Context Clinical trials using antihyperglycemic medications to improve glycemic control have not demonstrated the anticipated cardiovascular benefits. Low–glycemic index diets may improve both glycemic control and cardiovascular risk factors for patients with type 2 diabetes but debate over their effectiveness continues due to trial limitations.
Objective To test the effects of low–glycemic index diets on glycemic control and cardiovascular risk factors in patients with type 2 diabetes.
Design, Setting, and Participants A randomized, parallel study design at a Canadian university hospital research center of 210 participants with type 2 diabetes treated with antihyperglycemic medications who were recruited by newspaper advertisement and randomly assigned to receive 1 of 2 diet treatments each for 6 months between September 16, 2004, and May 22, 2007.
Intervention High–cereal fiber or low–glycemic index dietary advice.
Main Outcome Measures Absolute change in glycated hemoglobin A1c (HbA1c), with fasting blood glucose and cardiovascular disease risk factors as secondary measures.
Results In the intention-to-treat analysis, HbA1c decreased by −0.18% absolute HbA1c units (95% confidence interval [CI], −0.29% to −0.07%) in the high–cereal fiber diet compared with −0.50% absolute HbA1c units (95% CI, −0.61% to −0.39%) in the low–glycemic index diet (P < .001). There was also an increase of high-density lipoprotein cholesterol in the low–glycemic index diet by 1.7 mg/dL (95% CI, 0.8-2.6 mg/dL) compared with a decrease of high-density lipoprotein cholesterol by −0.2 mg/dL (95% CI, −0.9 to 0.5 mg/dL) in the high–cereal fiber diet (P = .005). The reduction in dietary glycemic index related positively to the reduction in HbA1c concentration (r = 0.35, P < .001) and negatively to the increase in high-density lipoprotein cholesterol (r = −0.19, P = .009).
Conclusion In patients with type 2 diabetes, 6-month treatment with a low–glycemic index diet resulted in moderately lower HbA1c levels compared with a high–cereal fiber diet.
Trial Registration clinicaltrials.gov identifier: NCT00438698
The need for implementation of effective dietary strategies in diabetes prevention and management has been emphasized by the success of diet and lifestyle changes in preventing diabetes in high-risk patients.1 There is also concern that use of antihyperglycemic medications to improve glycemic control in type 2 diabetes may not always significantly improve cardiovascular outcomes.2-7
One dietary strategy aimed at improving both diabetes control and cardiovascular risk factors is the use of low–glycemic index diets.8-10 These diets have been reported to benefit the control of diabetes11; increase high-density lipoprotein cholesterol (HDL-C)12,13; lower serum triglyceride, plasminogen activator inhibitor 1, and high-sensitivity C-reactive protein (CRP) concentrations14-16; and reduce diabetes incidence8,9 and overall cardiovascular events.10 Use of the α-glucosidase carbohydrate absorption inhibitor acarbose, which effectively creates a low–glycemic index diet by slowing the rate of carbohydrate absorption, similarly reduced not only the rate of progression to diabetes in high-risk individuals but also the incidence of hypertension and the risk of cardiovascular disease.17,18 Nonetheless, the relevance and practicality of applying the low–glycemic index dietary approach to the treatment of diabetes has been questioned.19-21 Furthermore, although meta-analyses of clinical studies have indicated a benefit in diabetes in terms of reduced glycated proteins,11 individual trials have often failed to demonstrate a clear benefit for low–glycemic index diets.11,22 However, these trials have usually been of short duration with relatively small numbers of participants.11 An exception is a recent larger and longer trial,22 but in that study, the mean baseline glycated hemoglobin A1c (HbA1c) concentration was already relatively low at 6.1%, thus making it difficult to demonstrate benefit from the intervention.
Our goal in this study was to assess the effect of a low–glycemic index diet in an adequately powered study of patients with type 2 diabetes controlled by oral medications with HbA1c concentrations between 6.5% and 8.0%. At these levels, a reduction in glycemia and associated risk factors for diabetes complications are likely to be observed more clearly. We selected a high–cereal fiber diet treatment for its suggested health benefits8,9,23-25 for the comparison so that the potential value of carbohydrate foods could be emphasized equally for both high–cereal fiber and low–glycemic index interventions.
Study participants were recruited from local newspaper advertisements. Out of 2220 responses by telephone, 981 participants were considered potentially eligible and were invited to attend an information session at the Risk Factor Modification Center, St Michael's Hospital, Toronto, Ontario, Canada, where all study clinical activity took place. Of those participants invited, 658 attended a screening appointment. A total of 389 participants were ineligible and 269 were eligible; however, 6 participants could not be contacted, 5 were unable to start the study immediately, and 48 declined to continue. The remaining 210 participants were randomized (Figure 1). Recruitment took place between July 8, 2004, and December 5, 2006, with the last follow-up visit on May 22, 2007. Eligible participants were men or postmenopausal women with type 2 diabetes who were taking oral medications other than acarbose to control their diabetes, with medications stable for the previous 3 months, and who had HbA1c values at screening between 6.5% and 8.0% (to convert to proportion of total hemoglobin, multiply by 0.01). None had clinically significant cardiovascular, renal, or liver disease (alanine aminotransferase >3 times the upper limit of normal) and none were undergoing treatment for cancer. Participants were accepted after surgery or myocardial infarction providing an event-free, 6-month period had elapsed before the study.
