Glycaemic index of selected staples commonly eaten in the Caribbean and the effects of boiling v. crushing

Abstract

Integrating information about the glycaemic index (GI) of foods into the Caribbean diet is limited by the lack of data. Therefore, we determined the GI of eight staple foods eaten in the Caribbean and the effect on GI of crushing selected tubers. Groups of eight to ten healthy volunteers participated in three studies at two sites. GI was determined using a standard method with white bread and adjusted relative to glucose. The mean area under the glucose response curve elicited by white bread was similar for the different groups of subjects. In study 1, the GI of cassava (Manihot esculenta; 94 (sem 11)) was significantly higher than those of breadfruit (Artocarpus altilis; 60 (sem 9)), cooking 'green' banana (Musa spp.; 65 (sem 11)) and sadha roti (65 (sem 9)) (P=0.018). There was no significant difference in the GI of the foods in study 2: dasheen (Colocasia esculenta var. esculenta; 77 (sem 10)), eddoes (Colocasia esculenta var. antiquorum; 61 (sem 10)), Irish potato (Solanum tuberosum; 71 (sem 8)), tannia (Xanthosoma sagittifolium; 60 (sem 5)) and white yam (Dioscorea alata; 62 (sem 6)), and, in study 3, crushing did not significantly affect the GI of dasheen, tannia or Irish potato. However, when the results from studies 2 and 3 were pooled, the GI of dasheen (76 (sem 7)) was significantly greater than that of tannia (55 (sem 5); P=0.015) with potato being intermediate (69 (sem 6)). We conclude that dasheen and cassava are high-GI foods, whereas the other tubers studied and sadha roti are intermediate-GI foods. Given the regular usage of cassava and dasheen in Caribbean diets we speculate that these diets would tend to be high GI, although this could be reduced by foods such as sadha roti and white yam. The range of GI between the staples is sufficiently large that health benefits may be accrued by replacing high-GI staples with intermediate-GI staples in the Caribbean diet.

Full text

Glycaemic index of selected staples commonly eaten in the
Caribbean and the effects of boiling v. crushing
D. Dan Ramdath
1
, Rene
´e L. C. Isaacs
2
, Surujpal Teelucksingh
3
and Thomas M. S. Wolever
2
*
1
Department of Preclinical Sciences, Faculty of Medical Sciences, The University of the West Indies, St Augustine,
Trinidad and Tobago
2
Department of Nutritional Sciences, University of Toronto, Faculty of Medicine, Toronto, Ontario, M5A 3E2,
Canada
3
Clinical Medical Sciences, Faculty of Medical Sciences, The University of the West Indies, St Augustine,
Trinidad and Tobago
(Received 22 July 2003 – Revised 22 January 2004 – Accepted 11 February 2004)
Integrating information about the glycaemic index (GI) of foods into the Caribbean diet is limited by the lack of data. Therefore, we deter-
mined the GI of eight staple foods eaten in the Caribbean and the effect on GI of crushing selected tubers. Groups of eight to ten healthy
volunteers participated in three studies at two sites. GI was determined using a standard method with white bread and adjusted relative to
glucose. The mean area under the glucose response curve elicited by white bread was similar for the different groups of subjects. In study
1, the GI of cassava (Manihot esculenta;94(
SEM 11)) was significantly higher than those of breadfruit (Artocarpus altilis;60(SEM 9)),
cooking ‘green’ banana (Musa spp.; 65 (SEM 11)) and sadha roti (65 (SEM 9)) (P¼0·018). There was no significant difference in the GI of
the foods in study 2: dasheen (Colocasia esculenta var. esculenta; 77 (SEM 10)), eddoes (Colocasia esculenta var. antiquorum; 61 (SEM
10)), Irish potato (Solanum tuberosum;71(SEM 8)), tannia (Xanthosoma sagittifolium;60(SEM 5)) and white yam (Dioscorea alata;62
(SEM 6)), and, in study 3, crushing did not significantly affect the GI of dasheen, tannia or Irish potato. However, when the results from
studies 2 and 3 were pooled, the GI of dasheen (76 (SEM 7)) was significantly greater than that of tannia (55 (SEM 5); P¼0·015) with potato
being intermediate (69 (SEM 6)). We conclude that dasheen and cassava are high-GI foods, whereas the other tubers studied and sadha roti
are intermediate-GI foods. Given the regular usage of cassava and dasheen in Caribbean diets we speculate that these diets would tend to
be high GI, although this could be reduced by foods such as sadha roti and white yam. The range of GI between the staples is sufficiently
large that health benefits may be accrued by replacing high-GI staples with intermediate-GI staples in the Caribbean diet.
