Filetype PDF | Posted on 05 Jan 2023 | 2 years ago
Authentication
digital resources
144x Filetype PDF File size 0.83 MB Source: diabetesjournals.org
Thomas M.S. Wolever, BM, PhD
Glycemic Index of Foods David |.A. Jenkins, DM, DSc
Vladimir Vuksan, PhD
in Individual Subjects Robert G. Josse, MD
Gerald S. Wong, MD
Alexandra L. Jenkins, BSc, RPDt
Downloaded from http://diabetesjournals.org/care/article-pdf/13/2/126/439321/13-2-126.pdf by guest on 04 January 2023
We studied 12 subjects with diabetes to determine elusion is not necessarily valid if the subjects tested each
how well the glycemic index (Gl) predicted the ranking food only once, and the lack of consistency was due to
of glycemic responses of different foods in individuals. within-individual variability Therefore, repeated testing
All subjects ate three mixed meals (bread, rice, or of the same meal in each subject is required to deter-
spaghetti with GIs of 100, 79, and 61, respectively) four mine whether the variability of glycemic responses is
times in a randomized complete block design. The mean due to consistent differences between subjects or to day-
glycemic response areas of the different meals ranked to-day variability within individuals.
according to the predicted Gl in every individual. The Our objective was to determine whether different in-
observed mean ± SD Gl values of the meals were dividuals shared common Gl values of foods, which
significantly different from each other (bread 100 ± 7,
rice 75 ± 9, spaghetti 54 ± 9), with no significant would allow the Gl to be used to predict the ranking of
difference in response between subjects. It is concluded glycemic responses in individual subjects. Therefore, we
that individuals share common mean Gl values for determined the glycemic responses of different test meals
different foods. Within confidence limits determined with a range of Gl values taken on four occasions each
by the variability of glycemic responses, the number by a group of diabetic subjects.
of repeated tests conducted, and the expected Gl
difference, the Gl can be used to predict the ranking of
the mean glycemic responses of mixed meals taken by RESEARCH DESIGN AND METHODS
individuals. Diabetes Care 13:126-32, 1990
Twelve patients with diabetes were studied on an out-
patient basis on 12 separate occasions over a 3- to 6-
here is much controversy about the validity of mo period with procedures that had been approved by
classifying foods according to their glycemic re- the University of Toronto Ethical Review Committee on
sponses with the glycemic index (Gl) or relative the use of human subjects. Volunteers entered the test-
Tglucose area (RGA) (1-4). It has been pointed out ing area at St. Michael's Hospital, Toronto, in the morn-
that the glycemic ranking of foods in individuals is not ing after overnight fasts. The subjects were chosen to be
always as expected (5-8), implying that glycemic re- dissimilar with respect to age, sex, body weight, treat-
sponses are idiosyncratic in different subjects. This con- ment, and type of diabetes to increase the chances of
detecting differences between the individuals (Table 1).
Only one of the seven patients taking insulin (subject 2)
had a known history of ketosis and could be classified
From the Division of Endocrinology and Metabolism, St. Michael's Hospital, on this basis as having insulin-dependent diabetes mel-
and the Department of Nutritional Sciences, Faculty of Medicine, University of litus (IDDM). The other six patients on insulin were di-
Toronto, Toronto, Ontario, Canada. agnosed in other centers, and records of ketosis were
Address correspondence and reprint requests to Thomas M.S. Wolever, As-
sistant Professor, Department of Nutritional Sciences, Faculty of Medicine, Uni- not available. In them, the diagnosis of IDDM or non-
versity of Toronto, Toronto, Ontario, Canada M5S 1A8. insulin-dependent diabetes mellitus (NIDDM) was based
Received for publication 7 June 1989 and accepted in revised form 13 Sep- on a fasting serum C-peptide below the lower limit of
tember 1989.
126 DIABETES CARE, VOL. 13, NO. 2, FEBRUARY 1990
T.M.S. WOLEVER AND ASSOCIATES
TABLE 1
Clinical details of subjects studied
Serum C-
Fasting blood peptide (pM)
Age Years BMI Type of glucose
2
Subject Sex (yr) diabetic (kg/m ) diabetes (mM) HbA Fasting 90 min Treatment
lc
1 27.2 I t 1.7 0.108 33 106 20 L a.m., 3-5 L p.m.
