点击显示 收起
1 From the Departments of Health Promotion and Chronic Disease Prevention (KAH, MES, JRV, JGE, M-LH, HKS, and LMV) and of Health and Functional Capacity (JES), the National Public Health Institute, Helsinki, Finland; the Department of Public Health, University of Helsinki, Helsinki, Finland (JGE); the Department of Environmental Health, the National Public Health Institute, Helsinki, Finland (M-LH); and the Department of Clinical Nutrition, University of Kuopio, Kuopio, Finland (HMM)
2 Supported by grants from the Ministry of Agriculture and Forestry, the Doctoral Programs in Public Health (to MES), the Juho Vainio Foundation, the Yrjö Jahnsson Foundation, and the Finnish Cultural Foundation. 3 Reprints not available. Address correspondence to L Valsta, Department of Health Promotion and Chronic Disease Prevention, National Public Health Institute, Mannerheimintie 166, FI-00300 Helsinki, Finland. E-mail: liisa.valsta{at}ktl.fi.
ABSTRACT
Background: Methodologic choices affect measures of the glycemic index (GI). The effects on GI values of blood sampling site, reference food type, and the number of repeat tests have been insufficiently determined.
Objective: The objective was to study the effect of methodologic choices on GI values. Comparisons were made between venous and capillary blood sampling and between glucose and white bread as the reference food. The number of tests needed for the reference food was assessed. Rye bread, oatmeal porridge, and instant mashed potato were used as the test foods.
Design: Twelve healthy volunteers were served each test food once and both reference foods 3 times at 1-wk intervals in a random order after they had fasted overnight. Capillary and venous blood samples were drawn at intervals for 3 h after each study meal.
Results: GIs and their CVs based on capillary samples were lower than those based on venous samples. Two tests of glucose solution as the reference provided stable capillary GIs for the test foods. The capillary GIs did not differ significantly when white bread was used as the reference 1, 2, or 3 times, but the variation was lower when tests were performed 2 and 3 times. Capillary GIs with white bread as the reference were 1.3 times as high as those with glucose as the reference. The capillary GIs of rye bread, oatmeal porridge, and mashed potato were 77, 74, and 80, respectively, with glucose as the reference.
Conclusions: Capillary blood sampling should be used in the measurement of GI, and reference tests with glucose or white bread should be performed at least twice.
Key Words: Glycemic index • capillary blood • venous blood • methods • reference food • rye bread • oatmeal porridge • mashed potato
INTRODUCTION
A diet producing a low glycemic response is associated with significantly less insulin resistance and significantly lower prevalence of the metabolic syndrome (1), risk of type 2 diabetes (2), and risk of coronary artery disease (3) than is a diet producing a high glycemic response. However, the results are ambiguous, and the evidence from randomized controlled trials showing that low-glycemic-response diets reduce coronary artery disease and its risk factors is weak (4).
Two decades ago, Jenkins et al (5) introduced the glycemic index (GI) as an alternative system for classifying carbohydrate-containing foods. The GI ranks foods according to their effects on postprandial glucose response. In the FAO/WHO recommendation on GI measurement, both capillary and venous blood sampling are considered acceptable (6). However, a recent recommendation favors capillary blood sampling (7). It has been shown that glucose concentrations in capillary blood are greater than those in venous blood (8, 9). Wolever et al (10) found that the use of venous plasma was associated with greater within-subject variation of both glycemic responses and GIs and with nonnormal distribution of GIs. Testing the reference 3 times has been recommended (6), but the recommendation has not been systematically followed. In a recent methodologic review, a recommendation was put forward to perform the tests of the reference food at least twice (7).
The aims of the current study were to compare glycemic responses and GIs analyzed from capillary and venous blood, to compare glucose with white bread as the reference food, and to study the effect of the number of reference tests on GI values. For each setting of the tested variables, we measured the GIs of rye bread, oatmeal porridge, and instant mashed potato. The rye bread and the white bread used were identical to those analyzed previously by Leinonen et al (11), and the instant mashed potato was the same as tested in a recent interlaboratory study (10).
SUBJECTS AND METHODS
Subjects
Twelve nonsmoking volunteers [11 women and 1 man;
Postprandial study
The study subjects were served each test food (rye bread, oatmeal porridge, and instant mashed potato) once and both reference foods (glucose solution and white bread) 3 times at 1-wk intervals in a random order. At baseline, a food-frequency questionnaire (FFQ) was administered; the food composition database of the Finnish National Public Health Institute was used to calculate energy and nutrient intakes (13). These intake data together with questionnaire data on physical activity at work and during leisure time were used to compose a standardized meal (55% of total energy from carbohydrates), individualized for each subject on the basis of his or her energy expenditure, for the evening before the study day to reduce the possible second-meal effect (14, 15). The size of the meal was designed to provide 15% of the subject’s calculated daily energy expenditure. The subjects were also asked to follow their usual diet throughout the study.