Glycaemic index: Caribbean: Diabetes: Glucose
The glycaemic index (GI) is a classification of the glu-
cose-rising potential of carbohydrate foods relative to glu-
cose (Wolever et al. 1991). Recent studies have shown
that the regular consumption of diets containing high-GI
foods is associated with an increased risk for type 2 dia-
betes mellitus (Salmeron et al. 1997a,b) and CHD (Ford
& Liu, 2001; Liu & Manson, 2001). In contrast, the
inclusion of low-GI foods in the diet, with no change
in the total amount of carbohydrate consumed, may
improve blood glucose control (Brand et al. 1992),
reduce serum triacylglycerols (Jenkins et al. 1987), pro-
long endurance during physical activity (Thomas et al.
1991) and improve insulin sensitivity (Byrnes et al.
1995; Frost et al. 1998). The GI may therefore provide
the rationale for choosing carbohydrate foods for meal
plans created for individuals with diabetes (Food and
Agriculture Organization of the United Nations, 1998).
In addition, low-GI diets could be incorporated into
health promotion messages for the reduction of risk for
type 2 diabetes and CHD.
In the Caribbean, type 2 diabetes and CHD are the lead-
ing causes of morbidity and mortality. In Trinidad and
Tobago the prevalence of diabetes is the highest in the
Caribbean and approximately six times higher than that in
North America (Gulliford et al. 1995; Gulliford, 1996).
The annual financial cost of admission for diabetes has
been conservatively estimated at TT$ 10·66 million (UK
£1·24 million) (Gulliford et al. 1995). Predominantly a
migrant population from Africa and South Asia, the Carib-
bean has in turn seeded a secondary wave of migrants to
many metropolitan areas of North America and Europe.
In developed countries, migrant Caribbean populations con-
tinue to have a high incidence of diabetes wherever they
have settled (Cruickshank et al. 1991; Cooper et al. 1997).
*Corresponding author: Professor Thomas Wolever, fax þ1 416 978 5882, email thomas.wolever@utoronto.ca
Abbreviations: AUC, area under the glucose response curve; GI, glycaemic index; RS, resistant starch.
British Journal of Nutrition (2004), 91, 971–977 DOI: 10.1079/BJN20041125
qThe Authors 2004

Caribbean guidelines for the dietary management of
diabetes and CHD risk reduction are based on six food
groups, and the main dietary staples are ‘ground pro-
visions’ and ‘grains and cereals’ (Caribbean Food and
Nutrition Institute, 1994). Foods in the former group
include tubers, breadfruit (Artocarpus altilis) and cooking
‘green’ banana (Musa spp.), which is a different variety
from dessert banana and is eaten in the immature stage.
These provisions are usually cooked by boiling, but may
also be crushed after being boiled. Another important
source of carbohydrate is sadha roti; a wheat-based lea-
vened bread, which is commonly eaten in Caribbean
islands with a large South Asian population. The GI of
these foods are not known and as such their inclusion in
the menus for individuals with diabetes has been based
mainly on reducing fat intakes and increasing intakes of
complex carbohydrates (Caribbean Food and Nutrition
Institute, 1994). An understanding of the physiological
basis of carbohydrate exchange, as classified by the GI,
may assist in optimising healthy food choices in the Carib-
bean and in emigrant populations. As such, knowledge of
the GI of staples commonly used by Caribbean individuals
could contribute to evidence-based meal planning and
implementation of culturally sensitive risk-reduction diet-
ary interventions.
In the present two-centre study the GI for boiled tubers,
breadfruit, cooking banana, and sadha roti were determined
in healthy volunteers. In another study the effect of crush-
ing on the GI of selected tubers was also evaluated, since
preparation methods and differing starch characteristics
can both affect GI (Soh & Brand-Miller, 1999).
Methods
Study design
Parallel studies were conducted at the Department of Pre-
clinical Sciences, University of the West Indies, Trinidad
and Tobago (site 1) and at the Department of Nutritional
Sciences, University of Toronto, Canada (site 2) using a
standard protocol as outlined by Wolever et al. (1991).
