Influence of and Optimal Insulin Therapy for a Low–Glycemic Index Meal in Children With Type 1 Diabetes Receiving Intensive Insulin Therapy

Children and adolescents with type 1 diabetes diagnosed for ≥1 year and using multiple daily injection therapy (≥4 injections/day) for >6 months were recruited from the John Hunter Children's Hospital Diabetes Clinic. Eligibility criteria included age between 7 and 17 years inclusive, that subjects not be obese, and recent A1C ≤8.5% (Primus PDQ A1c Analyzer; Primus, Kansas City, MO). Exclusion criteria included complications of diabetes such as gastroparesis or coexisting medical disorders such as celiac disease. Recruitment continued until a sample size of 20 was reached.

Ethics approval was obtained by the Hunter New England Human Research Ethics Committee in December 2006. Written consent was obtained from all participants and/or their parents.

Participants and their families were contacted by telephone 2 weeks before study commencement to review BGLs and adjust insulin therapy to optimize waking BGLs (4–8 mmol/l) and minimize the risk of hypoglycemia, defined as symptomatic or BGL <3.5 mmol/l. Four test conditions were administered to each child at breakfast on 4 consecutive days in a randomized way using a permuted block method with block size of four, stratified by test condition: 1) low–glycemic index meal (glycemic index 48), preprandial ultra-short-acting insulin; 2) high–glycemic index meal (glycemic index 84), preprandial ultra-short-acting insulin; 3) low–glycemic index meal (glycemic index 48), preprandial regular insulin; and 4) low–glycemic index meal (glycemic index 48), postprandial ultra-short-acting insulin.

Because of an increased risk of sensor failure on day 4, test conditions 1 and 2 were randomly assigned on the first 2 days to ensure conservation of these results. Test conditions 3 and 4 were randomly assigned on the final 2 days.

The preprandial injection of insulin was administered immediately before meal consumption. Postprandial insulin was administered 15 min after meal commencement. A kitchen timer was supplied to each participant to ensure accuracy. All children received their normal breakfast insulin dose based on the amount of carbohydrate in the test meal. If the preprandial BGL was >16 mmol/l, the subject was instructed to add a correction dose of insulin and was asked to repeat the study.

Participants were provided with four premade test meals. Each meal was consumed at breakfast after a minimum 10-h overnight fast. Evening meals and suppers the preceding nights were standardized for carbohydrate amount and type to limit the impact of these meals on the analysis of breakfast PPGE. Participants were supplied with measuring cups to facilitate compliance with preparation of these meals.

Children were required to fast and standardize their activity during the 4-h postprandial period. A food and activity diary was kept by each subject/parent to determine adherence to this protocol.

The low–glycemic index test meal (glycemic index 48) and the high–glycemic index test meal (glycemic index 84) consisted of a ham sandwich and a drink (Table 1). Sandwiches with a concentrated yeast extract spread (Vegemite) were offered as a vegetarian option, which two participants accepted. The carbohydrate, fat, protein, and fiber amounts were standardized for both meals (Table 1). The glycemic index of each food was obtained from published tables (20) where available and confirmed by manufacturers. The glycemic index of each meal was calculated using methods described previously (21).

Food was weighed using Salter kitchen scales (model 323; Salter, Kent, U.K.), which measure to ±1 g. Liquids were measured using a 250-ml measuring jug and a 10-ml disposable syringe.

All meals were supplied in an insulated bag the afternoon before study commencement. Sandwiches and drinks were individually wrapped and labeled with the day of consumption and instructions to refrigerate the drinks and freeze the sandwiches, defrosting them the night before consumption.

A continuous glucose monitoring system was used (Medtronic; MiniMed, Northridge, CA). The sensor was inserted in the abdominal subcutaneous tissue the afternoon before study commencement. Subjects were instructed to enter four BGLs per day into the monitor at a time when BGLs were stable for calibration. They were asked to enter a “meal marker” into the continuous glucose monitoring system immediately before meal consumption.