After randomization, it was found that 5 participants had not been excluded who were taking acarbose and had not subsequently been advised to stop taking acarbose, and 2 participants were not taking diabetes medications. In addition, 8 participants had changed their medications within 3 months before the start of the study and 2 women were not postmenopausal. Furthermore, at the start of the trial, 9 participants (6 in the low–glycemic index diet and 3 in the high–cereal fiber diet) did not receive the appropriate dietary advice for the treatment to which they had been randomized. This error was corrected within the first 4 weeks of the study. Nevertheless, all randomized participants were retained both for the intention-to-treat (ITT) and study completion analyses.
Race/ethnicity was determined by a member of the research team (S.M.) and included to allow for future subgroup analyses when genetic data (to be determined) become available. The study was approved by the research ethics board of St Michael's Hospital and the University of Toronto, Toronto, Ontario, Canada, and written consent was obtained from all participants.
Our study was a randomized, parallel study with 2 treatments, a low–glycemic index diet and a high–cereal fiber diet, each of 6 months’ duration. After stratification by sex and HbA1c (≤7.1%), randomization was performed by using participant identification (subject ID) by a statistician (E.V.) who was geographically separated from the center at which participants were observed. Neither the dietitians who were responsible for the day-to-day running of the study nor the participants could be blinded to the treatment allocation. The technical staff involved in the analyses were blinded to treatment, as was the statistician up to and during the preliminary assessment of the primary outcome, HbA1c. In addition, during the study, equally strong emphasis was placed by dietitians on the potential importance of either high–cereal fiber foods or low–glycemic index foods and appropriate weekly checklists were developed for each treatment.
Participants were observed at the Clinical Nutrition and Risk Factor Modification Center at baseline, weeks 2 and 4, and thereafter at monthly intervals until the end of the 6-month period. During the first month, participants received instructions regarding the diet to which they were allocated. Throughout the study, this advice was reinforced by the dietitians. At all center visits, participants were weighed in indoor clothing without shoes and a fasting blood sample was taken. Blood pressure was measured seated on 3 occasions at 1-minute intervals using an Omron automatic sphygmomanometer (OMRON Healthcare Inc, Burlington, Ontario, Canada) and the mean of the 3 measurements was taken. In addition, participants brought with them their 7-day food record covering the week before the visit; this was discussed with the dietitian together with their checklist of low–glycemic index or high–cereal fiber food items recorded on a daily basis throughout the study. Participants noted their overall feeling of satiety or hunger over the previous weeks on a scale of +4 to −4, where +4 was extremely satiated, 0 was neutral, and −4 was extremely hungry. For the last 114 participants who were enrolled in the study, adherence was rated formally under 1 of 3 categories (good, adequate but may need additional encouragement with a between-visit telephone call, or nonadherent with definite need for a telephone call between scheduled visits).
During the study, participants were asked to maintain their antihyperglycemic medications constant and letters were sent to family physicians for their support in this matter. When patients experienced symptoms of hypoglycemia with blood capillary glucose levels of less than 63.1 mg/dL (to convert to millimoles per liter, multiply by 0.0555) for no obvious reason, study endocrinologists (R.G.J., J.S., G.L.B., and L.A.L.) who were blinded to treatment were consulted. The patients were then referred back to their family physicians so that medications could be reduced according to a predetermined protocol. If HbA1c increased to more than 8.5% on 2 successive occasions, participants were withdrawn from the study and referred back to their own physicians. Only 1 participant was withdrawn due to HbA1c being more than 8.5% on 2 successive occasions. The participant's HbA1c was 7.6% at recruitment, but increased to 8.2% by baseline and was later withdrawn due to readings of more than 8.5% during the study.
General dietary advice conformed to the National Cholesterol Education Program Adult Treatment Panel III26 and the American Diabetes Association27 guidelines to reduce saturated fat and cholesterol intakes. Most of the participants were overweight (179/210 [85.2%], with body mass index [BMI, calculated as weight in kilograms divided by height in meters squared] of ≥25) or obese (113/210 [53.8%], BMI ≥30) and wished to lose weight. They were informed that this was not a weight-loss study but appropriate advice was given on portion size and fat intake to help them meet their body weight objectives. Participants were also provided with a checklist with either low–glycemic index or high–cereal fiber food options from different categories (breakfast cereals, breads, vegetables, fruit) as approximately 15-g carbohydrate servings. The number of carbohydrate servings prescribed covered 42% to 43% of total dietary calories.
In the low–glycemic index diet, the following foods were emphasized: low–glycemic index breads (including pumpernickel, rye pita, and quinoa and flaxseed) and breakfast cereals (including Red River Cereal [hot cereal made of bulgur and flax], large flake oatmeal, oat bran, and Bran Buds [ready-to-eat cereal made of wheat bran and psyllium fiber]), pasta, parboiled rice, beans, peas, lentils, and nuts (Table 1). In the high–cereal fiber diet, participants were advised to take the “brown” option (whole grain breads; whole grain breakfast cereals; brown rice; potatoes with skins; and whole wheat bread, crackers, and breakfast cereals) (Table 1). Six servings were prescribed for a 1500-kcal diet, 8 servings for a 2000-kcal diet, and 10 servings for a 2500-kcal diet. Detailed advice was also given to avoid starchy foods not directly recommended as part of the treatment, including those foods advised in the alternative treatment.
In both diets, participants were specifically advised to avoid foods such as pancakes, muffins, donuts, white buns, bagels, rolls, cookies, cakes, popcorn, french fries, and chips. Three servings of fruit and 5 servings of vegetables were encouraged on both treatments. In the low–glycemic index diet, temperate fruit was the focus, including apples, pears, oranges, peaches, cherries, and berries; and in the high–cereal fiber diet, tropical fruit, such as bananas, mangos, guavas, grapes, raisins, watermelon, and cantaloupe, were emphasized. Participants were also advised against eating fruit recommended in the alternative treatment.