F 62 13 5.3 d
2 M 36 22 19.1 I 6.8 dt 2.8 0.065 60 83 8 L + 7-10 R a.m., 8-10 L + 7-15 R p.m.
3 F 67 12 21.4 N 8.1 it 3.4 0.079 311 371 35 La.m.
4 F 63 4 24.2 N 7 A dt 3.1 0.077 877 3970 35 L + 5 Ra.m.
5 F 66 8 24.2 N 8.6 it 1.2 0.064 258 662 5 mg G
N t 2.1 0.072 570 38 L + 8 R a.m.
6 M 72 20 21.0 8.8 1 311
7 t 1.1 0.047
F 66 8 26.2 N 7.3 H 513 1523 15-19 La.m.
8 F 53 13 30.8 I 7.4 dt 2.7 0.078 252 238 40-52 N a.m.
9 F 48 4 33.7 N 10.7 ± 1.8 0.097 334 351 10 mgC Downloaded from http://diabetesjournals.org/care/article-pdf/13/2/126/439321/13-2-126.pdf by guest on 04 January 2023
10 M 71 4 23.8 N 9.6 ± 0.6 0.061 358 517 20 mgC
11 M 80 9 23.9 N 6.7 ± 0.4 0.061 3510 5300 Diet alone
12 F 83 8 25.2 N 8.2 ± 0.4 0.054 1490 3050 5 L a.m.
I, IDDM; N, NIDDM; units insulin (L, lente; R, regular; N, NPH); C, glyburide. Values are means ± SD.
normal for our laboratory (200 pM) or on no rise at 90 meals on four different occasions according to a ran-
min after a test meal containing 50 g carbohydrate as domized complete block design (11). The meals con-
white bread (Table 1). sisted of a 50-g carbohydrate portion of either white
We wanted to have an 80% chance of detecting an bread (B), polished white rice (R), or white spaghetti (S)
extreme range of Gl values of 20 between subjects. The to which cheese and tomato were added, so that fat and
pooled coefficient of variation (C.V.) of glycemic re- protein comprised 31 and 19% of calories. Duplicate
sponses for our nine NIDDM and three IDDM subjects test meals were analyzed for macronutrient composition
was estimated to be 18.7% based on published C.V.s of (12; Table 2). Test meals were served with 1-2 cups of
glycemic responses in NIDDM and IDDM patients (9). coffee or tea with or without 30 ml 2% butterfat milk
With this degree of variation, power analysis suggested (beverage standard for each subject).
that each subject should be tested 12 times (10). Subjects began to eat their test meals 5-10 min after
Each subject consumed each of three different test taking their usual dose of insulin or oral hypoglycemic
TABLE 2
Test meal composition based on proximate analysis
Test meal Weight (g) Carbohydrate (g) Fat (g) ProteiYi (g) Energy (kcal)
Bread
White bread 118 50.0 0.3 8.6 237
Cheddar cheese 41 0.6 13.7 10.3 166
Stewed tomato* 60 2.1 0.01 0.4 10
Total 52.7 14.0 19.3 413
Percentage of energy 51 31 19
Rice
Polished white ricet 65 50.0 0.1 4.6 219
Cheddar cheese 42 0.6 14.0 10.5 170
Skim cheese 11 0.4 0.4 3.7 20
Stewed tomato* 60 2.1 0.01 0.4 10
Total 53.1 14.5 19.2 419
Percentage of energy 51 31 19
Spaghetti
White spaghetti 72 50.0 0.1 8.6 236
Cheddar cheese 42 0.6 14.0 10.5 170
Stewed tomato* 60 2.1 0.01 0.4 10
Total 52.7 14.1 19.5 416
Percentage of energy 51 31 19
*Del Monte-brand canned stewed tomato.
tWeighed raw.
DIABETES CARE, VOL. 13, NO. 2, FEBRUARY 1990 127
GLYCEMIC INDEX IN INDIVIDUALS
agent (if any). Meals were eaten within 5-10 min. Fin- First Repeat Second Repeat
ger-stick capillary blood samples were obtained at fast-
ing and at 30-min intervals for 3 h after the start of the
test meal for analysis of glucose with a YSI model 27AM —7 8
glucose analyzer (Yellow Springs, OH).
