SEARCH

SEARCH BY CITATION

Keywords:

  • glycemic response;
  • glycemic index;
  • blood glucose

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

The Glycemic Index (GI) is a measure of the extent of the change in blood glucose content (glycemic response) following consumption of digestible carbohydrate, relative to a standard such as glucose. We have explored whether the reported GIs of foods are a sufficient guide to a person wishing to avoid large glycemic responses and thereby avoid hyperglycemia. For this purpose, volunteers carried out multiple tests of four foods, following overnight fasting, measuring the glycemic response over 2 H. The areas under the blood glucose/time curves (AUCs) were compared. Each food tester displayed individual, characteristic glycemic responses to each food, unrelated to any other tester's response. Wide variations (up to 5-fold) were seen between the average AUCs for the same test by different testers. The absolute magnitudes of the glycemic responses are important for individuals trying to control blood sugar and/or body weight, but using published GI lists as a guide to control the glycemic response is not fully informative. This is because in calculating the GI, individual glycemic responses to glucose are normalized to 100. GI values are, therefore, relative and are not necessarily a reliable guide to the person's actual individual AUC when consuming a food. Without knowledge of the person's characteristic blood glucose responses, reliance only on the GI may be misleading. © 2010 IUBMB IUBMB Life, 62(8): 637–641, 2010.

INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

The Glycemic Index (GI) is a well-established parameter, intended to reflect the magnitude of the blood glucose response elicited after consuming digestible carbohydrate. To assess glycemic responses, the areas under the blood-glucose/time curves (AUCs) for several individuals are measured after overnight fasting, averaged, and compared with the average AUC for glucose, the usual standard of reference, taken as having GI 100. Few foods elicit higher responses than glucose. AUCs are measured in mmol × min/L.

Tables of GI values of hundreds of solid and liquid foods are available for those interested in restricting spikes in blood glucose and some are displayed on the packages of the same foodstuffs. The general perception is that the GI provides an informed choice for managing blood glucose levels, and for selection of carbohydrate foods as recommended in many popular diets.

We have asked whether the blood glucose response by different individuals eating the same amount of the same food is uniform, around an average value, or is it personal to the individual and unlikely to bear any predictable relation to any other person's response? It is our contention that since values of GI are simply averages, relative to the average response to glucose, and individual glycemic responses show great variations between testers, they give no indication of what the actual response to a food might be for a particular person.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

The subjects were students and faculty of the University of Miami who gave informed consent to a protocol approved by the University of Miami Miller School of Medicine Human Subjects Research Office (Study number 20060725). Their ages ranged from 18 to 46, mostly 20–30 years.

The group included seven males and four females.

The testers consumed glucose (Sigma, 50 g in 355 mL of aqueous solution), Wonder Bread (two slices, 28 g of digestible carbohydrate), lemon-lime Gatorade (840 mL, 50 g of digestible carbohydrate), or brown rice syrup (42 g of digestible carbohydrate in 355 mL of aqueous solution). Except for glucose, the digestible carbohydrate contents were calculated from the product labels.

The measurements were made after an overnight fast. The food was consumed in equal portions at 0, 4, 8, and 12 min, after measuring blood glucose by fingerprick, using an Ascensia Breeze 2 glucose meter. Additional blood glucose measurements were then made at 15, 30, 45, 60, 90, and 120 min. The measurements were discontinued if the blood glucose level fell to the baseline earlier than 120 min. Frequently, this had happened by 90 min. All subjects made at least four tests with each food. The AUC was calculated by the trapezoidal method.

A few results are included from a to-be-published study compiled from the average of two tests made when eating/drinking in 2–4 min and two tests made when eating/drinking in five equal portions over 15 min. We showed that there was no statistical difference between the AUCs for slow and fast consumption, and these added results blend in with the 12 min consumption tests. They are marked with an asterisk.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

It was as a result of making many comparisons of AUCs when using the same subjects and comparing AUCs for rapid and slow consumption that we began to notice that each person, relative to others, displayed AUCs that in an absolute sense were characteristic of that person. Thus, of two subjects, one gave AUCs consistently higher than the other. In 23 tests with solid foods, this person was higher (often much higher) than the other in 22 tests, and with respect to liquid foods, was higher in 27 of 29 tests. Comparisons of data of other students revealed similar patterns and allowed us quickly to establish a hierarchy where the responses varied four-fold.