Only subjects who had acceptable traces for each of the four test conditions were accepted for analysis. If sensor failure occurred, then the test condition was repeated using a new sensor.

The 4-h postprandial period was analyzed using SPSS software (version 15; SPSS, Chicago, IL). A one-way repeated-measures ANOVA was used to analyze the following parameters: 1) blood glucose excursions at 30-min increments, 2) incremental area under the curve (AUC) as described previously (22), 3) peak blood glucose level, 4) time to peak BGL, and 5) time to return to baseline.

If significance was reached (P < 0.05) the test of least significant differences (LSD) was applied as the post hoc test. Sensitivity analysis was performed using Bonferroni correction to adjust for multiple pairwise comparisons. Results are presented as means with 95% CIs.

Thirty-four participants were recruited. Two participants withdrew before study commencement for personal reasons, and one participant was excluded because of an inappropriate A1C. Of the 31 children who participated, the results of 11 participants were excluded from analysis because of sensor failure (n = 4), inappropriate starting BGL on test days (n = 2), prolonged hypoglycemia (n = 2), failure to comply with the protocol (n = 1), diagnosis during the study period of a medical condition that would interfere with study protocol (n = 1), and sensor misplaced (n = 1).

The mean age of participants was 13.6 ± 2.7 (range 8.3–17.7) years. The subjects had been diagnosed with type 1 diabetes for 5.2 ± 3.8 (1.0–13.6) years and had good metabolic control (A1C 7.4 ± 0.7 [6.1–8.5] %). The starting BGL for each of the four test conditions was not significantly different (one-way repeated-measures ANOVA P > 0.05) (Table 2).

The high–glycemic index meal resulted in a significantly higher PPGE at all time points between 30 and 180 min compared with the low–glycemic index meal (LSD, P < 0.02) (Fig. 1A). The maximum difference between the PPGE after each test condition occurred at 60 min; the high–glycemic index meal was 4.2 mmol/l higher than the low–glycemic index meal (7.1 [range 5.2–9.0] vs. 2.9 [1.5–4.3] mmol/l; LSD, P < 0.0001).

The AUC and peak blood glucose excursion were greater for the high–glycemic index meal than for the low–glycemic index meal (LSD, P < 0.0001) (Table 2). After the high–glycemic index meal, the return to baseline BGL took significantly longer (LSD, P = 0.011) (Table 2). There was no difference in time to peak BGL for the two test conditions (Table 2).

Administering preprandial regular insulin compared with ultra-short-acting insulin resulted in a 1.1 mmol/l higher blood glucose excursion at 30 min only (3.4 [range 2.4–4.4] vs. 2.3 [1.3–3.3] mmol/l; LSD, P = 0.015) (Fig. 1B). Ultra-short-acting insulin reduced the AUC compared with administration of regular insulin (LSD, P = 0.046) (Table 2). There was no difference in the peak blood glucose excursion, time to peak, or time to baseline between the two test conditions (Table 2).

Administration of preprandial ultra-short-acting insulin resulted in a significantly lower PPGE compared with postprandial administration at 30 and 60 min (Fig. 1C). At 60 min the blood glucose excursion was 2.5 mmol/l lower after administration of preprandial insulin compared with postprandial insulin (2.9 [range 1.5–4.3] vs. 5.4 [4.2–6.6] mmol/l; LSD, P < 0.0001).

Preprandial administration resulted in a lower AUC for 0–1 h (1.9 [range 1.0–2.7] vs. 3.5 [2.8–4.2] mmol · h⁻¹ middot; l⁻¹; LSD, P < 0.0001) and 0–2 h (4.4 [1.8–7.0] vs. 7.1 [4.8–9.4] mmol · h⁻¹ · l⁻¹; LSD, P = 0.012); however, the total AUC was not significantly different. The peak BGL was 1.7 mmol/l lower when preprandial insulin was administered (LSD, P = 0.003) (Table 2). There was no difference in the time to peak BGL or time to return to baseline between the two test conditions (Table 2).