The respective institutional ethics review committee
approved the protocols and informed consent was obtained
from all volunteers. The power of the tests with ten sub-
jects was set at 80 % to detect a 22 % difference in areas
under the glucose response curves (AUC) between the
foods. This calculation assumed a variation of 22 %
within subjects (Wolever et al. 1991).
Reference food
White bread was prepared at both study sites using the
recipe previously described (Wolever et al. 1994). In the
absence of a bread maker at site 1 the ingredients were
manually mixed and kneaded over a 3 h period, and
baked in an electric oven at 1908C for 35 min. At both
study sites loaves were cooled at room temperature for
1 h, weighed, cut into 50 g available carbohydrate portions
(discarding the crust ends), placed in plastic bags and
frozen. Before consumption the bread was warmed for
1 min in a microwave oven.
Test foods and preparation
These included breadfruit, cassava (Manihot esculenta),
dasheen (Colocasia esculenta var. esculenta), eddoes
(Colocasia esculenta var. antiquorum), cooking ‘green’
banana, tannia (Xanthosoma sagittifolium), Irish potato
(Solanum tuberosum), and white yam (Dioscorea alata).
Additionally, the leavened wheat-based bread, sadha roti,
was tested. Foods eaten at site 1 were obtained on the
day of harvest. With the exception of Irish potatoes, the
foods eaten at site 2 were obtained from the local immi-
grant market and were usually about 1 week post-harvest.
In all cases the inedible portions of the tubers were
removed and discarded; for cassava and breadfruit this
included both the peel and core. For the remaining tubers
and cooking ‘green’ banana the peel was removed. The
edible portions were washed and allowed to air dry at
room temperature for 10 min; they were then cut into
chunks of approximately 25 mm, and 50 g available carbo-
hydrate portions (see Table 1) were boiled with minimal
water and a pinch of salt. With the exception of breadfruit
(18 and 8 min) and green banana (15 and 8min) the tubers
were cooked by gently boiling with the lid of the cooking
vessel on for 20 min, followed by simmering heat and the
lid off for a further 10 min. At site 1 the test foods
(except sadha roti) were prepared in weekly batches,
allowed to cool at room temperature (about 258C) for
10 min and stored in plastic bags at 2208C. These were
thawed at room temperature and warmed for 1 min in a
Table 1. Composition of test foods (per 50 g available carbohydrate)*
Food Weight (g) Protein (g) Fat (g) Total carbohydrate (g) Fibre (g)
Dasheen† 131 0·4 0·3 52·1 2·2
Eddoes† 221 4·9 0·7 53·4 3·4
Irish potato† 151 1·9 0·2 53·8 3·8
Tannia† 135 2·6 0·3 52·0 2·0
White yam† 300 5·8 0·2 52·1 2·1
White flour‡ 67 6·9 0·7 51·8 1·8
Cassava‡ 159 4·9 0·6 54·0 4·0
Green banana‡ 223 2·7 1·0 54·9 4·9
Breadfruit‡ 216 2·4 0·4 53·2 3·2
* Available carbohydrate was calculated by difference.
† Proximate and dietary analysis using standard AOAC methods.
‡ Taken from Food Composition for the Caribbean (Caribbean Food and Nutrition Institute, 1998).
D. D. Ramdath et al.972

microwave oven before consumption. At site 2 the test
tubers were prepared on the day before each test, allowed
to cool (228C) for 10 min and refrigerated; they were
heated in a microwave oven for 1 min before being
served. At site 1 sadha roti was prepared on the test day
from 67·3 g portions of white flour (Five Roses; ADM
Agri-Industries Ltd, Decatur, IL, USA). The flour was
dry mixed with 1 g salt and 1 g baking powder. Following
the addition of approximately 50 ml water the mixture was
kneaded to produce a soft dough. A ball of dough was then
flattened with a rolling pin to approximately 150 mm in cir-
cumference and 12·5 mm thick. It was then evenly cooked
on a hot plate at moderate heat for 5 min, turning occasion-
ally until done.
Study 1
At site 1, eight normal volunteers (age 25·4 (SEM 1·5)
years; four male, four female; BMI 21·5 (SEM 0·8) kg/m
2
)
ate cassava, breadfruit, cooking ‘green’ banana and sadha
roti. The portion sizes were based on published values
for total carbohydrate and dietary fibre (Caribbean Food
and Nutrition Institute, 1998) and available carbohydrate
was calculated by difference (Table 1).
