Are the Glycemic and Insulinemic Index Values of Carbohydrate Foods Similar in Healthy Control, Hyperinsulinemic and Type 2 Diabetic Patients?

ORIGINAL ARTICLE Are the glycemic and insulinemic index values of carbohydrate foods similar in healthy control, hyperinsulinemic and type 2 diabetic patients?

X Lan-Pidhainy1 and TMS Wolever1,2

1Department of Nutritional Sciences, Faculty of Medicine, University of Toronto, Toronto, Ontario, Cannada and 2Keenan Research Center of Li Ka Shing Knowledge Institute, St Michael’s Hospital, Toronto, Ontario, Cannada

Background/Objectives: a criticism of glycemic index (GI) is that it does not indicate the insulin response of foods (insulinemic index, II). However, it is unknown if the GI and II values of foods are equivalent in all subjects, a necessary criterion for clinical utility. We compared GI and II values in non-diabetic subjects with fasting-serum-insulin (FSI) o40 pmol/l (healthy control) or with FSI X40 pmol/l (hyper[I]) and subjects with type 2 diabetes (T2DM), and to see whether GI and II were related to the serum-glucose concentrations, insulin sensitivity, b-cell function and hepatic insulin extraction (HIE) of the subjects. Subjects/Methods: Serum-glucose, -insulin and -C-peptide responses after 50 g available-carbohydrate portions of glucose (tested three times by each subject), sucrose, instant mashed-potato, white-bread, polished-rice and pearled-barley were measured in healthy control (n ¼ 9), hyper[I] (n ¼ 12) and T2DM (n ¼ 10) subjects. Results: Food GI values did not differ significantly among the three subject groups, whereas II values were higher in T2DM (100±7) than healthy controls (78±5) and hyper[I] subjects (70±5) (mean±s.e.m., P ¼ 0.05). II was inversely associated with insulin sensitivity (r ¼À0.66, Po0.0001) and positively related to fasting- and postprandial-glucose (both r ¼ 0.68, Po0.0001) and HIE (r ¼ 0.62, P ¼ 0.0002). In contrast, GI was not related to any of the biomarkers (P40.05). Conclusion: The GI is a valid property of foods because its value is similar in healthy control, hyper [I] and T2DM subjects, and is independent of subjects’ metabolic status. However, II may depend upon the glycaemic control, insulin sensitivity and HIE of the subjects. European Journal of Clinical Nutrition (2011) 65, 727–734; doi:10.1038/ejcn.2011.28; published online 23 March 2011