In deciding to investigate this phenomenon further we increased our testing panel and used a standard regime of splitting the food into four equal portions, consumed over 12 min, a time used by the GI measuring laboratory at the University of Sydney (1). The characteristic inter-personal differences seen with fast and slow consumption were still seen with glucose when the 12-min regime was employed. We also tested Wonder Bread, Gatorade, and brown rice syrup in the same way. The collected data are displayed in a number of ways. In Fig. 1, the AUCs +/− SEM for 11 testers drinking glucose solution are displayed. The group displays average AUCs that differ by a factor of 4.8, low to high.

thumbnail image

Figure 1. The average AUCs +/− SEM from 11 subjects drinking 50 g of glucose, in at least four tests per person, arranged left to right in order of increasing average AUC. Results are expressed as mmol × min/L. These are the same volunteers whose AUCs are also to be seen in Tables 1–3 and Figs. 2–4, identified by a letter code.

Download figure to PowerPoint

In Table 1, the average AUCs for the same 11 subjects who had tested glucose and at least two of the other three foods are displayed. The results are classified as low or high on the basis of perceived gaps in the averages. With one exception, all subjects were uniformly low or uniformly high for all foods tested. Also, they generally ranked in the same order with C always lowest and D highest, three times out of four.

Table 1. Average AUCs for subjects carrying out at least four tests each with glucose, Wonder Bread, Gatorade, and brown rice syrup
 GlucoseWonder BreadGatoradeBrown rice syrup
  1. The Low responses are those with AUCs of 117 and below.

  2. Responses with an asterisk are the AUCs obtained by averaging those obtained from two rapid and two slow rates of consumption.

  3. Units are mmol × min/L.

C83Low65Low57Low55Low
A96Low89Low67Low  
M117Low70Low94Low107Low
T1188*High  140High165High
V1190High131High123High102Low
G203High129High162High120High
P291High136High126High198High
L291High126High194High187High
T2293*High153*High160High212*High
V2300*High143High  185High
D397High191High200High161High
Average223 (83–397)123 (65–191)132 (67–200)149 (55–212)

Table 2 shows the individual AUCs for subjects C and D for all four foods. Note that there was no instance of overlap for any food, while the average AUCs showed C to be about five times lower than D for glucose and three times lower for the other three foods (Table 2).

Table 2. Comparison of two subjects, C and D, with widely differing AUCs for four foods
 GlucoseWonder BreadGatoradeBrown rice syrup
  1. Note that when comparing the individual AUCs, there is no instance of overlap between C and D.

AUCs (mmol × min/L)
 C97519428
751035249
65582473
94465872
 Average83655755
 D445107334127
237274150200
446137160158
460247154160
 Average397191200161
 Ratio C:D1:4.81:2.91:3.51:2.9
Personal glycemic index
 C100786966
 D100485041

Table 3 shows the calculated “personal glycemic indexes” for the same 11 individuals for whom AUCs for the four foods are shown in Table 1, taking glucose as 100. Note that these “personal glycemic indexes” are not to be compared with published values of GI because, although 50 g of glucose was consumed, the available carbohydrate in the other test samples was not a consistent 50 g. Even though the AUCs obtained with glucose were normalized to 100, there was a wide variation between individual GIs (about two-fold) for each of the three other foods. There was a tendency for the personal GI to fall when proceeding from persons with a low AUC to those with a high AUC. The foods did not reliably stack up in the same order of GI.

The average AUCs +/− SEM for the tests with all four foods are shown in Figs. 1–4.

thumbnail image

Figure 2. The AUCs obtained when 10 subjects ate Wonder Bread (28 g of digestible carbohydrate), expressed as averages of at least four tests per person +/− SEM. See also Fig. 1.

Download figure to PowerPoint

thumbnail image

Figure 3. 10 subjects drank Gatorade containing 50 g of digestible carbohydrate in at least four tests each. Plotted here are the average AUCs +/− SEM. See also Fig. 1.

Download figure to PowerPoint

thumbnail image

Figure 4. The AUCs obtained when 10 subjects drank a solution of brown rice syrup containing 42 g of digestible carbohydrate. At least four tests were carried out by each person. Results are expressed as average AUCs +/− SEM. See also Fig. 1.

Download figure to PowerPoint

Table 3. Personal glycemic indexes of three foods, relative to glucose
 AverageGlucoseWonder BreadGatoradeBrown rice
AUC for glucosesyrup
  • These are the same subjects as in Table 1, where the GI values shown here have been calculated from the AUCs in Table 1.

  • a

    The number in parentheses is the average AUC for at least four tests with 50 g of glucose.

  • b

    AUCs for glucose obtained by averaging two tests each by fast and slow consumption (see Table 1).