Checklists were completed by participants on a daily basis throughout the study and 7-day diet records were completed before each visit. Participants were instructed on how to record using food models as examples of portion size and were asked to give actual weights or to express the amounts in terms of common measures, including cups, teaspoons, and dessert spoons. Adherence was assessed from the 7-day diet records. The daily checklists were of value in alerting the dietitian to problems with adherence to the diet plan over the month before center attendance. The overall goal was to achieve a 10% to 20% reduction in glycemic index on the low–glycemic index diet while keeping dietary fiber similar between treatments.
Blood glucose was measured in the hospital routine analytical laboratory by a glucose oxidase method using a Random Access Analyzer and reagents (SYNCHRON LX Systems; Beckman Coulter, Brea, California), with coefficient of variation between assays of 1.9%. HbA1c was analyzed within 2 days of collection on whole blood collected in EDTA Vacutainer tubes and measured by a designated high-performance liquid chromatography method (Tosoh G7 Automated HPLC Analyzer, Grove City, Ohio), with coefficient of variation of 1.7%.
Serum was analyzed for total cholesterol, triglycerides, and HDL-C by using a Random Access Analyzer and reagents (SYNCHRON LX Systems), with coefficient of variation of 1.5% to 2.4%. Low-density lipoprotein cholesterol (LDL-C) was calculated by the method of Friedewald et al28 [LDL-C = total cholesterol−(triglycerides/5 + HDL-C)] in mg/dL (to convert HDL-C and LDL-C to millimoles per liter, multiply by 0.0259). C-reactive protein was measured by end point nephelometry (Behring BN-100, N high-sensitivity CRP reagent; Dade-Behring, Marburg, Germany), with coefficient of variation of 2.3%. Diets were assessed for macronutrients, fatty acids, cholesterol, fiber, and glycemic index by using a computer program based on US Department of Agriculture data29 and international glycemic index tables,30 with white bread as the standard. Additional measurements were made on local foods, especially specialty breads used as part of the low–glycemic index diet. Glycemic load was calculated as the product of the mean daily available carbohydrate and glycemic index divided by 100.
The primary outcome measure was HbA1c, with glucose, HDL-C, triglycerides, CRP, blood pressure, and body weight as secondary outcome measures.
The original power calculation was based on a predicted HbA1c difference between end of treatments of 0.7 HbA1c units, with an SD of effect of 1.2 HbA1c units, α = .05, 1−β = .80, and a dropout rate of 25%. Our participant requirement was 67 for each treatment. However, the preliminary results of a large (but at the time unpublished) study that assessed the diet effect on HbA1c suggested our effect size would be smaller than we had originally predicted.22 In an unplanned interim analysis, we confirmed a smaller than predicted difference in the change in HbA1c between treatments for the first 58 participants to complete the study. Our revised goal therefore was to detect a difference of half of an SD (0.6%) in the change in HbA1c between treatments (change = 0.3%), with type I error of 2.53% (Bonferroni adjustment to allow for the initial analysis of the preliminary ITT HbA1c data with α = .05) and type II error of 20%. Assuming a loss of 25% of participants due to dropouts and medication changes, 104 participants were needed in each treatment group.
Results are expressed as means with 95% confidence intervals (CIs). All analyses were performed by using SAS version 188.8.131.52 Three participant groups were analyzed. The first and primary analysis was an ITT analysis, which included all 210 randomized participants, with the baseline observation carried forward, rather than the last observation, for all those who did not complete the study. This procedure was used to recognize that these individuals would most likely have reverted to their previous diets and behavioral patterns, resulting also in a reversion to their previous level of diabetic control. Participants who were randomized but did not receive intervention (n = 11) had their screening value used as baseline and this value was carried forward. The second analysis included 155 participants who completed the study (completer analysis), and the third analysis involved 124 participants who completed the study according to the protocol but who did not change their antihyperglycemic medications before or during the study period (per-protocol analysis).
The significance of treatment differences was assessed by using an analysis of covariance model (Proc GLM)31 with change from baseline to end of study as the response variable and diet and sex as main effects, with diet × sex as the interaction term and baseline as a covariate. Additional analyses were also performed to assess the effects of possible factors at baseline associated with diabetes outcome, including adequate glycemic control (HbA1c ≤7.0%), normal BMI (<25), younger age (<60 years), and smoking. The analysis of covariance model for these analyses was the basic model described above together with the main effect of interest and its diet interaction. The time trends were estimated by using Proc GLM in SAS and the repeated statement assuming a spatial power covariance structure to control for unevenly spaced time intervals between measurements. This is an autoregressive test assuming correlations between successive observations. A significant time trend therefore indicates a monotonic relationship between time and change in the study measurement. HbA1c, glucose, HDL-C, and the lipid ratios with HDL-C in the denominator showed significant time trends. No time trend was observed between weeks 2 to 24 for the other measurements.
