.
66 Hatonen KA, Virtamo J, Eriksson JG, Sinkko HK, Sundvall JE, Valsta LM. Protein
and fat modify the glycaemic and insulinaemic responses to a mashed potatobased meal. Br J Nutr 2011; 106: 248253.
67 Moghaddam E, Vogt JA, Wolever TMS. The effects of fat and protein on glycemic
responses in non-diabetic humans vary with waist-circumference, fasting plasma
insulin and dietary ber intake. J Nutr 2006; 136: 25062511.
68 Lan-Pidhainy X, Wolever TMS. The hypoglycemic effect of fat and protein is not
attenuated by insulin resistance. Am J Clin Nutr 2010; 91: 98105.
69 Wolever TMS, Bhaskaran K. Use of glycemic index to estimate mixed-meal glycemic response. Am J Clin Nutr 2012; 95: 256257.
70 Wolever TMS. The Glycaemic Index: A Physiological Classication of Dietary Carbohydrate. CABI Publishing: Wallingford, UK, pp 66, 2006.
71 Flood A, Subar AF, Hull SG, Zimmerman TP, Jenkins DJ, Schatzkin A. Methodology for adding glycemic load values to the National Cancer Institute Diet
History Questionnaire database. J Am Diet Assoc 2006; 106: 393402.
72 Simila ME, Valsta LM, Virtanen JM, Hatonen KA, Virtamo J. Glyaemic index
database for the epidemiological Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study. BrJ Nutr 2009; 101: 14001405.
73 Louie JC, Flood V, Turner N, Everingham C, Gwynn J. Methodology for adding
glycemic index values to 24-hour recalls. Nutrition 2011; 27: 5964.
74 Food and Agriculture Organization of the United Nations. FAO Food and Nutrition Paper 66. Carbohydrates in Human Nutrition. FAO: Rome, 1998.
75 Brouns F, Bjorck I, Frayn KN, Gibbs AL, Lang V, Slama G et al. Glycaemic index
methodology. Nutr Res Rev 2005; 18: 145171.
76 Australian Standards Organization. Glycemic Index of Foods AS4694-2007 http://
infostore.saiglobal.com/store/Details.aspx?DocN=AS0733779662AT
accessed
2 December 2009.
77 Wolever TMS, Vorster HH, Bjork I, Brand-Miller J, Brighenti F, Mann JI et al.
Determination of the glycaemic index of foods: interlaboratory study. Eur J Clin
Nutr 2003; 57: 475482.
78 Wolever TMS, Brand-Miller JC, Abernethy J, Astrup A, Atkinson F, Axelsen M et al.
Measuring the glycemic index of foods: interlaboratory study. Am J Clin Nutr
2008; 87(suppl): S247S257.
79 Foster-Powell K, Brand Miller J. International tables of glycemic index. Am J Clin
Nutr 1995; 62: S871S893.
80 Foster-Powell K, Holt SH, Brand-Miller JC. International table of glycemic index
and glycemic load values: 2002. Am J Clin Nutr 2002; 76: 556.
81 Atkinson FS, Foster-Powell K, Brand-Miller JC. International table of glycemic
index and glycemic load values: 2008. Diabetes Care 2008; 31: 22812283.

