The Effect of Flexible Low Glycemic Index Dietary Advice Versus Measured Carbohydrate Exchange Diets on Glycemic Control in Children With Type 1 Diabetes

RESEARCH DESIGN AND METHODS

Study design

Figure 1 summarizes the trial profile used in this study. Children attending the Melbourne Royal Children’s Hospital (RCH) Diabetes Clinic were selected using the following criteria: 1) age between 8 and 13 years, diagnosis of type 1 diabetes for longer than one year, and regular attendance at the clinic (3 monthly); 2) no additional dietary restrictions; 3) no other immediate family members with diabetes, 4) no medications that would affect appetite; and 5) family able to read and write English.

Agreement from the primary physician was sought. Of 112 eligible families, 104 agreed to participate, and letters were sent outlining their involvement in the study. Formal written consent was obtained. Subjects were assigned random number codes to ensure patient confidentiality. Ethics approval was granted by the RCH Ethics in Human Research Committee.

Diet assessment and education

Individual interviews with the research dietitian were used to collect initial data, instruct the child and parent in the use of food records, and develop a rapport to enhance participation over the 12-month period. Each subject was asked to complete a 3-day food diary. Two weekdays and one weekend day were specified to account for the variation in food intake at weekends (24). Families were encouraged not to alter their usual pattern of food intake during recording periods. A sample food diary and a contact phone number were provided.

At the beginning of the study, subjects were assessed by a dietitian to categorize their existing dietary regime and ensure correct stratification before being randomized to either the CHOx or low-GI diet. Computer-generated random numbers of 1 and 2 were generated in blocks of 10 and assigned consecutively to each subject upon recruitment to the study. Of the 104 subjects recruited, 49 were assigned to the CHOx diet, and 55 were assigned to the low-GI diet (Fig. 1). Education regarding the allocated study diet was then given to the child and parent (Table 1). The diet education session was structured similarly for both groups and executed in an outpatient setting by the same clinical dietitian. A specially made flipchart was used for each of the study diets explaining the principles of the diet. Literature was also provided to reinforce the advice (Table 1). No additional education sessions were planned over the 12-month period apart from usual clinic review. The food diaries were completed at 1, 3, 6, and 12 months, and phone calls were made 2 weeks before the clinic visit to ensure compliance.

All food diaries were analyzed by the same research dietitian using the Diet 3.12 computer program (Xyris Software). Portion sizes were estimated against standard portions within the software package based on the household measures recorded. If the food item was not included in the database or if the nutrient profile was incomplete, information from the manufacturer was sought or, alternatively, the most similar food item was substituted. The individual’s intake of energy, protein, fat, fiber, total carbohydrates, total sugars (with inclusion and exclusion, respectively, of sugars consumed for hypoglycemic treatment or used during exercise), nonmilk extrinsic sugars (NMES), GI, and carbohydrate distribution were calculated using the food diaries at each time point. NMES content was estimated from the food sources of total sugars, information from the food manufacturer, and food composition tables (25).

Energy intake was independently assessed as being below, within, or above range. Ranges were based on basal metabolic rate (BMR) calculations using cutoff points from published sources. BMR was calculated using Schofield’s equation (26). The minimum and maximum cutoff points were derived from the work of, respectively, Goldberg et al. (27), using a value of 0.8 × BMR × activity factor, and Torun et al. (28). Activity levels were individually assessed and defined as light (<2 organized activities per week), moderate (2–5 organized activities per week), and heavy (>5 organized activities per week). Activity factors were 1.55, 1.75, and 1.95, respectively.

For the purpose of dietary analysis, the daily GI (relative to a standard glucose value of 100) was calculated by summing the following: [grams of carbohydrate from the food item / total daily carbohydrate × 100 × GI of food item]. GI values were derived from published GI tables (8) and unpublished data from the Human Nutrition Unit, University of Sydney. Of 284 carbohydrate-containing foods, 194 were assigned a known GI, but 90 were given “estimated” values based on the GI of similar foods.

Outcome measures

The following measures were recorded at entry to the study and at 3, 6, and 12 months: HbA_1c level, weight, height, dietary intake information, and incidence of hypoglycemia (<3.5 mmol/l) and hyperglycemia (>15 mmol/l) as determined by preprandial breakfast, dinner, and supper levels charted in the child’s self-reported record book during the 1 month before each visit. In addition, a quality-of-life questionnaire was completed independently by the parent and child (via separate interview) at each time point. HbA_1c measurements were routinely performed in the clinic using the DCA 2000 Analyser (Bayer) on capillary blood samples obtained by fingerprick (mean coefficient of variation 3.8%).