Additional analyses were performed by using change in fiber, carbohydrate intake, and body weight as covariates in the analysis of covariance model (Proc GLM).31 For skewed data, the results were confirmed with nonparametric analysis but only parametric analyses are reported. The Mantel-Haenszel, χ2 test, and Fisher exact tests were used to assess differences in medication changes31 and baseline comparisons. Pearson product-moment correlations were calculated to examine associations of dietary fiber, glycemic index, and body weight change with other variables of interest. Partial correlations controlling for either change in body weight, fiber, or glycemic index were also undertaken to examine the independent association of fiber and glycemic index with the variables of interest.31
The participant flow diagram is shown in Figure 1. Eleven participants dropped out after randomization but before their baseline visit and were therefore unaware of their treatment allocation (5 in the high–cereal fiber diet group and 6 in the low–glycemic index diet group). After starting the high–cereal fiber diet, 23 of 99 participants (23%) dropped out compared with 19 of 100 participants (19%) in the low–glycemic index diet group.
There were no treatment differences at baseline (Table 2), with the exception of more carbohydrate and less fat consumed before the high–cereal fiber diet compared with the low–glycemic index diet (Table 3). By the end of the study, although carbohydrate intake increased similarly on both treatments, fiber intake increased slightly more with the low–glycemic index diet (18.7 g/1000 kcal at week 24) than with the high–cereal fiber diet (15.7 g/1000 kcal at week 24; P < .001). The glycemic index decreased with the low–glycemic index diet (from 80.8 to 69.6 glycemic index units) compared with an increase in the high–cereal fiber diet (from 81.5 to 83.5 glycemic index units), indicating adherence with the low–glycemic index diet (P < .001) (Table 3). In the assessment of the last 114 participants enrolled, the dietitians rated adherence similarly and poor in 8 of 58 participants (13.8%) in the high–cereal fiber diet group and in 6 of 56 participants (10.7%) in the low–glycemic index diet group. Satiety ratings (scale, +4 to –4) were also similar in both diets (high–cereal fiber diet [n = 94], 0.79; 95% CI, 0.60-0.98; and low–glycemic index diet [n = 93], 0.74; 95% CI, 0.57-0.91; P = .71).
Glycemic Control and Body Weight
In the ITT analysis, antihyperglycemic medication dosages increased similarly in both treatments (3 participants in low–glycemic index diet and 3 participants in high–cereal fiber diet), but dosage reductions were more frequent in the low–glycemic index diet group (13 participants in low–glycemic index diet and 4 participants in high–cereal fiber diet, P = .06).
In the ITT analysis, HbA1c decreased by –0.50% absolute HbA1c units (95% CI, –0.61% to –0.39%) in the low–glycemic index diet compared with –0.18% absolute HbA1c units (95% CI, –0.29% to –0.07%) in the high–cereal fiber diet (Figure 2). The relative change in absolute HbA1c units in the low–glycemic index diet compared with the high–cereal fiber diet was –0.33% (95% CI, –0.48% to –0.17%) (P < .001 and P = .001 after Bonferroni correction) (Table 4). A treatment difference was observed in glucose of –6.8 mg/dL (95% CI, –12.9 to –0.6 mg/dL; P = .02). Difference in body weight reduction was not significant (–0.9 kg; 95% CI, –1.71 to –0.04 kg; P = .053) (Table 4). An additional assessment determined that the reduction in HbA1c in the low–glycemic index diet was still significant after controlling for changes in body weight (P = .002), fiber (P < .001), or carbohydrate (P < .001) separately as covariates in the analysis of covariance model. No significant interactions were observed between diet and the following factors: sex, baseline glycemic control, age, or BMI in relation to the treatment effect on HbA1c.
The diet and smoking interaction approached significance (P = .06), although in the low–glycemic index diet group, the HbA1c response was similar in the 12 participants who were smokers (–0.65%) compared with the nonsmokers (–0.48%). However, the smokers were too few to allow meaningful conclusions. Exclusion of the smokers from the analysis of covariance model assessment of the change in HbA1c made no difference to the significance of the treatment effect favoring the low–glycemic index diet (nonsmokers [n = 196], –0.33%; P < .001; vs including smokers [n = 210], –0.32%; P < .001).
No significant treatment differences were observed in lipid medication changes, but significant treatment effects were observed for HDL-C and the LDL-C:HDL-C ratio (Figure 2). In the ITT analysis, HDL-C increased in the low–glycemic index diet group by 1.7 mg/dL (95% CI, 0.8 to 2.6 mg/dL) and decreased by –0.2 mg/dL (95% CI, –0.9 to 0.5 mg/dL) in the high–cereal fiber diet group (P = .005) (Table 4). Adjusting for change in body weight, carbohydrate, or fiber intake did not affect the significance of HDL-C in the ITT cohort (P = .01, P = .004, and P = .009, respectively). As a consequence, the LDL-C:HDL-C ratio showed a greater reduction in the low–glycemic index diet group (–0.12; 95% CI, –0.23 to –0.01) compared with the high–cereal fiber diet group (–0.01; 95% CI, –0.09 to 0.08; P = .047).
The number of participants who changed blood pressure medications was not different between treatments (low–glycemic index diet, 5 increased and 5 decreased; vs high–cereal fiber diet, 3 increased and 8 decreased; P = .18). Systolic and diastolic blood pressure dropped slightly in both diets, but the treatment differences did not reach significance (Table 4).
For the ITT analysis, the reductions from baseline at 24 weeks were –1.6 mg/L (95% CI, –2.9 to –0.3 mg/L; P = .02) in the low–glycemic index diet and –1.8 mg/L (95% CI, –3.9 to 0.4 mg/L; P = .11) in the high–cereal fiber diet (Table 4). The reductions in both treatments were similar (P = .78).
Associations Between Glycemic Index, Fiber, and Body Weight With Outcome Measures
In the ITT analysis (Table 5), the reduction in dietary glycemic index was positively associated with the reduction in HbA1c (n = 195, r = 0.35, P < .001) and negatively associated with HDL-C (n = 195, r = –0.19, P = .009).