The subjects were asked to avoid vigorous physical activity and not to drink alcohol during the day before the tests. They were asked to fast for 12 h after their standardized evening meal, to avoid exercise on the morning of the study, and to arrive at the clinic by car or public transportation.
In the study clinic, the subject’s weight was recorded, and an intravenous cannula was inserted into a vein in the antecubital fossa. A baseline venous blood sample was taken, and then a finger-prick capillary blood sample was taken. Next, the subject consumed the study meal within 10 min. Venous and finger-prick capillary blood samples were obtained at 15, 30, 45, 60, 90, 120, and 180 min after the start of the meal. Because the capillary sample was taken first, the actual sampling time of the venous sample was recorded to enable analyses that took into account the exact sampling times.
Study meals
Three different test foods—rye bread (Jälkiuunileipä; Oululainen Ltd, Lahti, Finland), oatmeal porridge (Elovena; Raisio Group Ltd, Raisio, Finland), and instant mashed potato (Idahoan Foods, Lewisville, ID)—were tested in the study. Oatmeal porridge (rolled oats; flake thickness 0.5–0.6 mm) and mashed potato were prepared according to package directions, except that the mashed potato was prepared with water instead of milk as in the interlaboratory study by Wolever et al (10). White bread (Ranskanleipä; Vaasan & Vaasan Ltd, Espoo, Finland) and glucose solution were used as the reference foods, and each was tested 3 times. All study meals, except instant mashed potato, were commercially available in Finland. Each of the test and reference foods was fed as a portion providing 50 g available carbohydrate. They were served with 40 g cucumber (except oatmeal porridge and glucose solution) and with a beverage of the subject’s choice, either water or noncaloric orange drink. Three subjects chose the noncaloric orange drink, and 9 subjects chose water. Each subject drank the same beverage throughout the study. The total water volume of the meals was standardized to be 500 mL by adjustment of the volume of the drink.
Chemical composition of the study meals
The chemical composition of the study meals and the evening meals was analyzed by VTT Biotechnology (VTT Technical Research Centre of Finland, Espoo, Finland). First, all samples were dried to constant weight. The protein content was estimated (nitrogen x 6.25) from quantitative analysis of nitrogen by using the Kjeldahl method (nitrogen x 5.7). The fats were measured gravimetrically by extraction in diethyl ether and petroleum ether after hydrolysis with acid (16). Total fiber and soluble and insoluble fibers were measured by using the method of Asp et al (17). Free sugars (ie, glucose, fructose, maltose, maltotriose, and sucrose) were measured by using an ion chromatograph system (Dionex, Sunnyvale, CA). Enzymatically available starch contents were analyzed by using the method of McCleary et al (18) and an assay kit (Megazyme, Bray, Ireland). The available carbohydrate was calculated as the sum of free sugars and enzymatically available starch. The nutrient composition of the study meals is shown in Table 1.
View this table:
TABLE 1. Nutrient composition and moisture of the test foods
Laboratory analysis
Capillary blood glucose was measured directly by using a glucose meter (HemoCue Glucose 201; HemoCue Ltd, Ängelholm, Sweden) that applies a modified glucose dehydrogenase method (19, 20). Results were automatically transformed to express those for plasma according to the instrumentation. A quality-control solution recommended by HemoCue was measured twice every study morning (28 d); the CV of these measurements was 1.1%. A fluoride-citrate tube was used to collect a blood sample for measurement of venous plasma glucose. The sample was centrifuged for 15 min at 4000 x g and 20 °C to separate the plasma. Plasma glucose was analyzed by a hexokinase method (Thermo Electron Ltd, Vantaa, Finland). The interassay and intraassay CVs for venous glucose measurement were 3.4% and 1.1%, respectively.