Keywords: insulin resistance; glycemic index; insulinemic index; type 2 diabetes

Introduction (Jenkins et al., 1983, Wolever et al., 1987, 1988), GI values tested in healthy control subjects can be applied in the The blood glucose raising potential of carbohydrate in foods nutritional management of diabetes (Atkinson, 2008). How- is numerically classified as glycemic index (GI), which is the ever, as GI has never been tested in non-diabetic subjects incremental area under the glycemic response curve (AUC) with insulin resistance/hyperinsulinemia, it is unknown elicited by a 50 g available-carbohydrate portion of a food whether GI is valid in this population. expressed as a percentage of that after 50 g glucose in the As hyperinsulinemia may have a role in the pathogenesis same subject (Jenkins et al., 1981). Many foods have been of insulin resistance and associated chronic diseases (Nilsson tested for their GI values in healthy control or diabetic et al., 2003, Takahashi et al., 2006), a major concern about subjects (Foster-Powell et al., 2002). As the GI values of the GI concept is that it does not consider concurrent insulin foods are similar in healthy control and diabetic subjects response. Thus, some investigators have begun to report insulinemic index (II) (Lee and Wolever, 1998, Miller et al., 1995, Holt et al., 1997). II is calculated similarly to GI in measuring the extent to which a food raises plasma insulin Correspondence: Dr TMS Wolever, Department of Nutritional Sciences, Faculty (Wolever et al., 1991). However, for II to be a valid property of Medicine, University of Toronto, Toronto, Ontario, Canada M5S 3E2. of a food, its value should be similar in different subjects E-mail: [email protected] Received 7 July 2010; revised 13 January 2011; accepted 7 February 2011; regardless of their degree of insulin sensitivity, b-cell published online 23 March 2011 function or glucose tolerance status. There is evidence that Effect of metabolic status on GI and II values X Lan-Pidhainy and TMS Wolever 728 relative insulin responses differ in lean and obese subjects assessment insulin sensitivity index, total and low-density with normal glucose tolerance, subjects with impaired lipoprotein cholesterol and triglycerides, and lower high- glucose tolerance and subjects with type 2 diabetes (T2DM) density lipoprotein cholesterol than those with FSI (Wolever et al., 1998). In addition, relative insulin responses o40 pmol/l (Moghaddam et al., 2006). were inversely related to subjects’ fasting insulin, suggesting Ten T2DM patients were recruited, with HbA1C of that II may be dependent on subject’s insulin sensitivity 7.3±0.3% and duration of diabetes of 7.0±1.0 years (Wolever et al., 2004). However, these studies used mixed (mean±s.e.m.). Eight patients were on metformin alone, meals rather than individual foods and did not use standard one on metformin and pioglitazone and one on metformin GI methodology; thus, their results cannot be used to draw and sulfonylurea. The patients took their usual medication conclusions about the validity of the II values of carbohy- on study days after the fasting blood sample and before drate foods. starting the test meal. None of the patients had a history of Therefore, we investigated whether the GI and II values of micro- or macro-vascular complications. a variety of carbohydrate foods are similar in healthy The protocol was designed to conform to standard GI control, hyperinsulinemic and T2DM subjects, and whether testing methodology for subjects with and without diabetes. metabolic status (insulin sensitivity, b-cell function, fasting- Subjects were instructed to maintain their usual daily routine and postprandial-glucose, hepatic insulin extraction (HIE) and food intake patterns between study days and refrain and plasma GLP-1 response) of the subjects influence the from exercise on the mornings of the test. After an overnight GI and II values. fast (10–14 h), they came to the Risk Factor Modification Center at St Michael’s hospital on eight separate mornings between 0730–0930 hours. Healthy control and hyper[I] Subjects and methods subjects had venous blood samples drawn just before and at 15, 30, 45, 60, 90 and 120 min after starting to eat. T2DM Subjects and study design had venous blood samples drawn at fasting and at 30, 60, 90, We recruited male and non-pregnant, non-lactating female 120 and 180 min after starting to eat. The protocol was subjects aged 18-70 yr (body mass index (BMI) o35 kg/m2) reviewed and approved by Research Ethics Boards at the with and without T2DM. Subjects were excluded for any of University of Toronto and St Michael’s Hospital. All subjects the following reasons: history of gastrointestinal disease or gave written informed consent. gastroparesis, liver disease (aspartate transaminase, alanine transaminase, or g-glutamyl transpeptidase42 times upper limit of normal) or kidney disease (creatinine41.2 times Test foods upper limit of normal), use of a-glucosidase or lipase Subjects were fed test meals consisting of 50 g available- inhibitors or insulin, or any acute medical or surgical event carbohydrate (total carbohydrate minus dietary fiber) as 50 g requiring hospitalization within 6 months. Subjects without anhydrous-glucose, 50 g sucrose, 71.8 g instant-mashed- diabetes had fasting glucose o7.0 mmol/l and were divided potato, 107 g white-bread, 62.5 g polished-rice and 80.6 g prospectively into those with normal fasting-serum-insulin pearled-barley. The nutrient composition of the foods was (FSI) (healthy control, n ¼ 9, FSI o40 pmol/l), or high-FSI shown in Table 1. Sugars (glucose þ sucrose) were dissolved (Hyper[I], n ¼ 12, FSI X40 pmol/l). Currently, there are no in 250 ml water. The test meals were served with a glass criteria by which an individual could be classified as insulin of water. Each subject tested glucose 3 times (first, fourth sensitive or resistant, the rationale for using FSI is that and last tests). The other five test meals were once each in FSI is strongly correlated with insulin resistance measured randomized order. by euglycemic hyperinsulinemic clamp (Laakso, 1993). The 40 pmol/l cut-off point was chosen because this represents approximately the 67th percentile for non-diabetic subjects Blood analysis in our laboratory (Wolever et al., 2004). In previous studies, Venous blood samples for glucose, insulin and C-peptide non-diabetic subjects with FSI 440 pmol/l had significantly were collected in BD Vacutainer SST tubes (BD, Franklin greater waist circumference, BMI, homeostasis model Lakes, NJ, USA). Serum glucose was measured by a glucose

Table 1 Nutrition composition of the test foods

Weight (g) Total CHO (g) Fat (g) Protein (g) Dietary fiber (g) Available CHO (g)

Oral glucose 50.0 50.0 0 0 0.0 50.0 Sucrose 50.0 50.0 0 0 0.0 50.0 Mashed potato 71.8 56.3 0.9 6.2 6.3 50.0 White bread 107 51.4 2.9 10 1.4 50.0 Rice 62.5 50.0 0 5.6 0.0 50.0 Barley 80.6 62.5 0.9 9.0 12.5 50.0