& 2013 Macmillan Publishers Limited

82 Wolever TMS. The Glycaemic Index: A Physiological Classication of Dietary Carbohydrate. CABI Publishing: Wallingford, UK, pp 120123, 2006.
83 van Bakel MM, Slimani N, Feskens EJ, Du H, Beulens JW, van der Schouw YT et al.
Methodological challenges in the application of the glycemic index in epidemiological studies using data from the European Prospective Investigation into
Cancer and Nutrition. J Nutr 2009; 139: 568575.
84 Wolever TMS, Nuttall FQ, Lee R, Wong GS, Josse RG, Csima A et al. Prediction of
the relative blood glucose response of mixed meals using the white bread
glycemic index. Diabetes Care 1985; 8: 418428.
85 Wolever TMS, Jenkins DJA, Josse RG, Wong GS, Lee R. The glycemic index:
similarity of values derived in insulin-dependent and non-insulin-dependent
diabetic patients. J Am Col Nutr 1987; 6: 295305.
86 Wolever TMS, Jenkins DJA, Jenkins AL, Josse RG. The glycemic index: methodology and clinical implications. Am J Clin Nutr 1991; 54: 846854.
87 Wolever TMS, Ip B, Moghaddam E. Measuring glycaemic responses: duplicate
fasting samples or duplicate measures of one fasting sample? Br J Nutr 2006; 96:
799802.
88 Aziz A. Health claims related to postprandial glycaemic control: Health Canadas
approach. Presented on 23 August, 2012 to the Program of Food Safety,
Nutrition and Regulatory Affairs, University of Toronto.
89 Wolever TMS, Leung J, Vuksan V, Jenkins AL. Day-to-day variation in glycemic
response elicited by white bread is not related to variation in satiety in humans.
Appetite 2009; 52: 654658.
90 Clips on Sugars For Up-to-Date Information on Sugars in Healthy Eating:
Understanding the Glycemic Index. Canadian Sugar Institute: Educators and
Students site http://www.sugar.ca/english/pdf/healthprofessionals/Clip9.pdf.
Updated 2008. Accessed 9 March 2012.
91 The Glycemic Index. Canadian Diabetes Association: For Professionals
Resources http://www.diabetes.ca/les/GlycemicIndex_09_Boehringer.pdf. Updated
September 2009. Accessed 9 March 2012.
92 Glycemic Index Foundation. Using the GI http://www.gisymbol.com.au/using.
php. Accessed 9 March 2012.
93 The National Diabetes Services Scheme (NDSS) administered by Diabetes
Australia. Talking diabetes: the glycemic index. Diabetes Australia: Diabetes
Information Sheets site http://www.australiandiabetescouncil.com/Resources/
PDFs/NDSS-Information-Sheets/GLYCEMIC-INDEX-2010.aspx. Updated August
2010 Accessed 9 March 2012.
94 Govindji A. Food Facts: Glycaemic Index is it just hype? The British Dietetic
Association: Food Facts Sheets site http://www.bda.uk.com/foodfacts/GIDiet.pdf.
Published April 2004. Accessed 9 March 2012.
95 Slavin J. Basic qualities of carbohydrates and GI/GL concept. EB2010 Symposium:
Are We Ready to Adopt Glycemic Index as an Eating Guide. Monday 26)Anaheim,
CA, 2010.
96 Ludwig DS, Peterson KE, Gortmaker SL. Relation between consumption of sugarsweetened drinks and childhood obesity: a prospective, observational analysis.
Lancet 2001; 357: 505508.
97 Forshee RA, Anderson PA, Storey ML. Sugar-sweetened beverages and body
mass index in children and adolescents: a meta-analysis. Am J Clin Nutr 2008; 87:
16621671.
98 Mattes RD, Shikany JM, Kaiser KA, Allison DB. Nutritively sweetened beverage
consumption and body weight: a systematic review and meta-analysis of
randomized experiments. Obes Rev 2011; 12: 346365.
99 Sievenpiper JL, de Souza RJ, Mirrahimi A, Yu ME, Carleton AJ, Beyene J et al.
Effect of fructose on body weight in controlled feeding trials: a systematic review
and meta-analysis. Ann Intern Med 2012; 156: 291304.
100 Marriott BP, Cole N, Lee E. National estimates of dietary fructose intake increased
from 19772004 in the United States. J Nutr 2009; 139: 1228S1235S.
101 Wang DD, Sievenpiper JL, de Souza RJ, Chiavaroli L, Ha V, Cozma AI et al. The
effects of fructose intake on serum uric acid vary among controlled dietary trials.
J Nutr 2012; 142: 916923.
102 Cozma AI, Sievenpiper JL, de Souza RJ, Chiavaroli L, Ha V, Wang DD et al. Effect
of fructose on glycemic control in diabetes: a systematic review and metaanalysis of controlled feeding trials. Diabetes Care 2012; 35: 16111620.
103 Ha V, Sievenpiper JL, de Souza RJ, Chiavaroli L, Wang DD, Cozma AI et al. Effect
of fructose on blood pressure: a systematic review and meta-analysis of controlled feeding trials. Hypertension 2012; 59: 787795.
104 Sievenpiper JL, Carleton AJ, Chatha S, Jiang HY, de Souza RJ, Beyene J et al.
Heterogeneous effects of fructose on blood lipids in individuals with type 2
diabetes: systematic review and metaanalysis of experimental trials in humans.
Diabetes Care 2009; 32: 19301937.
105 Livesey G, Taylor R. Fructose consumption and consequences for glycation,
plasma triacylglycerol, and body weight: meta-analyses and meta-regression
models of intervention studies. Am J Clin Nutr 2008; 88: 14191437.
106 Jenkins DJ, Srichaikul K, Kendall CW, Sievenpiper JL, Abdulnour S, Mirrahimi A et al.
The relation of low glycaemic index fruit consumption to glycaemic control and