Calculations and statistical analysis
The incremental area under the curve (IAUC) was calculated by using the trapezoidal method (6). The GI was defined as the IAUC of the blood glucose response curve of a 50-g carbohydrate portion of a test meal expressed as a percentage of the response to the same amount of carbohydrate from a reference food. The GI was calculated from the 2- and 3-h incremental glucose area by using glucose or white bread as a reference (ie, GI = 100). Five subjects were each missing one glucose value (ie, 1 venous value for rye bread, 1 capillary value for glucose solution, 1 venous value for glucose solution, 1 capillary value for white bread, and 1 venous value for white bread). We provided estimates for these points by using the corresponding average glucose value of the other subjects (rye bread) or the corresponding average of the subject’s other 2 reference tests (glucose solution and white bread). The estimate was corrected by the difference between the level of the incomplete curve and the mean level of the complete curves. (The levels were estimated as the mean of the blood glucose at all time points except the missing time point.) No values were estimated, nor were IAUCs calculated for 2 white bread test visits that had >1 missing venous sample. The IAUCs for 180 min were not calculated for 3 test visits because of missing capillary and venous samples at 180 min.
To ascertain the effect on GI values of using a single test rather than a double or a triple test of the reference food, 1 or 2 IAUCs of the reference food were selected at random from the 3 IAUCs of reference foods for each subject, and the GI values were calculated by using the IAUC for the randomly selected glucose or white bread tests.
The statistical significance of the differences between values was assessed by using a paired t test and repeated-measures analysis of variance (ANOVA). P values of < 0.05 were considered significant. Data were analyzed by using SAS software (version 8.2; SAS Institute, Cary, NC).
RESULTS
Glucose solution produced a faster initial rise and a higher peak at 30 min in mean capillary blood glucose than did the other study meals (Figure 1). The capillary blood glucose dropped below the fasting concentration at 120 min after subjects ate instant mashed potato, whereas after the consumption of glucose solution, white bread, and oatmeal porridge, it fell below the fasting concentration only at 180 min. Venous peak blood glucose concentrations were lower and dropped faster than did capillary glucose concentrations so that the fasting concentration was approached or undercut as early as 90 min after the meal; the glucose concentration remained below the fasting concentration at 180 min (Figure 2). The late glucose responses after the consumption of rye bread were significantly different from the responses after the other study meals, and neither capillary nor venous blood glucose concentrations dropped below the fasting concentration at any time point. Calculation of the means of all measurements for the glucose reference showed that fasting capillary blood glucose was 0.2 ± 0.4 mmol/L higher than was fasting venous blood glucose (5.1 ± 0.4 and 4.9 ± 0.3 mmol/L, respectively; P = 0.002). The difference increased to 1.6 mmol/L at 30 min (8.5 ± 0.9 and 6.9 ± 1.0 mmol/L, respectively; P < 0.0001) and to 1.7 mmol/L at 45 min (8.1 ± 1.2 and 6.4 ± 1.3 mmol/L, respectively; P < 0.0001) and decreased gradually thereafter to the point that no difference was found at 180 min.
FIGURE 1.. Mean fasting and postprandial capillary blood glucose responses to the study meals over 180 min. The values at different time points were based on 33 blood samplings for reference foods and on 11 blood samplings for test foods. Pooled SEMs of the glucose solution, white bread, rye bread, oatmeal porridge, and mashed potato were 1.00, 0.76, 0.68, 0.66, and 0.89, respectively.
FIGURE 2.. Mean fasting and postprandial venous blood glucose responses to the study meals over 180 min. The values at different time points were based on 33 blood samplings for reference foods and on 11 blood samplings for test foods. Pooled SEMs of the glucose solution, white bread, rye bread, oatmeal porridge, and mashed potato were 0.95, 0.72, 0.57, 0.61, and 0.87, respectively.
Capillary blood samples were scheduled to be taken as precisely as possible at 15, 30, 45, 60, 90, 120, and 180 min after the start of the meal; the largest mean difference from that timing was only 0.6 min (at the 60-min time point). The average delays for venous samples were 3.0–3.7 min at different points. The IAUCs based on real time were slightly higher than those based on fixed time, but the differences were not significant (data not shown). The CVs for the IAUCs were, however, slightly lower when real time was used than when fixed time was used. We used real time when calculating the IAUCs and GIs presented in this study.
The IAUCs measured in capillary blood samples were almost twice as large as those in venous blood samples (Table 2). The average IAUCs of either of the reference foods not differ significantly, regardless of whether they were calculated from 1, 2, or 3 tests, but the CVs were lower for 2 and 3 tests than for 1 test. Both for the glucose solution and white bread, the CVs were lower when calculated from capillary blood samples than from venous blood samples. The IAUC of glucose solution was 1.3 higher than that of white bread when the reference foods were tested 2 or 3 times. The CVs of the white bread were lower than the corresponding coefficients of glucose solution. The absolute percentage differences, however, were slightly lower for glucose solution than for white bread, but the difference was significant only for the venous sample tested twice.