European Journal of Clinical Nutrition Effect of metabolic status on GI and II values X Lan-Pidhainy and TMS Wolever 729 oxidase method (SYNCHRON LX Systems, Beckman Coulter, Data were expressed as mean±s.e.m. for normally Brea, CA, USA), with inter-assay coefficient of variation (CV) distributed variables or median (interquartile range) for of 1.9%. Insulin was measured using one-step immunoenzy- non-normally distributed variables. Normality was assessed matic assay (Beckman Access Ultrasensitive Insulin Assay, using the Shapiro and Wilk statistic and the normality plots Beckman Coulter), with inter-assay CV of 2.5–4.3%. Insulin (PROC UNIVARIATE). Skewed variables were log-transformed has no cross-reactivity with proinsulin. C-peptide was before analysis. The values of GI, II and C-peptide index were measured using double antibody competitive radioimmuno- subjected to repeated measures analysis of variance (PROC assay (Siemens Medical Solutions Diagnostics, Los Angeles, MIXED) to test for the main effects of food and subject-group CA, USA), with inter-assay precision of 10% or less. and the food Â subject-group interaction. Age, BMI and waist Venous blood samples for GLP-1 were collected in BD circumference were included in the model as covariates to Vacutainer EDTA tubes (BD). The dipeptidyl peptidase control their different variables among subjects. Tukey’s (DPP) IV inhibitor (Linco Research, St Charles, MO, USA) post hoc test was performed to compare individual means if was added immediately after collection. Plasma GLP-1 was the main effects or interactions were statistically significant. measured by capturing the active GLP-1 from the sample by The correlations between GI, II and the metabolic indices a monoclonal antibody (specific binding to the N-terminal (insulin sensitivity, b-cell function, HIE, severity of glycemia, region) using GLP-1 (active) ELISA Kit (Linco Research), with and GLP-1 response) were determined by simple linear inter-assay CV ranges from 1–13%. All the samples were regressions. Step-wise multiple regression analysis was used stored at À70 1C before analysis. to examine the extent to which the different variables accounted for the variability of GI or II. When GI is the dependent variable, the independent variables are age, BMI, OGIS, disposition index, HIE, and the AUC of glucose and Calculations and statistical analysis GLP-1. When II is the dependent variable, the independent The sample size was determined based on previous studies variables are GI, age, BMI, OGIS, disposition index, HIE, and (Wolever et al., 1998, Wolever, 2003), with a power of 80% the AUC of glucose and GLP-1. Collinearity was determined and Po0.05. The AUC for glucose, insulin, C-peptide and by including variance inflation factor in the model, with GLP-1 were calculated using the trapezoid rule (Wolever variance inflation factor of 5 or 10 and above indicating a et al., 1991). Hepatic insulin extraction (HIEauc) was multicollinearity problem (O’Brien, 2007). No collinearity determined by the AUC of C-peptide divided by that of was apparent for the variables included in the regression insulin (Polonsky and Rubenstein, 1984). GI was calculated analysis. All analyses were done using SAS 9.2, (SAS Institute as the AUC of the test food expressed as a percentage of the Inc., Cary, NC, USA). Differences were considered significant mean AUC of three tests of oral-glucose in the same subject if two-tailed Po0.05. (Jenkins et al., 1981); the mean of the resulting values was the GI of the food. II was calculated similarly to GI. C-peptide is co-secreted with insulin in equimolar amounts, but is not subjected to hepatic insulin clearance, which Results varies considerably; thus, C-peptide is regarded as a much better estimate of insulin secretion than levels of insulin T2DM subjects were significantly older and had higher BMI itself (Polonsky and Rubenstein, 1984). Therefore, C-peptide and waist circumference than both healthy control and index was also calculated using the same method for GI and hyper[I] subjects. Fasting glucose and postprandial glucose II calculation. responses (glucose AUC) were not significantly different The mean glucose and insulin values of the 3 oral-glucose- between healthy control and hyper[I] subjects but were tests were used to calculate oral-glucose-insulin-sensitivity significantly higher in T2DM. The intra-individual coeffi- (OGIS) index, which was constructed on established cient of variations (CV) of blood glucose AUC for repeated principles of glucose kinetics and insulin action and was tests of oral glucose for healthy control, hyper[I] and T2DM validated against the euglycemic-hyperinsulinemic clamp in subjects are 24±5, 25±3 and 17±3 (mean±s.e.m.), respec- healthy, obese and T2DM subjects (Mari et al., 2001). tively. Fasting insulin was significantly higher in hyper[I] The b-cell compensation for insulin resistance was esti- and T2DM than healthy control subjects. Fasting C-peptide mated using the insulin secretion/insulin resistance (disposi- increased in a step-wise manner from healthy control to tion) index derived from OGTT (Abdul-Ghani et al., 2007), hyper[I] to T2DM (Table 2). HIE was significantly higher in which was shown to be the best predictor of future T2DM than hyper[I] subjects. OGIS and disposition index development of T2DM in subjects with normal glucose decreased step-wise from healthy control to hyper[I] to tolerance compared with other predictive models such as T2DM (Table 2). Systolic- and diastolic-blood pressure, total San Antonio Diabetes Prediction Model (including age, cholesterol, triglyceride, total :high-density lipoprotein sex, ethnicity, BMI, blood pressure, fasting-plasma-glucose, cholesterol were similar in healthy control and hyper[I] triglycerides and high-density lipoprotein) and 2-hr plasma subjects, but were significantly higher in T2DM. Fasting- and glucose concentration (Stern et al., 2002). postprandial-GLP-1, high-density lipoprotein, low-density