Study 2
At site 2, ten normal volunteers (age 29 (SEM 4) years; four
male, six female; BMI 21·9 (SEM 0·8) kg/m
2
) ate dasheen,
eddoes, white yam, tannia and Irish potato. The portion
sizes were based on proximate (ash, acid hydrolysis for
fat, 1008C oven for moisture, Dumas method (N £5·70)
for protein, total carbohydrate by difference) and total diet-
ary fibre analysis (gravimetric method) using standard
AOAC methods (Association of Official Analytical Che-
mists, 1995) with available carbohydrate being defined as
total carbohydrate minus total dietary fibre (Table 1).
Study 3
At site 2 the tubers from study 2 with the highest and
lowest GI values were chosen. Irish potato was used as a
comparison or reference food in order to validate the pro-
cedure used to study the effects of crushing on the selected
tubers. Six of the volunteers from study 2 participated in
study 3. Available carbohydrate portions (50 g) of the
foods were boiled and either left uncrushed or crushed
before storage in the refrigerator, as in study 2.
Experimental procedure
Following 12 h overnight fasts volunteers ate 50 g available
carbohydrate portions of the foods weekly. They each
started and ended with white bread; the second white
bread and the test foods were randomly eaten. At site 1
all foods were taken with 250 ml water, at site 2 the volun-
teers chose water or unsweetened tea or coffee, but the
chosen beverage was constant for all studies. Fingerpick
capillary blood samples were taken before starting the
meals. Volunteers were then asked to consume the foods
within 10 min and to remain seated for the duration of
the test. Further blood samples were taken at 15, 30, 45,
60, 90, and 120 min. Blood samples were taken into
fluoro-oxalate tubes and either stored at 2208C (site 2)
or quickly centrifuged to obtain plasma which was stored
at 2208C (site 1). At site 1 capillary plasma glucose was
measured using a commercial hexokinase end-point
method with blank correction (Sigma Chemical Co., St
Louis, MO, USA). Capillary whole-blood glucose was
measured at site 2 using an automatic analyser (2300 Stat
Glucose Analyser; YSI Inc., Yellow Springs, OH, USA).
All glucose assays were performed within 48 h of
sampling. At both sites the inter-assay and intra-assay
CV for the glucose assay were approximately 3 %.
Statistical analysis
Incremental AUC ignoring area beneath the fasting level
were calculated geometrically (Wolever et al. 1991). The
AUC for each food was expressed as a percentage of the
mean area for the three white-bread tests and the resulting
values averaged to give the food GI based on white bread.
Because it is now the preferred practice to report GI values
using glucose, this was derived by multiplying the white-
bread-based GI by 0·71 (since the GI of white bread is
71 when glucose has the reference GI of 100; Wolever
et al. 2003). The test-food GI for each subject was aver-
aged to give the mean GI for each test food. The AUC
for pooled repeated white-bread tests were compared
using one-way ANOVA. The AUC and GI values were
not normally distributed and so statistical analysis was per-
formed by using Friedman’s ANOVA by ranks with Dun-
nett’s post hoc test. The blood glucose concentrations at
each time point were compared using conventional two-
way ANOVA, followed by Tukey’s test to protect for mul-
tiple comparisons. Differences were considered statistically
significant at P,0·05.
Results
The blood glucose response curves of the test foods and
white bread for studies 1 and 2 are shown in Fig. 1. The
mean, standard deviation and CV of the AUC values for
repeated white-bread tests are given in Table 2 for each
volunteer and each site. There was no significant difference
between the mean AUC (P¼0·5) or the mean CV
(P¼0·09) between the different centres. However, differ-
ences between subjects at each site were highly significant
(P,0·001).
Table 3 shows the AUC, the white-bread-based GI
values, as well as the glucose-based GI results for the
test foods. Two-way ANOVA showed there was no signifi-
cant effect of subject on GI in any of the three studies. In
study 1 cassava was found to have a significantly higher
(x
2
11·9, degrees of freedom 4; P¼0·018) ranked GI
than the other foods tested at site 1. In study 2 the differ-
ences in AUC and GI among the foods tested were not stat-
istically significant.