View this table:
TABLE 2. Incremental areas under the curve (IAUCs) for 1, 2, or 3 tests of the reference foods1
When calculating GIs, we found that the mean GIs as well as the CVs were significantly higher for venous samples than for capillary samples (Table 3). Capillary GI leveled off with 2 glucose references, although the CV diminished a little further still with 3 references. The capillary GIs did not differ significantly when white bread was tested as the reference 1, 2, or 3 times. The CVs were, however, significantly lower when 2 or 3 tests were used than when only 1 test was used. In the comparison of capillary GIs for reference foods, white bread provided GIs 1.3 times higher than did glucose solution. Extension of blood sampling to 180 min after the meal or the use of the glucose values from only 0, 30, 60, and 120 min did not provide capillary GI values significantly different from those based on all 7 sampling points (Table 4).
View this table:
TABLE 3. Glycemic indexes (GIs) for rye bread, oatmeal porridge, and mashed potato on the basis of capillary and venous blood sampling and of 1, 2, or 3 measurements of the reference foods1
View this table:
TABLE 4. Glycemic index (GIs) for rye bread, oatmeal porridge, and instant mashed potato on the basis of capillary sampling at 0, 15, 30, 45, 60, 90, and 120 min (ordinary), with an additional sample taken at 180 min (extended), or at 0, 30, 60, and 120 min only (reduced), with glucose solution as the reference, tested 3 times1
The GI of rye bread was 77, of oatmeal porridge 74, and of instant mashed potato 80 when measured from capillary blood with glucose solution tested 3 times as the reference (Table 3). The GI of white bread, tested once, was 89 when glucose solution was used as the reference (Table 5). When both white bread and glucose solution were tested 3 times, the GI of white bread was 79. The CV for test food measured 2 or 3 times was half that for food measured only once. Calculation of the capillary GIs for glucose solution with white bread as the reference, both tested 3 times, gave an estimate that white bread as the reference produced GI values 1.27 times as high as did glucose as the reference.
View this table:
TABLE 5. Capillary glycemic index (GI) for white bread with glucose solution as the reference to examine the effect of the number of measurements of the test food1
DISCUSSION
The main purpose of this study was to examine the effects of certain methodologic choices for testing GI. The results showed that capillary blood samples elicited significantly higher glucose responses than did venous blood samples and a significantly slower decline of the glucose concentration after its peak. Although the IAUC for capillary blood was almost twice that for venous blood, the CV of the IAUC was significantly lower for capillary than for venous blood. The GIs also were significantly lower when based on capillary blood than when based on venous blood, especially when white bread was used as the reference. The capillary GIs were significantly lower and had significantly smaller CVs with 2 or 3 tests of the reference food than with 1 test.
Our results regarding the difference between capillary and venous glucose concentrations are consistent with previous findings: in the fasting state, the concentrations of glucose in capillary and forearm venous blood samples are almost indistinguishable, but, after food intake and for up to 5 h thereafter, capillary blood glucose can be up to 50% higher than venous blood glucose (8, 9, 21-23). The size of the capillary-venous difference is a reflection of the efficiency of glucose uptake by the tissues of the forearm under the influence of insulin released in response to a meal. The difference has been found to be largest in persons without diabetes who have enhanced peripheral insulin sensitivity such as occurs in lean, fit, young athletes and to be smallest in insulin-resistant subjects, such as persons with diabetes (21, 24). In our study, all subjects had normal fasting plasma glucose (<6.1 mmol/L) and a normal glucose tolerance at baseline, so the differences in glycemic responses between capillary and venous samples were most likely due to glucose uptake by the forearm muscles. Because the venous samples were taken after the capillary samples, the venous sampling sometimes was done later than the targeted time point; the average delays for the venous samples were, however, modest. Therefore, the observed capillary-venous difference could not be explained by differences in sampling times.
To diminish variation in the response to the study meal, we asked subjects to control their diet and exercise on the preceding day, in keeping with recommendations in the literature (7). On the evening before each test, we served subjects a standardized evening meal containing 55% of energy as carbohydrates and adjusted to provide 15% of the subject’s daily energy. There is some evidence that the GI of an evening meal may affect glucose tolerance the next morning (14, 15, 25, 26) by causing a prolonged absorptive phase, which will favor more efficient suppression of free fatty acid absorption, thus improving insulin sensitivity (14).