European Journal of Clinical Nutrition Effect of metabolic status on GI and II values X Lan-Pidhainy and TMS Wolever 730 lipoprotein and C-reactive protein were not significantly (P ¼ 0.26); thus, the GI values were not significantly different different among the three subject groups (Table 2). among healthy control, hyper[I] and T2DM for any food For GI values, there was neither significant subject-group individually, nor for the mean of all carbohydrate foods effects (P ¼ 0.20) nor food Â subject-group interactions (Table 3). For II values, there was a tendency of significant

Table 2 Anthropometric and metabolic characteristics of the study groupsa

Control (n ¼ 9) Hyper[I] (n ¼ 12) T2DM (n ¼ 10) Pb

Age (year) 24 (23–29)a,c 26 (23–37)a 57 (53–58)b o0.0001 BMI (kg/m2) 22 (21–22)a 25 (22–30)b 32 (29–34)c o0.0001 Waist circumference (cm) 72 (70–76)a 88 (77–95)b 102 (94–115)c o0.0001 Fasting glucose (mmol/l) 4.6 (4.5–4.7)a 4.9 (4.7–5.2)a 9.8 (7.2–11.3)b o0.0001 Glucose AUC (mmol Â min/l)d 198 (138–291)a 247 (186–278)a 831 (699–978)b 0.0002 Fasting insulin (pmol/l) 29 (24–31)a 51 (47–57)b 51 (42–63)b 0.0002 Fasting C-peptide (pmol/l) 231 (213–241)a 412 (345–445)b 784 (711–878)c o0.0001 Fasting GLP-1 (pmol/l) 2.5 (2.2–6.1) 3.6 (2.3–5.1) 4.1 (3.5–4.5) 0.45 2hr-postprandial GLP-1 (pmol Â min/l)d 211 (94–340) 183 (113–294) 281 (133–420) 0.64 Hepatic insulin extractiond 3.8 (3.3–6.4)a,b 3.4 (2.7–3.8)a 6.8 (5.7–10.4)b 0.002 Oral glucose insulin sensitivityd 497 (478–521)a 459 (424–485)b 280 (259–362)c 0.001 Disposition indexd 63±5a,e 38±4b 29±4c 0.0008 Systolic BP (mm Hg) 101±2.4a 109±2.2a 130±4.1b o0.0001 Diastolic BP (mm Hg) 61±2.2a 67±2.2a 79±2.9b 0.0001 Total cholesterol (mmol/l) 4.1±0.3a 4.4±0.3a,b 5.5±0.5b 0.02 Triglycerides (mmol/l) 0.7 (0.5–0.9)a 0.9 (0.8–1.1)a 1.5 (1.2–1.9)b 0.02 HDL-C (mmol/l) 1.4±0.1 1.2±0.1 1.2±0.1 0.41 LDL-C (mmol/l) 2.4±0.3 2.7±0.2 3.1±0.3 0.27 Total:HDL cholesterol 2.8 (2.7–3.8)a 3.6 (3.2–3.9)a,b 3.9 (3.3–6.6)b 0.02 C-reactive protein (mg/l) 0.7 (0.2–1.2) 1.5 (0.5–3.8) 3.7 (1.1–5.4) 0.10

Abbreviations: AUC, incremental area under the curve; BMI, body mass index; BP, blood pressure; HDL, high-density lipoprotein; LDL, low-density lipoprotein. aHealthy control, with fasting serum insulin o40 pmol/l; hyper[I], non-diabetic subjects with FSIX40 pmol/l; T2DM, subjects with type 2 diabetes; BP; HDL; and LDL. bP represents overall significant differences across groups. P-values were derived from one-way analysis of variance except those for 2-hr postprandial GLP-1 response, hepatic insulin extraction, oral glucose insulin sensitivity and disposition index. cMedian; interquartile range in parentheses (all such values). Values in the same row with different superscript letters differ significantly, Po0.05 (Tukey’s post hoc test). dCalculated by using the mean of three oral-glucose-tolerance test results. PROC MIXED repeated measures analysis of variance, adjusted for age, BMI and WC. The units for oral glucose insulin sensitivity and disposition index are ml Â minÀ1 Â mÀ2 and (mmol)À2 Â kgÀ1 Â minÀ1 Â l, respectively. Hepatic insulin extraction does not have units. eMean±s.e.m. (all such values).