European Journal of Clinical Nutrition (2013) 1 10

Validity of glycaemic index


TMS Wolever

10

107

108
109

110

111

112

113

114

risk factors for coronary heart disease in type 2 diabetes. Diabetologia 2011; 54:
271279.
Sievenpiper JL, Chiavaroli L, de Souza R, Mirrahimi A, Cozma AI, Ha V et al.
Catalytic doses of fructose may benet glycemic control without harming
cardiometabolic risk factors: a small meta-analysis of randomized controlled
feeding trials. Br J Nutr 2012; 108: 418423.
Mellen PB, Walsh TF, Herrington DM. Whole grain intake and cardiovascular
disease: a meta-analysis. Nutr Metab Cardiovasc Dis 2008; 18: 283290.
de Munter JS, Hu FB, Spiegelman D, Franz M, van Dam RM. Whole grain, bran
and germ intake and risk of type 2 diabetes: a prospective cohort study and
systematic review. PLoS Med 2007; 4: e261.
van de Vijver LPL, van den Bosch LMC, van den Brandt PA, Goldbohm RA. Wholegrain consumption, dietary bre intake and body mass index in the Netherlands
cohort study. Eur J Clin Nutr 2009; 63: 3138.
ONeil CE, Zanovec M, Cho SS, Nicklas TA. Whole grain and ber consumption
are associated with lower body weight measures in US adults National Health
and Nutrition Examination Survey 19992004. Nutr Res 2010; 30: 815822.
Tighe P, Duthie G, Vaughan N, Brittenden J, Simpson WG, Duthie S et al. Effect of
increased consumption of whole-grain foods on blood pressure and other
cardiovascular risk markers in healthy middle-aged persons: a randomized
controlled trial. Am J Clin Nutr 2010; 92: 733740.
Brownlee IA, Moore C, Chateld M, Richardson DP, Ashby P, Kuznesof SA et al.
Markers of cardiovascular risk are not changed by increased whole-grain intake:
the WHOLEheart study, a randomised, controlled dietary intervention. Br J Nutr
2010; 104: 125134.
Kristensen M, Toubro S, Jensen MG, Ross AB, Riboldi G, Petronio M et al. Whole
grain compared with rened wheat decreases the percentage of body fat following a 12-week energy-restricted dietary intervention in postmenopausal
women. J Nutr 2012; 142: 710716.

European Journal of Clinical Nutrition (2013) 1 10

115 Jebb SA, Lovegrove JA, Grifn BA, Frost GS, Moore CS, Chateld MD et al. Effect
of changing the amount and type of fat and carbohydrate on insulin sensitivity
and cardiovascular risk: the RISCK (Reading, Imperial, Surrey, Cambridge, and
Kings) trial. Am J Clin Nutr 2010; 92: 748758.
116 McMillan-Price J, Petocz P, Atkinson F, ONeill K, Samman S, Steinbeck K et al.
Comparison of 4 diets of varying glycemic load on weight loss and cardiovascular risk reduction in overweight and obese young adults. Arch Intern Med 2006;
166: 14661475.
117 Jenkins DJA, Wolever TMS, Leeds AR, Gassull MA, Dilawari JB, Goff DV et al.
Dietary bres, bre analogues and glucose tolerance: importance of viscosity.
Br Med J 1978; 1: 13921394.
118 Tosh SM, Brummer Y, Wolever TMS, Wood PJ. Glycemic response to oat bran
mufns treated to vary molecular weight of b-glucan. Cereal Chem 2008; 85:
211217.
119 Lan-Pidhainy X, Brummer Y, Tosh SM, Wolever TM, Wood PJ. Reducing betaglucan solubility in oat bran mufns by freeze-thaw treatment attenuates its
hypoglycemic effect. Cereal Chem 2007; 84: 512517.
120 Wolever TMS, Jenkins DJA. Effect of dietary ber and foods on carbohydrate
metabolism. In: CRC Handbook of Dietary Fiber in Human Nutrition. 2nd edn, (Ed.).
Spiller GA) (CRC Press, Inc: Boca Raton, FL, USA, pp 111152, 1992.
121 Wolever TMS, Tosh SM, Gibbs AL, Brand-Miller J, Duncan AM, Hart V et al.
Physicochemical properties of oat b-glucan inuence its ability to reduce serum
LDL cholesterol in humans: a randomized clinical trial. Am J Clin Nutr 2010; 92:
723732.
122 Jenkins DJA, Wolever TMS, Rao AV, Hegele RA, Mitchell S, Ransom T et al. Effect
on serum lipids of very high ber intakes in diets low in saturated fat and
cholesterol. N Engl J Med 1993; 329: 2126.
123 Post RE, Mainous III AG, King DE, Simpson KN. Dietary ber for the treatment of
type 2 diabetes mellitus: a meta-analysis. J Am Board Fam Med 2012; 25: 1623.

& 2013 Macmillan Publishers Limited