The blood glucose response curves for the three boiled
(whole and crushed) tubers compared with the response
for white bread are shown in Fig. 1. Table 4 shows the
AUC and GI values for the boiled þuncrushed and
Glycaemic index of Caribbean staples 973

boiled þcrushed tubers compared with the response for
white bread. The AUC and the GI values of uncrushed
and crushed preparations of the foods tested were not sig-
nificantly different. When data for boiled whole dasheen,
Irish potato and tannia from both studies (study 2 and 3)
were pooled, their GI values, respectively, were 76 (SEM
7), 69 (SEM 6) and 55 (SEM 5), with the difference between
dasheen and tannia being statistically significant
(P¼0·015).
Discussion
The GI values of staples commonly eaten in the Caribbean
were determined in the present two-centre study using a
standard protocol. The foods selected for testing were
based on 24 h recalls (Wolever et al. 2002), food balance
sheets (Sinha, 1995) and discussions with nutrition prac-
titioners in the Caribbean. There was good concordance
between the two centres with respect to the mean and
within-subject variation of glycaemic responses elicited
by white bread. Although there was significant variation
in the glucose response to white bread between subjects,
there was no significant variation in GI values between
subjects. Additionally, in a recent study white bread
made from flour obtained at the respective study sites
was shown to have comparable GI (Wolever et al. 2003).
Based on the mean GI values obtained for dasheen (77
and 75), tannia (60 and 50) and Irish potato (71 and 66)
in studies 2 and 3, the average between-study standard
deviation was 6·6, which is less than the average
between-laboratory standard deviation of 9 previously
reported (Wolever et al. 2003). We did not undertake
measurements of insulin in the present study due to finan-
cial constraints but also because our main objective was to
generate GI data for Caribbean foods.
Fig. 1. Blood glucose responses in groups of eight to ten normal subjects elicited by test foods (green banana (a), breadfruit (b), cassava (c),
roti (d), tannia (e), Irish potato (f), dasheen (g), eddoe (h), crushed tannia (i), crushed Irish potato (j), crushed dasheen (k), yam (l)) (W,O)
compared with white bread (X) consumed by the same subjects. For (i) to (k), the results for uncrushed tubers are shown (W) as well as those
for crushed tubers (O). Mean values are shown, with standard errors of the mean represented by vertical bars. Bars have not been shown if
they are smaller than the symbol or overlap other symbols or bars. Mean values were significantly different from that for white bread
(P,0·05): * uncrushed foods; † crushed foods.
Table 2. Incremental areas under the glucose response curve
(mmol £min/l) for white bread
Volunteer
Code no. 1 no. 2 no. 3 Mean SD CV (%)
Site 1
101 203 179 218 200 20 10
102 167 156 223 182 36 20
103 117 105 102 108 8 7
104 103 120 96 106 12 11
105 252 251 231 245 12 5
106 221 266 199 228 34 15
107 158 171 187 172 15 9
108 238 288 197 241 46 19
Mean 182 192 182 185 23 12
SEM 20 24 19 19 5 2
Site 2
201 146 148 123 139 14 10
202 462 393 452 435 37 9
203 85 97 179 120 51 43
204 71 80 86 79 7 10
205 212 135 134 160 45 28
206 121 125 205 150 47 32
207 167 127 147 147 20 14
208 75 114 120 103 24 24
209 116 131 96 114 18 15
210 158 190 142 163 24 15
Mean 161 154 168 161 29 20
SEM 36 28 33 32 5 4
D. D. Ramdath et al.974

The method used to measure glucose and the type of
blood samples analysed differed at the two study sites.
However, results from a multi-centre study, which included
both study sites, showed that derivation of GI is not
affected by the analytical method used for glucose determi-
nation (Wolever et al. 2003). Additionally, GI values
obtained from capillary whole blood or plasma were not
different and produced a smaller standard deviation in
comparison with venous blood (Wolever et al. 2003). At
site 2, volunteers were allowed the choice of water,
unsweetened tea or coffee, whereas at site 1 all meals
were eaten with water. The former practice does not
affect the glucose AUC significantly (Young & Wolever,
1998), and is considered acceptable providing that volun-
teers keep constant their chosen drink for the duration of
the GI studies (Wolever et al. 1991).