Adjusting for multiple comparisons using the Bonferroni correction led to a P = 0.02. Almost all of the P values in the analysis remained significant using this threshold, and this adjustment did not change the conclusions.

To our knowledge, this study is the first to examine whether changing the glycemic index of a mixed meal alters the PPGE in children receiving multiple daily injections. The low–glycemic index meal produced a significantly lower PPGE for 30–180 min, a lower AUC, a smaller peak blood glucose excursion, and reduced time to reach baseline BGLs compared with the high–glycemic index meal when preprandial ultra-short-acting insulin was administered (LSD, P < 0.02).

Previous studies in children receiving conventional insulin therapy and in adults receiving intensive insulin therapy (11,23,24) support our findings. Furthermore, Nansel et al. (13) recently postulated that reductions in postprandial excursions may be responsible for the improved daytime glycemic control they found in youths consuming low–glycemic index diets. However, one study in children failed to show a difference in the PPGE when the glycemic index of a test breakfast was altered (25). This study was influenced by numerous confounders, including failure to standardize fiber and macronutrient content. Additionally, the results were not applicable to children receiving intensive insulin therapy.

It may be hypothesized that the substantial differences demonstrated in the postprandial glucose excursion and AUC between the low–and high–glycemic index test meals in this study should extrapolate to improved A1C. This theory is supported by Brand-Miller's meta-analysis, which detected a 10.6% reduction in the A1C of adults and children with type 1 diabetes when a low–glycemic index diet was consumed over a 2- to 52-week period (10).

The test meals in this study had significantly different glycemic index values (glycemic index 48 vs. 84). A limitation of the study is that more moderate differences in glycemic index values may result in less significant alterations in PPGE. In addition, as discussed by Mohammed and Wolever (24), the shape of the postprandial blood glucose curve may be different if a starchy food of the same glycemic index value was substituted for the fruit juice given in this study. Further studies in mixed meals are warranted to explore these issues.

After consideration of the effect of glycemic index on PPGEs, the question of whether insulin type (regular insulin or ultra-short-acting insulin) or timing of insulin administration should be altered when a low–glycemic index meal is consumed remains. Administration of preprandial ultra-short-acting insulin and regular insulin produced clinically similar PPGEs. Administration of regular insulin produced a 1.1 mmol/l higher blood glucose excursion at 30 min (P = 0.015). Therefore, the use of preprandial regular insulin did not offer an advantage over that of preprandial ultra-short-acting insulin when a low–glycemic index meal was consumed.

This study demonstrated that postprandial administration of ultra-short-acting insulin produced significantly higher postprandial blood glucose levels. These findings are supported by the results of other studies in adults (17,18). Thus, we believe that postprandial administration of ultra-short-acting insulin should not be recommended, as it potentates deviations from normoglycemia.

Dietary focus remains on carbohydrate counting as a strategy to reduce the risk and progression of long-term complications in the context of intensive insulin therapy. The current study supports the integration of low–glycemic index dietary advice into medical nutrition therapy for children and adolescents with type 1 diabetes receiving multiple daily injections. Preprandial ultra-short-acting insulin remains the optimal insulin therapy when a low–glycemic index meal is consumed and postprandial injection of insulin is not recommended as a standard management technique.

We thank the Australian Diabetes Educators Association for delivering the independent grant from Eli Lilly that supported this study.

We thank Kathy Hodge for her assistance throughout the study and Dr. Brent McSharry for his assistance in data collation.

A pilot study was previously published in abstract form in the proceedings of the 33rd Annual Scientific Meeting for the International Society of Pediatric and Adolescent Diabetes, Berlin, Germany, 26–29 September, 2007 (Pediatr Diabetes 8 [Suppl. 7]:60, 2007).