C(83)a100786966
A(96)1009370 
M(117)100608091
T1(188)b100 7588
V1(190)100696554
G(203)100648060
P(291)100474368
L(291)100436764
T2(293)b100525572
V2(300)b10048 62
D(397)100485041
Average 10060 (43–93)65 (43–80)67 (41–91)

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

Despite the wide variation between the AUCs displayed by any one individual in the same test (see Table 2), it is our impression that for any such individual, repeated testing provides an average AUC that represents that person's characteristic/individualistic glycemic response to a particular food. The results from repeated measurements of AUCs will revolve around a mean characteristic of that person, not around a mean derived from a group of individuals. Is it not therefore obscuring the huge differences in actual AUCs that, for the purpose of calculating the GI, all the glucose AUCs are normalized to 100? In evidence of this contention, we offer Fig. 1. These are the average AUCs + SEM of the 11 individuals making at least four tests each with 50 g of glucose (Table 1). They have been arranged in ascending order. Clearly there is no common mean to which the AUCs of all individuals will move closer with repeated testing. We also found similar large differences occurred in the AUCs for Wonder Bread (Fig. 2), Gatorade (Fig. 3), and brown rice syrup (Fig. 4).

It is important that where D had an average AUC of 191 for Wonder Bread (Table 2), half that recorded for glucose (397), this Wonder Bread blood glycemic response, in absolute terms, is 2.3 times higher than the response shown by C for glucose, which has the highest glycemic index of the foods tested. Even if the GIs derived from these AUCs were the same for C and D (which they are not, Table 3), is it not important, even more than knowing the GIs, to acknowledge the widely differing AUCs on which the GIs are based?

We conclude that each individual has a characteristic glycemic response to foods that contain digestible carbohydrate and that these responses can vary widely between individuals. Thus, average values of AUC from a group of individuals and the published GIs derived therefrom are unable to predict what will be the actual glycemic response of an individual.

This idea of individual reproducible and characteristic glycemic responses is not that which generally prevails. Testing laboratories that have compared AUCs and GIs from the same foods admit of wide variation in AUCs for the same test, but do not seem to acknowledge a link between the AUC and personal characteristics. Rather, they contend that “apparent variations between individuals reflect not real differences, but day-to-day variability within individuals. With repeated testing of a given food, all individuals move closer to the mean of the group” (2; p. 42). This comment was made with respect to the GI values and may be correct. But it overlooks the fact that the AUCs themselves do not move closer to the mean of the group. That can only occur when all the glucose AUCs are normalized to 100.

And, “subject characteristics such as age, sex, BMI, and ethnicity may have contributed to the highly significant differences in absolute glycemic responses between subjects. However, these factors had no significant effects on the GI values obtained in this study. This is consistent with previous data in subjects with diabetes, which suggests that most of the variation of GI values is due to within-subject variation” (3; p. 481). Our approach is quite different with respect to individualistic responses. The differences between individuals are real, consistent, and reproducible.

The GI testing labs also write that “Between-laboratory variation in GI appears to be related to random, day-to-day variation in glycemic responses within subjects. Thus, finding ways to reduce the within-subject variation may be the most effective strategy to improve the precision of measurement of GI values” (3; p. 481).

The “highly significant” differences in AUCs were exemplified in an interlaboratory study where seven labs tested glucose and other foods using 8–12 subjects each, who each made eight tests (4). The average AUCs for glucose ranged from 97 to 281 (note our score of 223 in Table 1). In calculating GIs from these AUCs, both 281 and 97 were assigned the same GI value of 100. It is not surprising that the GIs of the other foods, calculated relative to a common GI of 100, were closer to each other than were the parent AUCs, but even so, 2-fold differences in the GIs for 2 foods were noted between different labs.

The authors then state, “GI is the same in different subjects and therefore is a property of the food and not of the subject in whom this was measured” (4; p. 253S). To claim that “GI is the same in different subjects” is highly misleading. It amounts to saying that the GI of glucose is the same in different subjects because no matter what is the AUC a person displays when consuming glucose, it is normalized to 100. This approach is based on the belief that no matter how much the AUCs vary, if tested often enough, everyone will display average AUCs for glucose and the test food that are in the same ratio, and therefore, have the same GI. However, the results of several thousand individual tests (4) contain some glaring exceptions. The claim that the GI is not influenced by the subject consuming the food obscures the huge differences in individual AUCs. GI is a relative measure where AUC is an absolute measure. Which is more important? That the consumer should know that food X has a GI of 70, versus glucose = 100? Or that a person like D may have an AUC at the high end of a scale in which another person consuming the same food has an AUC several times smaller?

“A high blood glucose concentration…has undesirable consequences…is a strong independent risk factor for morbidity and mortality, …creates oxidative stress etc” (2; p. 41). This is not in contention. We believe that for maintaining blood glucose at a safe, low level, and/or to select foodstuffs for a particular diet program, a glucose tolerance test is essential. For a person like our subject C (Table 2), who displays low glycemic responses, one could conclude that the GI is not really relevant in the effort to avoid hyperglycemia. But for a person like our subject D, the GI would be a very useful guide.