We were unable to detect significant differences in gly-
caemic response between the tubers in the individual
studies, despite apparently large-enough differences in
mean AUC: 26 % in study 2, dasheen v. eddoe; 34 % in
study 3, uncrushed dasheen v. crushed tannia. The original
power analysis indicates that the probabilities of failing to
detect differences of 26 and 34 %, respectively, are 0·16
and 0·04. Thus, in study 2 the largest difference between
tubers, 26 %, was such that there was a relatively high
chance (0·16, or about 1:6) of failing to detect it. On the
other hand, in study 3 it is improbable (P,0·05) that the
failure to detect the 34 % difference in AUC was due to
chance, and hence is more probably due to the existence
of greater variability than expected. However, when the
results for the whole tubers in studies 2 and 3 were
pooled, the difference in GI between dasheen and tannia
became significant. Even if a very conservative Bonferroni
correction is applied to the pooled data to adjust for the fact
that the same data were used for two statistical tests, the
difference is still significant (P¼0·03).
The tubers reported in the present paper have not been
studied previously, so it was important to calibrate our
studies with a food, preferably a tuber, with a known
GI. Irish potato was chosen for this purpose. The GI
values obtained for Irish potato in studies 2 and 3 (71
and 66) not only agree with each other but also are similar
to GI values we have previously published for boiled or
baked Prince Edward Island, white and new potatoes
(range 55– 70) (Jenkins et al. 1981, 1983; Wolever et al.
1994).
The studies showed that the nine different test foods
have GI values which varied over a 1·5-fold range. The
reason for the relative differences is not known; however,
one possibility is that the physical and chemical character-
istics of the starch in these foods vary. Further work is
therefore needed to determine the characteristics of the
starch in the various tubers. The cook – cooling– re-warm-
ing cycle to which the tubers were subjected could have
affected the amount of resistant starch (RS) they contained.
About 7 % of the starch in reheated boiled potatoes escapes
digestion in the small intestine starch compared with about
3 % in freshly cooked potato (Englyst & Cummings, 1987),
a difference which seems too small to have a detectable
effect on glycaemic responses. The RS content of foods
similar to those reported here (roti, yam, eddoes, cassava,
Table 3. Incremental areas under the glucose response curve (AUC) and glycaemic index (GI) for test foods and white bread in study 1 and study 2
(Mean values and standard errors of the mean)
Study 1 Study 2
Green
banana Breadfruit Cassava Roti White bread Dasheen Eddoe Irish potato Tannia White yam White bread
Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM
AUC 171 38 142 12 243* 39 172 33 185 19 168 33 125 23 151 24 137 33 134 20 161 32
GI
wb
92 16 83 13 133* 16 91 13 100† – 109 14 86 14 101 11 84 7 88 9 100† –
GI 65 11 60 9 94* 11 65 9 71† – 77 10 61 10 71 8 60 5 62 6 71† –
GI
wb
, GI based on bread reference (i.e. GI of white bread ¼100).
* Mean value was significantly different from those for the other test foods (P,0·01).
†NoSEM shown because of zero degrees of freedom for a defined value.
Glycaemic index of Caribbean staples 975

potato) is low, with a range of 1·8 –3·4 g/11 g dry weight
(Platel & Shurpalekar, 1994). More recently, the effect of
RS on GI was examined using barley (about 15 % RS),
and there was no significant effect on GI after adjusting
for RS content (Wolever et al. 2003).
The finding that crushing did not result in significant
differences in GI suggests that this procedure did not
affect the foods’ physical structure sufficiently to produce
elevated glycaemic responses. The GI for Irish potato did
not increase when it was crushed in these studies. This is
consistent with the findings of a previous study in which
the GI of boiled Prince Edward Island potatoes (64) did
not increase significantly when the potatoes were mashed
(74) (Wolever et al. 1994).
Using published criteria (Brand-Miller et al. 2003b)itis
possible to classify dasheen and cassava as being high-GI
foods, whereas the other tubers studied (i.e. yam, tannia,
eddoes, and Irish potato, cooking ‘green’ banana, and
breadfruit) and roti can be designated intermediate-GI
foods. Food consumption patterns are not available for
most Caribbean countries; however, of the staple foods
tested, cassava, dasheen, Irish potato, yam and roti are
probably eaten more often than ‘green’ banana, breadfruit,
eddoes and tannia. When the former are added to the other
frequently used staples (bread and rice) it becomes appar-
ent that the Caribbean diet is probably one with a high GI.
This is supported by the finding of a mean diet GI of 58 in
a sample of healthy individuals in Trinidad and Tobago
(Wolever et al. 2002), which can be compared with
median values of 52 and 50, respectively for middle-aged
men and women in the USA (Salmeron et al. 1997a,b).