Subject 8 did not complete her last two tests because
of an exacerbation of her irritable bowel. The missing
values (1 B and 1 R test) were estimated in the analysis
of variance by use of the means of the other three re- LU
spective test meals. cr
Incremental areas under the glycemic response curves Third Repeat Fourth Repeat
were calculated geometrically (13). To assess differ- LU
en
ences between test meals and between subjects, anal- o
C_3
ysis of variance with repeated measures (ANOVARM),
with subject and test meal as the variables, was per- o
formed on all the glycemic response areas and on the o Downloaded from http://diabetesjournals.org/care/article-pdf/13/2/126/439321/13-2-126.pdf by guest on 04 January 2023
o
GIs, i.e., the glycemic response areas expressed as a ^4
percentage of each subject's own mean response to the
B meal (9). To assess the significance of differences within
subjects, the glycemic areas and CIs of each subject
were analyzed by ANOVARM with test meal as the vari- 0 60 120 180 0 60 120 180
able. After demonstration of significant heterogeneity, TIME (min)
the significance of differences between foods was de-
termined by Tukey's Q method to adjust for multiple FIG. 1. Mean blood glucose increments of 12 patients with
comparisons (14). non-insulin-dependent diabetes mellitus for 1st, 2nd, 3rd,
The ranking of glycemic responses was assessed within and 4th repeats of bread (solid lines), rice {dashed lines),
each subject by making all possible three-way compar- and spaghetti (dotted lines) test meals.
isons between the glycemic responses of the B, R, and
S meals. For each subject taking four B, four R, and four
S meals, there were 64 possible comparisons of individ- other, whether expressed as incremental areas (1105 ±
ual areas (36 in subject 8). To determine the effect of 246, 819 ± 204, and 581 ± 142 mM min, F(222) =
increasing the number of repeated tests performed on 29.21, P < 0.01) or GIs (100 ± 7, 75 ± 9, and 54 ±
the proportion of correct predictions, the means of all 9 mM min, F = 88.49, P < 0.01).
(2;22)
possible combinations of four tests taken two at a time Comparison of glycemic responses between sub-
were calculated for each test meal. This gave six values jects. There were highly significant differences between
for each test meal and 216 possible comparisons be- the incremental glycemic response areas of the different
tween B, R, and S (54 in subject 8). Next, the means of subjects (F(11(106) = 54.64, P < 0.01, Table 3). In addi-
all the combinations of the four tests taken three at a tion, there was a significant interaction between the ef-
time were calculated, resulting in 64 comparisons be- fects of test meal and subject (F 6) = 3.90, P < 0.01),
(22J0
tween B, R, and S (4 in subject 8). Finally, there was indicating that the differences between the test meals'
one comparison in each subject between the means of incremental glycemic response areas varied in the dif-
the four tests of B, R, and S. For each group of com- ferent subjects. However, when the results were ex-
parisons there were six possible outcomes: B > R > S, pressed as the Gl, there was no significant difference
B>S>R, R>B>S, R>S>B, S>B>R, andS between the subjects (F = 1.57, NS) and no inter-
(11106)
> R > B. If the glycemic responses of B, R, and S were action between the effects of test meal and subject
randomly distributed, 17% of the total comparisons would F(22,io6) = 1-12, NS, Table 3).
be correct (i.e., B > R > S) and 83% incorrect. The Comparison of different meals within subjects. In all
observed number of correct rankings in each subject 12 subjects, the mean glycemic response area for B was
2 greater than R, which was greater than S (Table 3). The
was compared with that expected by chance with x -
analysis (14). differences between test meals were not statistically sig-
nificant in subjects 2, 8, and 11 because of large within-
individual variability of glycemic responses. For the re-
RESULTS maining 9 subjects, B was significantly greater than R
in 5 and greater than S in all 9, with R being significantly
The mean glycemic responses for the first, second, third, greater than S in 4 (Table 3).
and fourth repeats of the three test meals are shown in Ranking of glycemic responses within subjects. The
Fig. 1. The pattern of the mean glycemic responses was overall proportion of correct rankings of individual tests
similar in every case with B > R > S. The overall was 71 % (P < 0.001, Table 4). In 11 of 12 subjects, the
mean ± SD glycemic responses of B, R, and S test meals, proportion of correct rankings was significantly greater
respectively, were all significantly different from each than would have been expected by chance (Table 4).