Table 3 The GI, II and C-peptide index of different carbohydrate foods in the study groupsa

Subjectsa Glucose Sucrose Potato Bread Rice Barley Meanb

Glycemic index (%) Control 100±068±9 101±763±11 61±758±670±4 Hyper[I] 100±069±681±10 70±672±744±667±3 T2DM 100±068±598±571±670±10 47±570±4 Pc 0.95 0.20 0.13 0.10 0.55 0.20

Insulinemic index (%) Control 100±070±8 128±10 86±9a,b 60±646±578±5 Hyper[I] 100±074±8 114±10 74±7a 56±735±470±5 T2DM 100±081±10 137±12 138±13b 73±770±16 100±7 P 0.85 0.28 0.01 0.81 0.31 0.05

C-peptide index (%) Control 100±065±11 142±18 85±10 45±11 47±777±7 Hyper[I] 100±074±996±12 77±877±10 44±674±4 T2DM 100±083±9 123±14 113±11 72±857±690±6 P 0.44 0.07 0.14 0.24 0.70 0.40

Abbreviation: GI, glycemic index; II, insulinemic index; T2DM, type 2 diabetes. aHealthy control with fasting serum insulin o40 pmol/l (n ¼ 9); hyper[I], non-diabetic subjects with fasting serum insulinX40 pmol/l (n ¼ 12); T2DM, subjects with type 2 diabetes (n ¼ 10). Values (mean±s.e.m.) in the same column with different superscript letters are significantly different, Po0.05 (Tukey’s post hoc test). bRefers to the mean of all carbohydrate foods. cP-values refer to overall significant differences among subject groups, and were derived from repeated measures of analysis of variance (PROC MIXED, Tukey’s post hoc test). Age, body mass index and waist circumference were included in the model as covariates.

European Journal of Clinical Nutrition Effect of metabolic status on GI and II values X Lan-Pidhainy and TMS Wolever 731 subject-group effects (P ¼ 0.05) and food Â subject-group regression analysis using GI as dependent variable and age, interaction (P ¼ 0.09); thus the II values of white-bread was BMI, OGIS, disposition index, HIE and AUC of glucose significantly higher in T2DM than healthy control and and GLP-1 as independent variables found that none of hyper[I] subjects (P ¼ 0.01) and the mean II value of all the variables met the 0.05 significance level for entry into carbohydrate foods was much higher in T2DM than healthy the model. control and hyper[I] subjects (P ¼ 0.05) (Table 3). The values II was inversely associated with OGIS (r ¼À0.66, of C-peptide index were not significantly different among Po0.0001) and positively related to HIE (r ¼ 0.62, healthy control, hyper[I] and T2DM for any food individu- P ¼ 0.0002). GI was not related to either OGIS (r ¼À0.07, ally, nor for the mean of all carbohydrate foods (Table 3). P ¼ 0.73) or HIE (r ¼ 0.09, P ¼ 0.65) (Figure 1). C-peptide GI was not related to any of the anthropometric (age and index was not significantly related to OGIS (r ¼À0.17, BMI) or metabolic indices (OGIS, HIE, disposition index and P ¼ 0.37) and HIE (r ¼ 0.35, P ¼ 0.057). The GI, II and GLP-1 response) for each food individually, nor for the mean C-peptide index were not related to either GLP-1 response of all carbohydrate foods (P40.05) (Figure 1). Multiple or disposition index (P40.05). Multiple regression analysis showed that OGIS and HIE together predicted II B Control FSI < 40 pmol/L (II ¼ 106À0.09 Â OGIS þ 2.75 Â HIE), and explained 51% 2 Hyper[I] FSI ≥ 40pomol/L of the variation in II (r ¼ 0.51, Po0.0001). In particular, 2 T2DM OGIS alone explained 43% of the variation in II (r ¼ 0.43, Po0.001). 150 GI was not related to any of the markers of the severity of r=-0.07 r=-0.09 glycemia (P40.05) (Figure 2) whereas II was positively p=0.73 p=0.65 associated with fasting- and mean-postprandial-glucose 100 (both r ¼ 0.68, Po0.0001) and glucose AUC (r ¼ 0.46, P ¼ 0.009).

