Effect of Varying Glycemic Index Meals on Blood Glucose Control Assessed With Continuous Glucose Monitoring in Youth With Type 1 Diabetes on Basal-Bolus Insulin Regimens

Participants were recruited from a pediatric endocrinology practice; inclusion criteria included diagnosis of type 1 diabetes ≥1 year, insulin dose ≥0.5 units · kg⁻¹ · day⁻¹, and age 7–16 years. Informed consent and assent were obtained. The study was approved by the institutional review board of the National Institutes of Health. A within-subject crossover trial was used; participants consumed 1 day of HGI meals and 1 day of LGI meals in a controlled setting. The order of conditions was counterbalanced, with a washout day between and a standardized evening meal before each condition. Diets were matched for calories and macronutrients; mean glycemic index of the HGI diet was 64 (e.g., corn flakes, white bread, mashed potatoes) and of the LGI diet was 40 (e.g., peaches, kidney beans, brown basmati rice) (online-only appendix table [available at http://dx.doi.org/10.2337/dc07-1879]). Meal timing and activity levels were consistent across conditions. The CGMS (Medtronic MiniMed, Northridge, CA) was used to assess BG profiles. Subjects were given standard BG meters, and BG checks were performed before each meal and 2 h postprandial. BG values <80 mg/dl were treated with 15 g carbohydrate.

Daytime and nighttime values were calculated from the CGMS data for each of the following parameters: mean BG, BG area >180 mg/dl, BG area <70 mg/dl, low BG index (9), and high BG index (9). Frequency of hypoglycemia was calculated from BG meter data. Because insulin dose could vary based on application of the correction factor, and additional carbohydrate could be provided for hypoglycemia treatment, the ratio of actual insulin taken to carbohydrate consumed was calculated. Paired sample t tests were conducted to assess differences between conditions on each continuous outcome except for hypoglycemia frequency, where a skewed distribution necessitated use of Wilcoxon's signed-rank test.

A total of 22 youths participated, although two participants experienced CGMS equipment failure. All subjects used a basal-bolus insulin regimen (65% insulin pump, 35% injections). The mean ± SD duration of diabetes was 5.3 ± 4.5 years, A1C was 8.3 ± 1.8%, and age was 13.1 ± 2.6 years, and 55% were female, 80% were white, and 20% were black or mixed race.

Under the LGI condition, BG levels were in the target range (70–180 mg/dl) 66% of the time versus 47% under the HGI condition (P = 0.002). Participants demonstrated lower daytime mean BG, BG area >180 mg/dl, and high BG index under the LGI condition (Table 1). No differences were observed in BG area <70 mg/dl, low BG index, and nighttime parameters. Mild hypoglycemia occurred more frequently during the LGI condition (one or more episodes in 13 subjects during LGI vs. 8 subjects during HGI, P = 0.007) (supplemental figure [available in an online appendix at http://dx.doi.org/10.2337/dc07-1879]). Comparison of actual insulin taken (including application of correction factor) to carbohydrate consumed (including hypoglycemia treatment) indicated a trend for lower insulin required during the LGI condition. During the HGI condition, participants took 1 unit of insulin for every 10.4 g of carbohydrate consumed. During the LGI condition, participants took 1 unit of insulin for every 12.3 g of carbohydrate consumed (t = −2.07, P = 0.05).

While the use of glycemic index to guide carbohydrate choice has been criticized as minimally beneficial (10), these findings indicate that it has utility for improving glycemic control to a clinically meaningful degree above that obtained by careful carbohydrate counting and contemporary insulin regimens. The LGI diet resulted in significantly lower mean daytime BG, as well as lower scores on two indexes of high BG risk, both which have been associated with A1C (11,12). This effect was observed despite a greater actual amount of carbohydrate being consumed per unit of insulin in the LGI condition. The absence of a difference in nighttime parameters supports the understanding that an LGI diet effects BG through reduction of postprandial excursions.

It is notable that a greater frequency of mild hypoglycemia was observed during the LGI condition. Therefore, insulin dose on an LGI diet may need to be reduced to prevent excessive hypoglycemia. Consistent consumption of an LGI diet may reduce insulin requirement while improving BG control, but careful attention should be given to BG monitoring and adjustment of insulin dose.

Strengths of this study include 1) the use of CGMS to capture the BG profile, 2) a controlled setting to ensure compliance and consistency across conditions, and 3) a sample of youth using insulin-to-carbohydrate regimens. Primary limitations are the study's short duration and small sample size, precluding assessment of effect modification by disease duration or pubertal status.

This study adds to the evidence supporting the utility of an LGI diet in optimizing diabetes management. Furthermore, considering the prevalence of cardiovascular risk factors in youth with diabetes (13) and the increasing prevalence of comorbid type 2 diabetes (14), the additional benefits of an LGI diet shown in previous research (15–17) are also highly relevant. Promoting LGI eating may offer substantial health benefits to people with type 1 diabetes.

This research was supported by the Intramural Research Program of the National Institutes of Health, the National Institute of Child Health & Human Development.

The authors acknowledge Linda Zeitzoff, David Greenberg, Ellie Centenio, and Donna Franz of Mt. Washington Pediatric Hospital for their assistance in the conduct of this study.