Statistical analysis

The sample size of 104 families allowed for a 15% dropout rate and provided 80% power to reject the null hypothesis, which stated that mean HbA_1c levels after 12 months were the same in the two groups if a 10% difference (i.e., a difference in measured HbA_1c of ≃1%) was found. Significance was set at 5%. The sample size calculation was based on an effect size of 0.625 SD. A subgroup analysis was planned prospectively that divided the subjects into four subgroups, as illustrated by Fig. 1. An intent-to-treat analysis was performed on the assumption that subjects adhered to the dietary advice provided at entry to the study. Data analysis and clinical outcome measures were assessed by researchers unaware of the treatment allocation.

Results were expressed as the mean ± SD, unless otherwise stated, and were analyzed using multiple linear regression or Spearman’s correlation for continuous variables or with a combination of logistic regression or Pearson’s χ² analysis for categorical data. Nonnormal data were analyzed using Wilcoxon’s rank-sum test and expressed as medians and ranges. Clinical data were adjusted for baseline values where specified. All statistical analysis was performed using the Intercooled Stata 5.0 Statistical package (Stata Corporation).

RESULTS

There were no significant differences in baseline measures or demographic data between the two study groups (Table 2). A total of 15 subjects (14%) dropped out during the study period: 11 from the CHOx group and 4 from the low-GI group. There was a significantly higher dropout rate from the CHOx group (22 vs. 7% in the CHOx and low-GI groups, respectively, P = 0.03). Specifically, there was a differential dropout rate of 39% for subgroup GC, who had been assigned to the CHOx diet for the study after previously following the low-GI diet, versus subgroups CC (13%), GG (15%), and CG (3%) (P < 0.01), who were, respectively, only instructed in the CHOx diet, only instructed in the low-GI diet, or randomized to the low-GI diet after following a CHOx diet prior to the study. Apart from dietary assignment, there were no other significant differences at baseline between these subjects.

At baseline, 3, and 6 months, there were no statistically significant differences in HbA_1c levels between the groups. However, by 12 months the mean HbA_1c level had fallen in those provided low-GI dietary instruction while remaining essentially unchanged in the CHOx group (Fig. 2). The difference between groups at 12 months was significant, even after adjusting for baseline values (P = 0.05, Table 2). Subjects who had been given low-GI dietary instruction both before and during the study (subgroup GG) had markedly lower HbA_1c values at the end of 12 months (7.77 ± 0.79%) compared with subjects who had only received CHOx advice (subgroup CC) (8.76 ± 1.07%, P = 0.002).

The differences in HbA_1c levels were not related to variations in insulin therapy because there were no significant differences in insulin dose at 12 months nor were there changes to insulin dose over the 12 months (Table 2). The majority of subjects from both groups were on a twice-daily insulin regimen (Table 2), using regular short- and long-acting insulin types at baseline (96 vs. 98% in the CHOx and low-GI groups, respectively, P = 0.27) and at 12 months (89 vs. 88%, P = 0.21).

The proportions of subjects from each study group within acceptable and unacceptable HbA_1c ranges were compared. The acceptable cutoff was <8% (29) and levels >9% were considered suboptimal. At 12 months, twice as many subjects from the low-GI group (45%) had HbA_1c values within the acceptable range compared with subjects in the CHOx group (22%) (P = 0.02 after adjustment for baseline values). Subjects in the CHOx group were ∼3 times more likely to have undesirable HbA_1c values compared with the low-GI group (47 vs. 18%, P = 0.004 after adjustment for baseline values).

A significantly larger proportion of subjects from the CHOx group reported more frequent episodes of hyperglycemia at 12 months, defined as >15 episodes per month (66 vs. 35%, P = 0.006 after adjustment for baseline values). There were no significant differences in hypoglycemic episodes (Table 2).

Of the 89 subjects who completed the study, 6 did not complete a food diary. The two study groups showed no significant differences in any of the dietary variables measured at 12 months. On average, the children consumed 17% energy as protein, 34% as fat, 49% as carbohydrate, and 7% as NMES, and they consumed 21 g fiber per day. However, there was a high proportion of subjects who appeared to underreport food intake in both groups (53 vs. 46% in the CHOx and low-GI groups, respectively), casting doubt on the reliability of the food diaries. The dietary variables were reanalyzed, excluding underreporters but otherwise remaining essentially unchanged.