The testing lab consortium (4) would view very high individual values like those of D in Tables 1–3 as “outliers”, to be excluded (5) and indeed did exclude outliers that were more than two SDs above the mean in their research and did not specify whether they were considering “implausible” results above or below the mean (5; p. 163). This is exactly what we have found. Some of our subjects provided test averages that were internally consistent in being low, medium, or high, when compared with averages of other subjects. Using the rule of two SDs, we would have to exclude D from Fig. 1.

It seems to us that investigators have assumed that the variations occur because of imperfect measuring techniques (3; p. 481), with results influenced by within- and between-person differences, with the implication that the GI is somehow a fixed quantity. What is more, calculating SDs for any set(s) of results with multiple testers is meaningless, because there is no fixed mean. That will depend on the group(s) of individuals selected for the test(s).

DeVries (6) concluded that “GI results are unpredictable.” He too seemed to be working on the premise that the GI of a food ought to be a precise quantity, but the unpredictability led him to dismiss GI as having any meaning. We agree about the unpredictability. To repeat, the results are reflective of the group of individuals chosen to make the test. The huge variations seen in the interlaboratory study (4) bear this out.

For Williams et al. (7; p. 364) “the unpredictability of individual responses, even if they are the results of day-to-day variation, places limitations on the clinical usefulness of GI” and “the measures of reliability of the GI of foods indicate that their repeatability is unacceptably low” (7; p. 367).

Accepting that glycemic responses are indeed unpredictable and dependent on the individual test subject would herald a new outlook on the usefulness, and the limitations, of GI. It would stress that the GI is not a substitute for a glucose tolerance test and that the glycemic response to a food is a property of the individual who consumes it. This would encourage exploration of the reasons for the intra-personal differences in AUCs, so as to refine the testing procedures in the hope that the differences might diminish or disappear.

Thus, we must concentrate on the actual AUCs recorded by individuals because those are absolute values that pertain directly to possible hyperglycemia and the obvious consequences.

The GI is simply a relative value useful in developing eating habits to obtain lower blood glucose responses. We asked if the fasting blood glucose was a predictor of the size of the AUC. The averages of four fasting glucose values each by testers C, G, and D (Fig. 1) were C = 83 mg/dL, G = 86 mg/dL, and D = 89 mg/dL, where the average AUCs for four tests with 50 g of glucose were 83, 203, and 397 mmol × min/L, respectively.

In summary, while the GI is a valuable parameter, its use in the absence of knowledge of an individual's characteristic glycemic response is imperfect. No-one would expect a dietitian, armed with the AUCs, to render the same advice to subject C as to subject D. Yet the one-size-fits-all concept is what consumers of digestible carbohydrate are being encouraged to follow.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

The research described here was funded in part by the Agatston Research Foundation. The authors thank Drs. P. Kurlansky and R. B. Goldberg for their critical analyses of our interpretation of the data. They also thank many undergraduate, graduate and M.D. students, and faculty of the University of Miami who took part in these tests.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  • 1
    Personal communication from Dr Jennie Brand-Miller. University of Sydney, NSW, Australia.
  • 2
    Brand-Miller, J. ( 2007) The glycemic index as a measure of health and nutritional quality: an Australian perspective. Cereal Foods World 52, 4144.
  • 3
    Wolever, T. M. S., Vorster, H. H., Bjorck, I., Brand-Miller, J., Brighenti, F., Mann, J. I., Ramdath, D. D., Granfeldt, Y., Holt, S., Perry, T. I., Venter, C., and Wu, X.. ( 2003) Determination of the glycaemic index of foods: interlaboratory study. Eur. J. Clin. Nutr. 57, 475482.
  • 4
    Wolever, T. M. S., et al. ( 2008) Measuring the glycemic index of foods: interlaboratory study. Am. J. Clin. Nutr. 87, 24752575.
  • 5
    Brouns, F., Bjork, I., Frayn, K. N., Gibbs, A. L., Lang, V., Slama, G., and Wolever, T. M. S. ( 2005) Glycemic index methodology. Nutr. Res. Rev. 18, 154171.
  • 6
    De Vries, J. W. ( 2007) Glycemic index: the analytical perspective. Cereal Foods World 52, 4549.
  • 7
    Williams, S. M., Venn, B. J., Perry, T., Brown, R., Wallace, A., Mann, J. I., and Green, T. ( 2008) Another approach to estimating the reliability of the glycemic index. Br. J. Nutr. 100, 364372.