Change in body weight was significantly related to HbA1c (n = 210, r = 0.50, P < .001), blood glucose (n = 210, r = 0.36, P < .001), systolic (n = 209, r = 0.14, P = .04) and diastolic blood pressure (n = 209, r = 0.20, P = .005), and negatively associated with HDL-C (n = 210, r = –0.19, P = .007). No significant associations were observed with dietary fiber apart from systolic blood pressure (n = 195, r = –0.16, P = .03) (Table 5).
Partial correlations indicated that after controlling for change in body weight the association between change in glycemic index and change in HbA1c remained significant (r = 0.28, P < .001), as it did after controlling for change in fiber (r = 0.33, P < .001). In contrast, the association between change in fiber and change in HbA1c was still nonsignificant after controlling for change in glycemic index (r = –0.02, P = .75).
There were no serious adverse events directly related to the study; however, of the 11 participants who reduced their diabetes medications, all 6 who had clear evidence of hypoglycemic symptoms or low blood glucose levels were taking low–glycemic index diets. In addition, 1 participant in the low–glycemic index diet was withdrawn due to high HbA1c (>8.5%) on 2 successive measurements but the HbA1c had increased from 7.6% at screening to 8.2% at the start of the study. The further increase during the study did not therefore appear related to the intervention.
Completer and Per-Protocol Results
In general, the completer and per-protocol analyses confirmed the ITT analysis but showed larger effect sizes and greater significance levels (Figure 3). The completer analysis showed a significant reduction in antihyperglycemic medication use in the low–glycemic index diet group (low–glycemic index diet, 13/15 medication changes vs high–cereal fiber diet, 3/6 medication changes; P = .02). Greater reductions in HbA1c were observed in the low–glycemic index diet (–0.66%; 95% CI, –0.79% to –0.53%) compared with in the high–cereal fiber diet (–0.24%; 95% CI, –0.40% to –0.09%; P < .001), and an increase in HDL-C in the low–glycemic index diet (2.2 mg/dL; 95% CI, 1.0 to 3.5 mg/dL) compared with in the high–cereal fiber diet (–0.3 mg/dL; 95% CI, –1.3 to 0.7 mg/dL; P = .01).
The results of the per-protocol analysis were similar except that the reduction in the blood glucose was no longer significant but a significantly greater reduction was observed in body weight in the low–glycemic index diet (–3.6 kg; 95% CI, –4.6 to –2.5 kg) compared with in the high–cereal fiber diet (–2.3 kg; 95% CI, –2.9 to –1.6 kg; P = .04).
Lowering the glycemic index of the diet improved glycemic control and risk factors for coronary heart disease (CHD). These data have important implications for the treatment of diabetes where the goal has been tight glycemic control to avoid complications.32
The reduction in HbA1c was modest, but we think it has clinical relevance. Reductions in HbA1c of 1% and 0.67%, respectively, in the United Kingdom Prospective Diabetes Study2 and ADVANCE6 studies resulted in 37% and 21% reductions in microvascular complications of diabetes. Our 0.33% HbA1c treatment difference might therefore be expected to reduce microvascular complications by 10% to 12% and possibly more in the longer term, because there was a significant time trend for HbA1c with a tendency for separation of low–glycemic index and high–cereal fiber values over time. Furthermore, the US Food and Drug Administration recognizes a 0.3% to 0.4% reduction in HbA1c as clinically meaningful in the development of drugs for diabetes treatment,33 and the reduction in HbA1c observed in our study was achieved in individuals already treated with 1 or more antihyperglycemic medications, and without weight gain, which often accompanies treatment with glucose-lowering medications.6,7,23
Pharmacological interventions to improve glycemic control in type 2 diabetes have often failed to show a significant reduction in cardiovascular events.2-5 In view of the 2- to 4-fold increase in CHD risk in participants with type 2 diabetes, the ability of a low–glycemic index diet to address both glycemic control and CHD risk factors increases the clinical relevance of this approach for patients with type 2 diabetes, such as those in this study, who are overweight and also taking statins for CHD risk reduction.
Low HDL-C is one of the characteristics of the dyslipidemia associated with type 2 diabetes and may be part of the reason for the increased CHD risk observed in type 2 diabetes.34 Current data suggest a 1:1 to 1:3 relationship between the percentage increase in HDL-C and the reduction in CHD risk.35-37 By this reasoning, the 4.7% increase in HDL-C observed in our study in the low–glycemic index treatment group might reduce CHD risk by 4% to 11% in the ITT group and 6% to 19% in the completer group. Previous cross-sectional studies have noted a negative relationship between glycemic index12,13 or glycemic load14 and HDL-C. One short-term randomized controlled trial of patients with type 2 diabetes also observed an increase in HDL-C in a low–glycemic index diet,38 which was also observed in our study.