Surprisingly, despite differences in dietary instruction, there was no difference in mean GI between the two groups (56.5 ± 4.0 and 55.3 ± 4.8 in the CHOx and low-GI groups, respectively, P = 0.26). Because GI is not a precise measure, we compared the proportions of subjects that had a low (<55), moderate (55–60), and high (>60) GI. At all time points, the low-GI group showed a greater proportion of subjects within the low (GI <55) range (low GI: 62, 47, 47, and 44%; CHOx: 37, 40, 36, and 30% at 1, 3, 6, and 12 months, respectively), but the difference reached statistical significance only at one month (P = 0.02). The low-GI group also contained more subjects who achieved an extremely low average GI (<50%), (0 vs. 14% in the CHOx and low-GI groups, respectively, P = 0.04 at 12 months). From a food variety perspective, the average number of different carbohydrate food choices per day was similar between the two groups (9.9 ± 2.5 vs. 10.8 ± 2.4, P = 0.10).

In correlation analyses, no dietary factor correlated significantly with HbA_1c levels. The reported physical activity level was not different between the dietary groups (median number of organized activities per week [range]: 4.0[1–10] and 4.5[1–10] in the CHOx and low-GI groups, respectively, P = 0.75) and showed no relationship to HbA_1c level.

Quality of life was influenced significantly by type of dietary instruction. Twice as many parents in the low-GI group stated that their child had no difficulties in selecting their own meals at the 12-month time point (51 vs. 24% P = 0.01). Almost twice as many parents from the low-GI group reported that diabetes never limited the types of activities pursued as a family (53 vs. 27% P = 0.02) and that diabetes had never been a source of tension or conflict within the family (55 vs. 27% P = 0.01). There was a trend for the majority of parents from the low-GI group to believe that diabetes never interrupted various everyday family activities (53 vs. 32% P = 0.06).

The 53 children (and their parents) that had experienced both types of dietary approaches (subgroups CG and GC) expressed an overall preference for the low-GI diet compared with the CHOx diet (Table 3, P < 0.01 and P < 0.001 for the children and parents, respectively). The same subgroup of parents believed that the low-GI diet led to better control of blood glucose levels compared with the CHOx diet (P < 0.001). The low-GI diet was the dietary regime that most parents and children selected to continue after completion of the study (P < 0.001 and P < 0.001 for the children and parents, respectively).

CONCLUSIONS

This study demonstrated that children with type 1 diabetes who were given flexible dietary instruction based on the food pyramid and low-GI choices achieved significantly better HbA_1c levels after 12 months compared with those who received more traditional dietary advice. Twice as many children in the low-GI group achieved acceptable HbA_1c levels at 12 months without any increase in the frequency of hypoglycemic episodes. Insulin dose and injection regime did not change or differ between the groups.

The difference in glycemic control was greater and more significant in the children who had received the same dietary instruction both before and throughout the study. This suggests that some aspect of the flexible low-GI approach was associated with gradual but sustained improvement in HbA_1c levels. However, it is difficult to attribute the lower HbA_1c levels solely to differences in diet, particularly when there was no apparent difference in mean GI.

The dietary records showed that macronutrient and fiber intakes were very close to that of children in the general Australian population (30) and were similar in both groups. However, the dietary data need to be interpreted cautiously because of the unacceptably high prevalence of underreporting. About half the records revealed energy intakes that were not likely to reflect the child’s habitual intake. This criticism plagues all dietary assessment studies (31,32) and is especially true of those in children. The high rate of underreporting may have affected the ability to detect subtle differences in carbohydrate quality (mean GI) between the two study groups. The large number of foods with estimated GI values may also have contributed. Despite these limitations, the assumption that a low-GI diet might restrict food variety and increase fat or sugar intake was not borne out by the data (33). Many low-GI foods were particularly favored by the children, including peanut butter, pasta, baked beans, and dairy foods. Given the documented differences in hyperglycemic episodes between the two groups, the dietary data on the whole suggest that individuals in the low-GI group were consuming more low-GI foods than those in the CHOx group.

The quality-of-life questions revealed an obvious preference for the flexible low-GI dietary regimen among the children (and parents) who had experienced both types of dietary instruction. There were significant differences in measures such as family conflict, limitations placed on family activities, difficulties for the child in selecting their own meals, and a clear preference to continue on the low-GI diet. These findings are important and challenge the belief that low-GI diets might be difficult to follow and a burden for people with diabetes. Furthermore, they need to be considered alongside the fact that glycemic control was superior among those who received the simpler, less regimented dietary advice. Using CHOx is often quoted as the most challenging aspect of managing diabetes (2,3).

The findings of this large long-term prospective study suggest that more flexible dietary instruction based on the food pyramid and with an emphasis on low-GI foods has demonstrable benefits for children with diabetes. Weighed-carbohydrate exchange dietary advice was associated with inferior glycemic control and measures of quality of life.