The GI values of the foods were measured by using the method recommended by the Food and Agricultural Organization and the World Health Organization (6). Because within-subject variation of blood glucose response may be considerable, the World Health Organization recommends that the reference food be tested 3 times for each subject so as to obtain representative responses. Wolever et al (10) investigated how the number of tests of the reference food affected GI values. When they used 3 reference tests to calculate the GI values, the resulting mean ± SD and median values were lower than when they used only 1 reference test. The results of the current study support the view that one reference test is not enough. In their review, Brouns et al (7) recommended repeating the tests of the reference food at least once. Our results are in accord with this recommendation. Repeating measurements of the test food improves the accuracy of the GI estimate. However, repeating all measurements adds expenses, and thus repeating only the reference test food has been recommended, because it has an influence on the GI of every test food in the series (7).
The GI values of the test foods tend to be smaller when calculated from capillary blood than when calculated from venous blood. The interlaboratory study of Wolever et al (10) recommended the use of capillary blood to measure GIs because remarkable within-subject variations were found for venous blood samples. Similarly, in the current study, the variations in IAUCs were significantly larger when we used venous blood samples than when we used capillary samples.
Either glucose solution or white bread can be used as the reference food. The GI values obtained by using white bread are 1.4 times as high as those obtained by using glucose (6). In the current study, this coefficient was 1.3 when capillary blood samples were used, whereas with venous blood samples the coefficient was highly variable. These differences in values also highlight the need, in studies using white bread as the reference, to calibrate white bread against glucose (7).
Rye bread is considered to have a beneficial influence on glucose metabolism. We found that rye bread provided an initial blood glucose peak similar to that of the other study meals, but the decline in blood glucose was more moderate and did not extend below baseline at any sampling point. This resulted in a larger IAUC for rye bread; therefore the GI of rye bread, 77, was of the same magnitude as that of oatmeal porridge, 74, and instant mashed potato, 80. These findings are compatible with those of Leinonen et al (11), who used the same kind of rye bread as in the current study. They found that rye bread caused a lower insulin response, which could explain the delayed glucose decline and the resulting higher GI.
The GI for oatmeal porridge in our study was 74 for capillary blood and 95 for venous blood. It differs somewhat from the GI value for oatmeal porridge in the international table of GIs (ie, 58), which is based on the mean of 7 published observations and 1 unpublished observation (27). Methodologic differences exist between the international table and the current study, however: the GI values in the international table were calculated in part from 3-h values (28), the tests included subjects with diabetes (29-31), and only 2 studies tested up to 10 subjects (27, 32). Such methodologic discrepancies may explain the large variation between published GI values, which range from 42 to 75 (5, 28-31, 32). In a Swedish study (8), GI values were reported as the 1.5-h value (64 from venous blood and 65 from capillary blood), because of the similarity of the 1.5-h and 2-h values.
We tested the same instant mashed potato as Wolever et al (10) used in their interlaboratory study. They found that the GI varied from 86 to 98 in the 5 laboratories using capillary blood sampling and from 65 to 74 in the 2 laboratories using venous blood sampling. These results are significantly different from the results of the current study, because we found the capillary GI to be 80 and the venous GI to be 92. We do not have an explanation for the differences.
No general rule exists in statistical analysis to exclude outliers from the calculation of GI values (7). Some advice on how to handle outliers has appeared in the literature, eg, the guideline that, if an individual subject’s GI value is >2SD from the mean, this value may be excluded (33). When we used this heuristic to exclude outliers in the current study, GI values did not change essentially.
In conclusion, the findings of the current study strengthen the statement that capillary blood sampling should be used when measuring GI. Glucose or white bread as the reference food should be tested at least twice.
ACKNOWLEDGMENTS
We thank Thomas Wolever for providing the instant mashed potato sample; Karin Autio and Kirsi-Helena Liukkonen for expert help with chemical analyses of the experimental foods; Heikki Pakkala and Jukka Lauronen for assisting with the technical aspects of the study; Jaana Leiviskä, Laura Råman, Sari Salonen, and Pia Turunen for laboratory assistance; Christine Bartels for editing the English wording and grammar in the manuscript; and volunteers for their cooperation and interest.
KAH, MES, JRV, JGE, HMM, and LMV contributed to the conception and design of the study. KAH, MES, and LMV carried out the postprandial studies. JES was responsible for the analysis of the blood samples. M-LH and HKS carried out the statistical analyses. KAH wrote the first draft of the manuscript. KAH, MES, JRV, and LMV participated in the writing of the final draft of the manuscript. KAH, MES, JRV, JGE, M-LH, HMM, and LMV were responsible for the final manuscript. None of the authors had any personal or financial conflict of interest.
REFERENCES