In our study, the reduction in CRP in the high–cereal fiber diet group was not significant, although significant reductions have been shown previously.39 The low–glycemic index diet reduced CRP levels similarly but the reduction from baseline was significant. This effect has been reported in earlier studies.16,22
Our study has a number of limitations. The dropout rate during the treatment periods was high at 21% (42/199 participants). However, these numbers are in keeping with many nutritional studies.40 Furthermore, the significance of the data was not materially altered when only the completers were assessed. As observed in other studies,41 a substantial number of participants (11%, n = 23) had antihyperglycemic medication changes during the study. Only 6 participants had clear evidence of repeated hypoglycemic episodes or low blood glucose levels, but all of these occurred in the low–glycemic index diet group. Furthermore, retention of participants with medication changes in the analysis did not result in loss of significance in established treatment differences. Dietary fiber intakes were not completely balanced between treatments with an approximately 4.6 g/d higher fiber intake in the low–glycemic index diet than in the high–cereal fiber diet at week 24. Viscous fibers or diets high in fiber from a variety of sources have been shown to improve blood lipids and markers of glycemic control.24,42 However, with the exception of oat and barley fiber, cereal fibers are largely without metabolic effect. Increasing dietary intake of wheat fiber, even by as much as 20 g/d, has been shown not to influence HbA1c or other biomarkers of chronic disease in patients with type 2 diabetes.43 In addition, controlling for fiber intake as a covariate in the analysis of covariance analysis did not alter the significance of the results. Similarly, controlling for fiber in the partial regression analysis did not alter the significance of the association of the change in glycemic index with change in HbA1c. A further potential weakness was that our study was a single-site study, which may be seen as a limitation to its generalizability.
Study strengths include the independence of the effect of glycemic index on HbA1c from the fiber or carbohydrate intake and the similarity of the observed HbA1c effect with the magnitude of reduction in glycated proteins observed in a meta-analysis.11 Another strength of our study was the comparison of the low–glycemic index diet with a high–cereal fiber diet representing another treatment rather than simply a control. Increased cereal fiber intakes have been associated with reduced incidence of diabetes and CHD in the longer term8,9,23-25 and so provided a reason for dietary adherence.
Body weight is recognized as a major determinant of glycemic control.27,44 However, in our study, the treatment difference in HbA1c remained even after controlling for body weight. Neither the effect of overweight nor the interaction of diet with overweight was significant. Low–glycemic index or low–glycemic load diets have been associated with weight loss,45,46 although the weight loss was not significant (P = .053) in the ITT analysis of our study. At the same time, weight loss resulting from altered insulin economy may be one of the mechanisms through which low–glycemic index diets improve glycemic control.47 Furthermore, interventions that improve glycemic control without increasing body weight confer an additional benefit in diabetes management by avoiding the macrovascular complications that may be associated with weight gain.7
In conclusion, in this study of patients with type 2 diabetes, 6-month treatment with a low–glycemic index diet resulted in moderately lower HbA1c levels compared with a high–cereal fiber diet. Low–glycemic index diets may be useful as part of the strategy to improve glycemic control in patients with type 2 diabetes taking antihyperglycemic medications.27,48
Corresponding Author: David J. A. Jenkins, MD, Department of Nutritional Sciences, Faculty of Medicine, University of Toronto, 150 College St, Toronto, ON, M5S 3E2, Canada (email@example.com).
Author Contributions: Drs Jenkins and Kendall had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Jenkins, Kendall, McKeown-Eyssen, R. Josse, Booth, Vidgen, Leiter.
Acquisition of data: Jenkins, Kendall, Vidgen, A. Josse, Nguyen, Ares, Mitchell, Emam, Augustin, Parker.
Analysis and interpretation of data: Jenkins, Kendall, McKeown-Eyssen, R. Josse, Silverberg, Booth, Vidgen, Nguyen, Corrigan, Banach, Parker.
Drafting of the manuscript: Jenkins, Kendall, McKeown-Eyssen, Corrigan, Banach, Augustin.
Critical revision of the manuscript for important intellectual content: Jenkins, Kendall, McKeown-Eyssen, R. Josse, Silverberg, Booth, Vidgen, A. Josse, Nguyen, Ares, Mitchell, Emam, Parker, Leiter.
Statistical analysis: Vidgen.
Obtained funding: Jenkins, Kendall, McKeown-Eyssen.
Administrative, technical, or material support: Kendall, A. Josse, Nguyen, Corrigan, Banach, Ares, Mitchell, Emam, Augustin, Parker.
Study supervision: Jenkins, Kendall, R. Josse, Silverberg, Booth, Leiter.
Financial Disclosures: Dr Jenkins reported serving on the Scientific Advisory Board of Unilever, the Sanitarium Company, and the California Strawberry Commission; receiving honoraria for scientific advice from the Almond Board of California, Barilla, and Unilever Canada; being on the speaker's panel for the Almond Board of California; and receiving research grants from Loblaws, Unilever, Barilla, and the Almond Board of California. His wife is a director of Glycemic Index Laboratories, Toronto, Ontario, Canada. Dr Kendall reported having been on the speaker's panel for the Almond Board of California and receiving partial salary funding from research grants provided by Unilever, Loblaws, and the Almond Board of California. Mr Vidgen reported receiving partial salary funding from research grants provided by Unilever, Loblaws, and the Almond Board of California. No other authors reported any financial disclosures.
Funding/Support: This work was supported by the Canadian Institutes of Health Research, Canada Research Chair Endowment of the Federal Government of Canada, and Barilla (Italy).
Role of the Sponsors: None of the funding organizations or sponsors played any role in the design and conduct of the study, in the collection, management, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.
Additional Contributions: We thank Roberto Ciati, PhD, and Marta Bianchi, PhD (Barilla, Italy), for their assistance with the study, and to the study participants for their attention to detail and enthusiasm. No compensation was given.