The range of GI between staples is sufficiently large, and
could result in beneficial health effects if those consuming
a Caribbean diet reduced their intakes of staples with high
GI and increased the consumption of those with intermedi-
ate GI. This is particularly important since small changes
in diet GI are associated with a significant reduction in
CHD risk (Liu et al. 2000), diabetes risk (Salmeron et al.
1997a,b) and improvements in insulin sensitivity and gly-
caemic control (Frost et al. 1998; Wolever & Mehling,
2002; Brand-Miller et al. 2003a). Ideally, low-GI Carib-
bean staples need to be identified and their usage pro-
moted; however, this will require further studies. The
findings of the present study can provide useful guidance
for health workers involved in meal planning for diabetics
and diabetes education programmes, especially in Carib-
bean countries and in Caribbean migrants. It can also be
used to achieve healthy eating and plan chronic-disease
risk-reduction programmes in high-risk populations.
It should now be possible to advocate for evidence-based
changes in the type and frequency of staples used by Car-
ibbean individuals.
Acknowledgements
The present work was partially supported by grants from
the Caribbean Health Research Council and the University
of the West Indies (St Augustine) Research and Publication
Funds. We thank Mr B. Mahabir, Ms N. Ramdhanie, Ms
A. Williams, Mr C. Bridgemohan, the National Herbarium
of Trinidad and Tobago, and the volunteers for their valu-
able contributions.
References
Association of Official Analytical Chemists (1995) Official
Methods of Analysis, 16th ed. Arlington, VA: AOAC.
Brand JC, Colagiuri S, Crossman S, Allan A, Roberts DCK &
Truswell AS (1992) Low-glycemic index foods improve
long-term glycemic control in NIDDM. Diabetes Care 15,
562–566.
Brand-Miller J, Hayne S, Petocz P & Colagiuri S (2003a) Low-
glycemic index diets in the management of diabetes: a meta-
analysis of randomized controlled trials. Diabetes Care 26,
2261–2267.
Brand-Miller J, Wolever TMS, Foster-Powell K & Colagiuri S
(2003b)The New Glucose Revolution: the Authoritative
Guide to the Glycemic Index. New York: Marlowe & Com-
pany.
Byrnes SE, Brand-Miller JC & Denyer GS (1995) Amylopectin
starch promotes the development of insulin resistance in rats.
J Nutr 125, 1430–1437.
Caribbean Food and Nutrition Institute (1994) Meal Planning for
Diabetes, 4th ed., Kingston, Jamaica: Caribbean Food and
Nutrition Institute Press.
Caribbean Food and Nutrition Institute (1998) Food Composition
Tables for Use in the English-Speaking Caribbean, 2nd ed.,
Kingston, Jamaica: Caribbean Food and Nutrition Institute
Press.
Cooper RS, Rotimi CN, Kaufman JS, Owoaje EE, Fraser H,
Forrester T, Wilks R, Riste LK & Cruickshank JK (1997)
Prevalence of NIDDM among populations of the African
Diaspora. Diabetes Care 20, 343–348.
Cruickshank JK, Cooper J, Burnett M, MacDuff J & Drubra U
(1991) Ethnic differences in fasting plasma C-peptide and insu-
lin in relation to glucose tolerance and blood pressure. Lancet
338, 842–847.
Englyst NH & Cummings JH (1987) Digestion of polysaccharides
of potato in the small intestine of man. Am J Clin Nutr 45,
423–431.
Table 4. Effect of crushing tubers on incremental area under the glucose curve (AUC; mmol £min/l) and glycaemic index (GI)
Uncrushed
dasheen
Uncrushed
Irish potato
Uncrushed
tannia
Crushed
dasheen
Crushed Irish
potato
Crushed
tannia White bread
Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM
AUC 155 40 130 23 112 29 129 15 127 13 103 13 145 22
GI
wb
105 17 93 14 71 11 93 8 94 9 74 7 100* –
GI 75 12 66 10 50 8 66 6 66 7 52 5 71* –
GI
wb
, GI based on bread reference (i.e. GI of white bread ¼100).
*NoSEM shown because of zero degrees of freedom for a defined value.
D. D. Ramdath et al.976

Food and Agriculture Organization of the United Nations (1998)
Carbohydrates in Human Nutrition. FAO Food and Nutrition
Paper no. 66. Rome: FAO.
Ford ES & Liu S (2001) Glycemic index and serum high-density
lipoprotein cholesterol concentration among US adults. Arch
Intern Med 161, 572–576.