Glycemic Index (GI) 50 Discussion

This study confirms that the GI values of carbohydrate foods 200 r = -0.66 r = 0.62 are similar in all subjects regardless of the severity of p<0.0001 p=0.0002 glycemia or degree of insulin sensitivity. This shows that 150 GI is a property of foods and affirms its clinical utility in a broad population. However, the II values of carbohydrate 100 foods were inversely associated with insulin sensitivity and positively related to the severity of glycemia and HIE, 50 suggesting that II is not solely a property of foods but also

Insulinemic Index (II) depends on the metabolic status of the subjects. 0 150 We found that between-subject variation of GI values was not explained by any demographic, anthropometric or metabolic variables. Previous studies have shown that GI 100 values are similar in healthy control vs T2DM (Jenkins et al., 1983), individuals with type 1 diabetes (T1DM) vs individuals with T2DM (Wolever et al., 1987), adults vs children 50 with T1DM (Wolever et al., 1988), T2DM on oral agents vs C-peptide Index r = -0.17 r = 0.35 insulin (Jenkins et al., 1986) and T2DM in good vs poor p=0.37 p=0.057 metabolic control (Wolever et al., 1986). We showed here 0 200 300 400 500 600 0 5 10 15 that the GI values of foods in hyperinsulinemic subjects were similar to those in healthy control and T2DM subjects. OGIS HIE (ml/min/m2) Although not unexpected, this is important because GI may be particularly useful for obese and/or insulin resistant Figure 1 The linear correlation between GI (top) or II (middle) subjects to assist with weight management (Ebbeling et al., or C-peptide index (bottom) and oral glucose insulin sensitivity (OGIS) (left) and hepatic insulin extraction (HIE) (right) in all subject 2007) and/or the prevention of T2DM (Barclay et al., 2008) –groups (healthy control, FSIo40 pmol/l, white circle; hyper[I], and stroke (Oh et al., 2005). Thus, it is valid to utilize the GI FSI X40 pmol/l, gray circle; and T2DM, black circle). r: Pearson’s of foods tested in healthy control subjects in the dietary correlation coefficient. The values of GI, II and C-peptide index management of hyperinsulinemic/insulin resistant subjects. are the mean of five carbohydrate foods for each subject (n ¼ 31). OGIS and HIE are calculated from the mean of three oral- The GI values of starchy carbohydrate foods depend on glucose-tests. Lines are regression line. differences in their relative rates of digestion and absorption

European Journal of Clinical Nutrition Effect of metabolic status on GI and II values X Lan-Pidhainy and TMS Wolever 732 Control FSI < 40 pmol/L Hyper[I] FSI ≥ 40pomol/L T2DM

150 r=-0.10 r=0.0003 r=0.19 p=0.58 p=0.99 p=0.31

100

50 Glycemic Index (GI)

0 150

100

50 r=0.68 r=0.68 r=0.46

Insulinemic Index (II) p<0.0001 p<0.0001 p=0.009

0 Fasting Glucose Mean Postprandial Glucose Glucose AUC (mmol/L) (mmol/L) (pmol× min/L) Figure 2 The linear correlation between GI (top) or II (bottom) and fasting glucose (left) and mean postprandial glucose (middle) and glucose AUC (right) in all subject groups (healthy control, FSI o40 pmol/l, white circle; hyper[I], FSI X40 pmol/l, gray circle; and T2DM, black circle). r: Pearson’s correlation coefficient. The values of GI and II are the mean of five carbohydrate foods for each subject (n ¼ 31). Fasting glucose, mean postprandial glucose and glucose AUC are calculated from the mean of three oral-glucose tests. The lines are regression line.

(that is, the rate of glucose appearance from the gut), carbohydrate foods vary depending on the metabolic status, (Wolever et al., 1991) which, presumably, do not differ in thus limiting its clinical utility. healthy, hyperinsulinaemic and diabetic subjects. We There are physiological basis for the observed higher mean included sucrose as a test meal because its glycemic response II values of the five carbohydrate foods for T2DM than depends, at least in part, on the hepatic metabolism of healthy control and hyperinsulinemic subjects. Glucose is fructose, which, in turn, may depend on insulin sensitivity. not the only stimulus for insulin secretion, gastrointestinal We previously showed that the GI of fruit leather, over 50% hormones, mainly GIP and GLP-1 are known to potentiate of the available-carbohydrate of which consisted of fructose, the stimulatory effect of glucose and mediate postprandial was inversely related to fasting insulin and to waist insulin secretion (Meier and Nauck, 2005, Drucker and circumference (Wolever et al., 2009). However, the present Nauck, 2006). In addition, the activity of the entero-insulin results are not consistent with this, in that the mean GI of axis and HIE all have a role in modulating postprandial sucrose in hyper[I] subjects was, if anything, slightly higher insulin concentration. We found that OGIS and b-cell than in the healthy control. function decreased from healthy control to hyperinsuline- One of the major criticisms of GI is that it does not mic to T2DM and HIE was much higher in T2DM than take into account the concurrent insulin response (Xavier healthy control and hyperinsulinemic subjects. These meta- Pi-Sunyer, 2002); thus, some researchers have advocated bolic aberrations, especially reduced b-cell function (less using II in the treatment of diabetes (Holt et al., 1997, Bao insulin secretion) and increased HIE may reduce plasma et al., 2009); however, for II to have clinical utility, it must be insulin response to oral glucose (the reference meal) in applicable to a broader population and be similar in all T2DM patients; thus other the component of foods (the test subject groups regardless of their degree of insulin sensitivity meals) on insulin secretion becomes more prominent, which and glucose tolerance status. We found that the mean II resulted in increased II in T2DM. values of all five carbohydrate foods were higher in T2DM Though the mean II value of the carbohydrate foods was than healthy control and hyperinsulinemic subjects higher in T2DM than healthy control and hyperinsulinemic (P ¼ 0.05). Furthermore, II values were inversely related subjects, for each individual food, only in white-bread a to OGIS and positively related to HIE and the severity of significant difference in II was observed. How can the fact glycemia. In addition, OGIS alone explained 43% that II was correlated with metabolic status (OGIS and of the variation in II. These results suggest that the II of the severity of glycemia) be reconciled with the lack of