Knowler WC, Barrett-Connor E, Fowler SE,
et al; Diabetes Prevention Program Research Group. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med
. 2002;346(6):393-40311832527PubMedGoogle ScholarCrossref
UK Prospective Diabetes Study (UKPDS) Group. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet
. 1998;352(9131):837-8539742976PubMedGoogle ScholarCrossref
Meinert CL, Knatterud GL, Prout TE, Klimt CR. A study of the effects of hypoglycemic agents on vascular complications in patients with adult-onset diabetes, II: mortality results. Diabetes
Nissen SE, Wolski K. Effect of rosiglitazone on the risk of myocardial infarction and death from cardiovascular causes. N Engl J Med
. 2007;356(24):2457-247117517853PubMedGoogle ScholarCrossref
Patel A, MacMahon S, Chalmers J,
et al; ADVANCE Collaborative Group. Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes. N Engl J Med
. 2008;358(24):2560-257218539916PubMedGoogle ScholarCrossref
Gerstein HC, Miller ME, Byington RP,
et al; Action to Control Cardiovascular Risk in Diabetes Study Group. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med
. 2008;358(24):2545-255918539917PubMedGoogle ScholarCrossref
Salmerón J, Ascherio A, Rimm EB,
et al. Dietary fiber, glycemic load, and risk of NIDDM in men. Diabetes Care
. 1997;20(4):545-5509096978PubMedGoogle ScholarCrossref
Salmerón J, Manson JE, Stampfer MJ, Colditz GA, Wing AL, Willett WC. Dietary fiber, glycemic load, and risk of non-insulin-dependent diabetes mellitus in women. JAMA
. 1997;277(6):472-4779020271PubMedGoogle ScholarCrossref
Liu S, Willett WC, Stampfer MJ,
et al. A prospective study of dietary glycemic load, carbohydrate intake, and risk of coronary heart disease in US women. Am J Clin Nutr
. 2000;71(6):1455-146110837285PubMedGoogle Scholar
Brand-Miller J, Hayne S, Petocz P, Colagiuri S. Low-glycemic index diets in the management of diabetes: a meta-analysis of randomized controlled trials. Diabetes Care
. 2003;26(8):2261-226712882846PubMedGoogle ScholarCrossref
Ford ES, Liu S. Glycemic index and serum high-density lipoprotein cholesterol concentration among us adults. Arch Intern Med
. 2001;161(4):572-57611252117PubMedGoogle ScholarCrossref
Frost G, Leeds AA, Dore CJ, Madeiros S, Brading S, Dornhorst A. Glycaemic index as a determinant of serum HDL-cholesterol concentration. Lancet
. 1999;353(9158):1045-104810199351PubMedGoogle ScholarCrossref
Liu S, Manson JE, Stampfer MJ,
et al. Dietary glycemic load assessed by food-frequency questionnaire in relation to plasma high-density-lipoprotein cholesterol and fasting plasma triacylglycerols in postmenopausal women. Am J Clin Nutr
. 2001;73(3):560-56611237932PubMedGoogle Scholar
Järvi AE, Karlström BE, Granfeldt YE, Björck IE, Asp NG, Vessby BO. Improved glycemic control and lipid profile and normalized fibrinolytic activity on a low-glycemic index diet in type 2 diabetic patients. Diabetes Care
. 1999;22(1):10-1810333897PubMedGoogle ScholarCrossref
Liu S, Manson JE, Buring JE, Stampfer MJ, Willett WC, Ridker PM. Relation between a diet with a high glycemic load and plasma concentrations of high-sensitivity C-reactive protein in middle-aged women. Am J Clin Nutr
. 2002;75(3):492-49811864854PubMedGoogle Scholar
Chiasson JL, Josse RG, Gomis R, Hanefeld M, Karasik A, Laakso M. Acarbose for prevention of type 2 diabetes mellitus: the STOP-NIDDM randomised trial. Lancet
. 2002;359(9323):2072-207712086760PubMedGoogle ScholarCrossref
Chiasson JL, Josse RG, Gomis R, Hanefeld M, Karasik A, Laakso M. Acarbose treatment and the risk of cardiovascular disease and hypertension in patients with impaired glucose tolerance: the STOP-NIDDM trial. JAMA
. 2003;290(4):486-49412876091PubMedGoogle ScholarCrossref
Coulston AM, Reaven GM. Much ado about (almost) nothing. Diabetes Care
. 1997;20(3):241-2439051364PubMedGoogle Scholar
Franz MJ. The glycemic index: not the most effective nutrition therapy intervention. Diabetes Care
. 2003;26(8):2466-246812882880PubMedGoogle ScholarCrossref
Pi-Sunyer X. Glycemic index in early type 2 diabetes. Am J Clin Nutr
. 2008;87(1):3-418175729PubMedGoogle Scholar
Wolever TMS, Gibbs AL, Mehling C,
et al. The Canadian Trial of Carbohydrates in Diabetes (CCD), a 1-y controlled trial of low-glycemic-index dietary carbohydrate in type 2 diabetes: no effect on glycated hemoglobin but reduction in C-reactive protein. Am J Clin Nutr
. 2008;87(1):114-12518175744PubMedGoogle Scholar
Anderson JW, Randles KM, Kendall CW, Jenkins DJ. Carbohydrate and fiber recommendations for individuals with diabetes: a quantitative assessment and meta-analysis of the evidence. J Am Coll Nutr
. 2004;23(1):5-1714963049PubMedGoogle ScholarCrossref
Chandalia M, Garg A, Lutjohann D, von Bergmann K, Grundy SM, Brinkley LJ. Beneficial effects of high dietary fiber intake in patients with type 2 diabetes mellitus. N Engl J Med
. 2000;342(19):1392-139810805824PubMedGoogle ScholarCrossref