Frost G, Leeds A, Trew G, Margara R & Dornhorst A (1998)
Insulin sensitivity in women at risk of coronary heart disease
and the effect of a low glycemic index diet. Metabolism 47,
1245–1251.
Gulliford MC (1996) Epidemiological transition in Trinidad and
Tobago, West Indies 1953– 1992. Int J Epidemiol 25, 357 – 365.
Gulliford MC, Ariyanayagam-Baksh SM, Bickram L, Picou D &
Mahabir D (1995) Counting the cost of diabetic hospital admis-
sions from a multi-ethnic population in Trinidad. Diabet Med
12, 1077–1085.
Jenkins DJA, Wolever TMS, Jenkins AL, Thorne MJ, Lee R, Kal-
musky J, Reichert R & Wong GS (1983) The glycaemic index
of foods tested in diabetic patients: a new basis for carbo-
hydrate exchange favouring the use of legumes. Diabetologia
24, 257–264.
Jenkins DJA, Wolever TMS, Kalmusky J, et al. (1987) Low-gly-
cemic index in hyperlipidemia: use of traditional starchy foods.
Am J Clin Nutr 46, 66–71.
Jenkins DJA, Wolever TMS, Taylor RH, Barker HM, Fielden H,
Baldwin JM, Bowling AC, Newman HC, Jenkins AL & Goff
DV (1981) Glycemic index of foods: a physiological basis
for carbohydrate exchange. Am J Clin Nutr 34, 362 – 366.
Liu S & Manson JE (2001) Dietary carbohydrates, physical
activity, obesity, and the ‘metabolic syndrome’ as predictors
of coronary heart disease. Curr Opin Lipidol 12, 395 – 404.
Liu S, Willett W, Stampfer M, Hu F, Franz M, Sampson L,
Hennekens C & Manson J (2000) A prospective study of diet-
ary glycemic load, carbohydrate intake and risk of coronary
heart disease in US women. Am J Clin Nutr 71, 1455 – 1461.
Platel K & Shurpalekar KS (1994) Resistant starch content of
Indian foods. Plant Foods Hum Nutr 45, 91 – 95.
Salmeron J, Ascherio A, Rimm EB, Colditz GA, Spiegelman D,
Jenkins DJ, Stampfer MJ, Wing AL & Willett WC (1997a)
Dietary fiber, glycemic load, and risk of NIDDM in men.
Diabetes Care 20, 545–550.
Salmeron J, Manson JE, Stampfer MJ, Colditz GA, Wing AL &
Willett WC (1997b) Dietary fiber, glycemic load, and risk of
non-insulin-dependent diabetes mellitus in women. JAMA
227, 472–477.
Sinha DP (1995) Changing patterns of food, nutrition and health
in the Caribbean. Nutr Res 15, 899– 938.
Soh NL & Brand-Miller J (1999) The glycaemic index of pota-
toes: the effect of variety, cooking method and maturity. Eur
J Clin Nutr 53, 249–254.
Thomas DE, Brotherhood JR & Brand JC (1991) Carbohydrate
feeding before exercise: effect of glycemic index. Int J
Sports Med 12, 180–186.
Wolever TMS, Isaacs RLC & Ramdath DD (2002) Lower diet
glycaemic index in African than South Asian men in Trinidad
and Tobago. Int J Food Sci Nutr 53, 297 – 303.
Wolever TMS, Jenkins DJA, Jenkins AL & Josse RG (1991) The
glycemic index: methodology and clinical implications. Am J
Clin Nutr 54, 846–854.
Wolever TMS, Katzman-Relle L, Jenkins AL, Vuksan V,
Josse RG & Jenkins DJA (1994) Glycaemic index of 102 com-
plex carbohydrate foods in patients with diabetes. Nutr Res 14,
651–669.
Wolever TMS & Mehling C (2002) High-carbohydrate-low-gly-
caemic index dietary advice improves glucose disposition
index in subjects with impaired glucose tolerance. Br J Nutr
87, 477–487.
Wolever TMS, Vorster HH, Bjo
¨rck I, et al. (2003) Determination
of the glycaemic index of foods: interlaboratory study. Eur J
Clin Nutr 57, 475–482.
Young KWH & Wolever TMS (1998) Effect of volume
and type of beverage consumed with a standard test meal
on postprandial blood glucose responses. Nutr Res 18,
1857–1863.
Glycaemic index of Caribbean staples 977