European Journal of Clinical Nutrition Effect of metabolic status on GI and II values X Lan-Pidhainy and TMS Wolever 733 significant difference in II among the subject groups for Barclay AW, Petocz P, McMillan-Price J, Flood VM, Prvan T, Mitchell P each individual food (except white-bread)? This may be et al. (2008). Glycemic index, glycemic load, and chronic disease risk— Am J Clin Nutr because II, OGIS and markers of the severity of glycemia are a metaanalysis of observational studies. 87, 627–637. Drucker DJ, Nauck MA (2006). The incretin system: glucagon-like continuous variables, whereas the subject groups (healthy peptide-1 receptor agonists and dipeptidyl peptidase-4 inhibitors control, hyperinsulinemic and T2DM) are categorical vari- in type 2 diabetes. Lancet 368, 1696–1705. ables. Continuous variables (regression analysis) has greater Ebbeling CB, Leidig MM, Feldman HA, Lovesky MM, Ludwig DS statistical power than categorical variables (analysis of (2007). Effects of a low-glycemic load vs low-fat diet in obese young adults: a randomized trial. JAMA 297, 2092–2102. variance) due to increased precision, a simpler and more Foster-Powell K, Holt SHA, Brand-Miller JC (2002). International informative interpretation of the results, and greater parsi- table of gylcemic index and glycemic load values: 2002. Am J Clin mony (Lazic, 2008); therefore, categorical variables analysis Nutr 76,5–56. runs the risk of missing significant effects. Holt SHA, Brand Miller JC, Petocz P (1997). An insulin index of foods: the insulin demand generated by 1000-kJ portions of It is concluded that the GI of carbohydrate foods is not common foods. Am J Clin Nutr 66, 1264–1276. significantly different among healthy control, hyperinsuli- Jenkins DJA, Wolever TMS, Jenkins AL (1986). Low glycemic nemic and T2DM patients, and the GI is not influenced by response to traditionally processed wheat and rye products: Bulgur subjects’ metabolic status. This finding supports the clinical and pumpernickel bread. Am J Clin Nutr 43, 516–520. Jenkins DJA, Wolever TMS, Jenkins AL (1983). The glycaemic index utility of GI in the prevention and management of diabetes. of foods tested in diabetic patients: a new basis for carbohydrate On the contrary, II values were inversely related to OGIS and exchange favouring the use of legumes. Diabetologia 24, 257–264. positively associated with the severity of glycemia and HIE, Jenkins DJA, Wolever TMS, Taylor RH (1981). Glycemic index of suggesting that II is not a property of food but is subject foods: a physiological basis for carbohydrate exchange. Am J Clin Nutr 34, 362–366. dependent. The results suggest that II may have limited Laakso M (1993). How good a marker is insulin level for insulin clinical utility. resistance? Am J Epidemiol 137, 959–965. Lazic SE (2008). Why we should use simpler models if the data allow this: relevance for ANOVA designs in experimental biology. BMC physiol 8, 16. Conflict of interest Lee BM, Wolever TMS (1998). Effect of glucose, sucrose and fructose on plasma glucose and insulin responses in normal humans: XLP declares no conflict of interest. TMSW is the president comparison with white bread. Eur J Clin Nutr 52, 924–928. and part owner of Glycemic Index Laboratories Inc. and Mari A, Pacini G, Murphy E, Ludvik B, Nolan JJ (2001). A model- based method for assessing insulin sensitivity from the oral Glycemic Index Testing Inc. glucose tolerance test. Diabetes Care 24, 539–548. Meier JJ, Nauck MA (2005). Glucagon-like peptide 1(GLP-1) in biology and pathology. Diabetes Metab Res 21, 91–117. Miller JB, Pang E, Broomhead L (1995). The glycaemic index of foods Acknowledgements containing sugars: comparison of foods with naturally-occurring v. Added sugars. Br J Nutr 73, 613–623. XLP designed and ran the study, collected and analyzed data, Moghaddam E, Vogt JA, Wolever TMS (2006). The effects of fat and and wrote the manuscript. TMSW designed and secured protein on glycemic responses in nondiabetic humans vary with waist circumference, fasting plasma insulin, and dietary fiber funding for the study and wrote the manuscript. XLP was intake. JNutr136, 2506–2511. supported by St Michael’s Hospital-Li Ka Shing Knowledge Nilsson P, Nilsson J, Hedblad B, Eriksson K, Berglund G (2003). Institute Scholarship and Ontario Graduate Scholarship Hyperinsulinaemia as long-term predictor of death and ischaemic (OGS). The study was supported by Canadian Institute heart disease in nondiabetic men: The Malmo¨ Preventive Project. J Intern Med 253, 136–145. of Health Research (CIHR) operating grant MOP-79382. O’Brien RM (2007). A caution regarding rules of thumb for variance We thank Melissa Kwong for GLP-1 measurements and inflation factors. Qual Quant 41, 673–690. Cindy Huang, Michelle Liu for assistance with the running Oh K, Hu FB, Cho E, Rexrode KM, Stampfer MJ, Manson JE et al. of the study and data entry. The study was presented in (2005). Carbohydrate intake, glycemic index, glycemic load, and dietary fiber in relation to risk of stroke in women. poster form at the 28th International Symposium on Am J Epidemiol 161, 161–169. Diabetes and Nutrition of the Diabetes and Nutrition Study Polonsky KS, Rubenstein AH (1984). C-peptide as a measure of the Group (DNSG) of the EASD, 1–4 July, 2010, Oslo, Norway. secretion and hepatic extraction of insulin. Pitfalls and limita- tions. Diabetes 33, 486–494. Stern MP, Williams K, Haffner SM (2002). Identification of persons at high risk for type 2 diabetes mellitus: do we need the oral glucose References tolerance test? Ann Intern Med 136, 575–581. Takahashi F, Hasebe N, Kawashima E, Takehara N, Aizawa Y, Abdul-Ghani MA, Williams K, DeFronzo RA, Stern M (2007). What is Akasaka K et al. (2006). Hyperinsulinemia is an independent the best predictor of future type 2 diabetes? Diabetes Care 30, predictor for complex atherosclerotic lesion of thoracic aorta in 1544–1548. non-diabetic patients. Atherosclerosis 187, 336–342. Atkinson FS (2008). International tables of glycemic index and Wolever TMS (2003). Determination of the glycaemic index of foods: glycemic load values. Diabetes Care 31, 2281–2283. Interlaboratory study. Eur J Clin Nutr 57, 475. Bao J, De Jong V, Atkinson F, Petocz P, Brand-Miller JC (2009). WoleverTMS,JenkinsAL,VuksanV,CampbellJ(2009).Theglycaemic Food insulin index: physiologic basis for predicting insulin index values of foods containing fructose are affected by metabolic demand evoked by composite meals. Am J Clin Nutr 90, 986–992. differences between subjects. Eur J Clin Nutr 63, 1106–1114.