Schulze MB, Liu S, Rimm EB, Manson JE, Willett WC, Hu FB. Glycemic index, glycemic load, and dietary fiber intake and incidence of type 2 diabetes in younger and middle-aged women. Am J Clin Nutr
. 2004;80(2):348-35615277155PubMedGoogle Scholar
National Cholesterol Education Program. Executive Summary of the Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). JAMA
. 2001;285(19):2486-249711368702PubMedGoogle ScholarCrossref
Franz MJ, Bantle JP, Beebe CA,
et al; American Diabetes Association. Nutrition principles and recommendations in diabetes. Diabetes Care
. 2004;27:(suppl 1)
Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem
. 1972;18(6):499-5024337382PubMedGoogle Scholar
US Department of Agriculture. Composition of Foods, Agriculture Handbook No. 8: The Agricultural Research Service. Washington, DC: US Dept of Agriculture; 1992
Foster-Powell K, Holt SH, Brand-Miller JC. International table of glycemic index and glycemic load values: 2002. Am J Clin Nutr
. 2002;76(1):5-5612081815PubMedGoogle Scholar
SAS/STAT User's Guide, version 9.1.2. Cary, NC: SAS Institute Inc; 2006
US Food and Drug Administration. Guidance for industry: diabetes mellitus: developing drugs and therapeutic biologics for treatment and prevention. http://www.fda.gov/cder/guidance/7630dft.pdf. Released February 13, 2008. Accessed November 6, 2008
Ginsberg HN. Lipoprotein physiology in nondiabetic and diabetic states: relationship to atherogenesis. Diabetes Care
. 1991;14(9):839-8551959476PubMedGoogle ScholarCrossref
Robins SJ, Collins D, Wittes JT,
et al. Relation of gemfibrozil treatment and lipid levels with major coronary events: VA-HIT: a randomized controlled trial. JAMA
. 2001;285(12):1585-159111268266PubMedGoogle ScholarCrossref
Schaefer EJ, Lamon-Fava S, Ordovas JM,
et al. Factors associated with low and elevated plasma high density lipoprotein cholesterol and apolipoprotein A-I levels in the Framingham Offspring Study. J Lipid Res
. 1994;35(5):871-8828071609PubMedGoogle Scholar
Manninen V, Tenkanen L, Koskinen P,
et al. Joint effects of serum triglyceride and LDL cholesterol and HDL cholesterol concentrations on coronary heart disease risk in the Helsinki Heart Study: implications for treatment. Circulation
. 1992;85(1):37-451728471PubMedGoogle ScholarCrossref
Luscombe ND, Noakes M, Clifton PM. Diets high and low in glycemic index versus high monounsaturated fat diets: effects on glucose and lipid metabolism in NIDDM. Eur J Clin Nutr
. 1999;53(6):473-47810403584PubMedGoogle ScholarCrossref
Ajani UA, Ford ES, Mokdad AH. Dietary fiber and C-reactive protein: findings from National Health and Nutrition Examination Survey data. J Nutr
. 2004;134(5):1181-118515113967PubMedGoogle Scholar
Dansinger ML, Gleason JA, Griffith JL, Selker HP, Schaefer EJ. Comparison of the Atkins, Ornish, Weight Watchers, and Zone diets for weight loss and heart disease risk reduction: a randomized trial. JAMA
. 2005;293(1):43-5315632335PubMedGoogle ScholarCrossref
Barnard ND, Cohen J, Jenkins DJ,
et al. A low-fat vegan diet improves glycemic control and cardiovascular risk factors in a randomized clinical trial in individuals with type 2 diabetes. Diabetes Care
. 2006;29(8):1777-178316873779PubMedGoogle ScholarCrossref
Jenkins DJ, Newton C, Leeds AR, Cummings JH. Effect of pectin, guar gum, and wheat fibre on serum-cholesterol. Lancet
. 1975;1(7916):1116-111749474PubMedGoogle ScholarCrossref
Jenkins DJ, Kendall CW, Augustin LS,
et al. Effect of wheat bran on glycemic control and risk factors for cardiovascular disease in type 2 diabetes. Diabetes Care
. 2002;25(9):1522-152812196421PubMedGoogle ScholarCrossref
Klein S, Sheard NF, Pi-Sunyer X,
et al; American Diabetes Association; North American Association for the Study of Obesity; American Society for Clinical Nutrition. Weight management through lifestyle modification for the prevention and management of type 2 diabetes: rationale and strategies: a statement of the American Diabetes Association, the North American Association for the Study of Obesity, and the American Society for Clinical Nutrition. Diabetes Care
. 2004;27(8):2067-207315277443PubMedGoogle ScholarCrossref
Ebbeling CB, Ludwig DS. Treating obesity in youth: should dietary glycemic load be a consideration? Adv Pediatr
. 2001;48:179-21211480757PubMedGoogle Scholar
Maki KC, Rains TM, Kaden VN, Raneri KR, Davidson MH. Effects of a reduced-glycemic-load diet on body weight, body composition, and cardiovascular disease risk markers in overweight and obese adults. Am J Clin Nutr
. 2007;85(3):724-73417344493PubMedGoogle Scholar
Ebbeling CB, Leidig MM, Feldman HA, Lovesky MM, Ludwig DS. Effects of a low-glycemic load vs low-fat diet in obese young adults: a randomized trial. JAMA
. 2007;297(19):2092-210217507345PubMedGoogle ScholarCrossref
Miles JM. A role for the glycemic index in preventing or treating diabetes? Am J Clin Nutr
. 2008;87(1):1-218175728PubMedGoogle Scholar