128 DIABETES CARE, VOL. 13, NO. 2, FEBRUARY 1990
T.M.S. WOLLVER AND ASSOCIATES
TABLE 3
Incremental areas under the curve and glycemic index values for four test meals of bread, rice, and spaghetti taken
by each subject
Area under the curve mM min Glycemic index
Subject Bread Rice Spaghetti Bread Rice Spaghetti
1 1925 ± 98 1493 ± 153* 802 ± 129t 100 ± 5 78 ± 8* 42 ± 7t
2 1740 ± 78 1373 ± 531 1243 ± 206 100 ± 4 79 ± 31 72 ± 12
3 1709 ± 197 1133 ± 281* 696 ± 83t 100 ± 12 66 ± 16* 41 ± 5+
4 1196 ± 115 855 ± 199* 111 ± 92* 100 ± 10 72 ± 17* 65 ± 8*
5 1224 ± 148 761 ± 268* 472 ± 148* 100 ± 12 62 ± 22* 39 ± 12*
6 1214 ± 236 1033 ± 107 544 ± 151t 100 ± 19 85 ± 9 45 ± 12t
7 970 ± 101 620 ± 68* 567 ± 83* 100 ± 10 64 ± 7* 59 ± 9*
8 748 ± 341 607 ± 204 553 ± 184 100 ± 46 81 ± 27 74 ± 25
9 909 ± 96 730 ± 150 365 ± 25t 100 ± 11 80 ± 16 40 ± 3t Downloaded from http://diabetesjournals.org/care/article-pdf/13/2/126/439321/13-2-126.pdf by guest on 04 January 2023
10 601 ± 76 420 ± 114 282 ± 90* 100 ± 13 70 ± 19* 47 ± 15*
11 603 ±115 451 ± 131 413 ± 101 100 ± 19 75 ± 22 69 ± 17
12 420 ± 117 350 ± 42 258 ± 43* 100 ± 28 83 ± 10 62 ± 10*
Values are means ± SD.
*P < 0.05 vs. bread.
+P < 0.05 vs. bread and rice.
The proportion of incorrect rankings was greater for sub- proportion of correct predictions increased to 83, 92,
jects with the most variable glycemic responses (r = and 100%, respectively.
0.644, P < 0.05). However, none of the clinical char-
acteristics of the subjects were significantly related to
the proportion of correct rankings (age, r = 0.379; du- DISCUSSION
ration of diabetes, r = -0.383; fasting C-peptide, r =
-0.318; postprandial C-peptide, r = -0.328; treat- he results demonstrate that different individuals
ment with insulin, r = 0.239; classification of diabetes, share the same mean Gl values of foods. Thus,
r = 0.392; mean fasting blood glucose, r = 0.346; SD differences in Gl observed when individual tests
of fasting blood glucose, r = -0.210; HbA , r = 0.338). Tare compared are due largely to day-to-day var-
1c
When comparisons were made between the means of iability within the same subject. When tests of B, R, and
two, three, and four repeated tests of each meal, the S were repeated four times by each subject, the Gl cor-
TABLE 4
Pooled SD of glycemic index values for each subject and number (percentage) of 3-way comparisons for which ranking
of glycemic responses in each individual was as predicted by Gl (i.e., bread > rice > spaghetti)
Number of area values averaged for comparisons
Subject Pooled SD 1 2 3 4
1 6.6 64 (100)* 216 (100) 64 (100) (100)
2 19.1 16(25) 126 (58) 48 (75) (100)
3 11.9 60 (94)* 216 (100) 64 (100) (100)
4 12.0 36 (56)* 156 (72) 56 (88) (100)
5 16.0 60 (94)* 210(97) 64 (100) (100)
6 14.3 48 (75)* 192 (89) 64 (100) (100)
7 8.8 44 (69)* 186 (86) 56 (88) (100)
8 29.5 14 (39)* 24 (44) 3(75) (100)
9 11.4 56 (88)* 210(97) 64(100) (100)
10 15.7 56 (88)* 204 (94) 64(100) (100)
11 19.4 28 (44)* 144 (67) 48 (75) (100)
12 18.1 48 (75)* 198 (92) 64 (100) (100)
Mean percentage
correct predictions 71 83 92 100
Comparisons were conducted with individual area values and all possible combinations of means of 2, 3, and 4 values for each meal. Values
in parentheses are percentages.
*P < 0.001.
DIABETES CARE, VOL. 13, NO. 2, FEBRUARY 1990 129
no reviews yet
Please Login to review.