European Journal of Clinical Nutrition Effect of metabolic status on GI and II values X Lan-Pidhainy and TMS Wolever 734 Wolever TMS, Campbell JE, Geleva D, Anderson GH (2004). High- Wolever TMS, Jenkins DJA, Jenkins AL, Josse RG (1991). The glycemic fiber cereal reduces postprandial insulin responses in hyperinsu- index: methodology and clinical implications. Am J Clin Nutr 54, linemic but not normoinsulinemic subjects. Diabetes Care 27, 846–854. 1281–1285. Wolever TMS, Jenkins DJA, Josse RG, Wong GS, Lee R (1987). The Wolever TMS, Chiasson J, Csima A, Hunt JA, Palmason C, glycemic index: similarity of values derived in insulin-dependent Ross SA et al. (1998). Variation of postprandial plasma glucose, and non-insulin-dependent diabetic patients. J Am Coll Nutr 6, palatability, and symptoms associated with a standardized 295–305. mixed test meal versus 75 g oral glucose. Diabetes Care 21, Wolever TMS, Jenkins DJA, Kalmusky J (1986). Comparison of 336–340. regular and parboiled rices: explanation of discrepancies between Wolever TMS, Jenkins DJA, Collier GR, Ehrlich RM, Josse RG, reported glycemic responses to rice. Nutr Res 6, 349–357. Wong GS et al. (1988). The glycaemic index: effect of age in Xavier Pi-Sunyer F (2002). Glycemic index and disease. Am J Clin Nutr insulin dependent diabetes mellitus. Diabetes Res 7, 71–74. 76, 290S–298S.

European Journal of Clinical Nutrition