The Carbohydrate Counting in Adolescents With Type 1 Diabetes (CCAT)Study

The study included adolescents aged 12–18 years with type 1 diabetes(as defined by ADA criteria)¹²  for> 1 year who received care from the Barbara Davis Center for Childhood Diabetes in Denver, Colo., and were using insulin:carbohydrate (I:C) ratios for at least one meal per day. The center treated > 1,400 patients between the ages of 12 and 18 years in 2005. Patients coming in for routine follow-up exams were contacted about possibly participating in the study. Patients with celiac disease, type 1 diabetes of < 1 year's duration, an A1C > 12%, or who were non–English speaking were excluded from the study. Fifty individuals provided consent, although one declined to complete the study and one was later deemed ineligible. Therefore, data were analyzed from 48 participants.

Participants assessed the carbohydrate content for 32 foods presented as typical breakfasts, lunches, dinners, and snacks (e.g., banana [snack] or cereal with milk [breakfast]) commonly consumed by youths. Registered dietitians (RDs) selected the foods to be included after reviewing diet records and dietary data collected from youths with diabetes.¹³ 

Foods were presented either as food models or as actual food, with some items presented as standard serving sizes and some self-served by the study participants. Packaged real food items were presented to participants with their Nutrition Facts labels, and participants were allowed to use these food labels if they wished. Participants recorded their estimate of portion size,carbohydrate content, and frequency of consumption. Study staff recorded the use of nutrition labels by participants during the study visit (for foods that were presented with a nutrition label). For self-served foods, the actual weight of the food served was recorded out of sight of the participants. The amount of carbohydrate in each food was determined by either the nutrition label for packaged foods, the Nasco Food Replica Nutrition Guide (based on the U.S. Department of Agriculture's A Standard Reference for Nutrient Composition for food models, or the Nutrient Data System for Research(Version 2007) software licensed from the Nutrition Coordinating Center at the University of Minnesota for foods that were self-served by study participants.

An interviewer-administered questionnaire assessed participants'carbohydrate counting education, family/friend support for carbohydrate counting, and patterns of carbohydrate counting (i.e., at which meals and snacks they count carbohydrates and how often). A1C levels were measured as part of patients' clinical visits. All participants provided informed consent and/or assent, and the study was approved by the local institutional review board.

From the 32 commonly consumed foods, typical breakfasts, lunches, dinners,and snacks were constructed along with typical whole-day menus (three meals and a snack). Accuracy of carbohydrate counting was determined as the difference between actual and participant-estimated carbohydrate content (in grams) for each food, meal, and daily menu. Researchers determined the number of participants who estimated carbohydrates accurately within 5 grams for individual food/food combinations or within 10 grams for meals or snacks based on clinical application of a common I:C ratio for adolescents of 1 unit of insulin for each 10 grams carbohydrate consumed.

T-tests were used to assess the significance of over- or underestimation of carbohydrate content. To describe accuracy for whole days(three meals and a snack), levels of accuracy were considered as accuracy within 10 g/day, within 20 g/day, and within 30 g/day. Comparisons of mean A1C levels across categories of accuracy were made by analysis of variance.

All analyses were performed using SAS 9.0 software (SAS Institute Inc.,Cary, N.C.). Despite some concern for multiple comparisons, because this was a feasibility pilot study and the project was not formally powered to test hypotheses related to accuracy of carbohydrate counting, investigators simply considered a P value of < 0.05 as statistically significant.

As shown in Table 1,participants were categorized by their carbohydrate counting accuracy level as accurate (within 10 g of true amount), overestimated (by > 10 g of true amount), or underestimated (by > 10 g of true amount) based on the commonly used I:C ratio of 1:10. Participants (mean age 15.2 ± 1.8 years, 52%female, 85% non-Hispanic white, 55% using CSII) had an average A1C of 8.0± 1.0% (range of 6.2–10.8%) and an average diabetes duration of 67 ± 45 months (range 13–183 months).

Seventy-four percent of participants received carbohydrate counting instruction from an RD, although 72% of these instructions were > 1 year before participation in this study. The remainder of participants received carbohydrate counting education from their physician, nurse, books/printed materials, and the Internet (Table 1).

Participants were also asked at which meals they were most or least likely to count carbohydrates, but no statistically significant pattern was found(Table 1). Additionally,participants were asked how often they receive help with carbohydrate counting, but no statistically significant differences were found among the three carbohydrate counting categories. In other words, the participants who received the most help with carbohydrate counting did not do significantly worse in carbohydrate counting by themselves than those who always counted carbohydrate for themselves.

Statistically significant overestimation of carbohydrate content was observed for 15 of 32 foods in the quiz(Table 2). Syrup, hash browns,rice, spaghetti, and chips were overestimated by more than 5 g/serving. Statistically significant underestimation of carbohydrate content was observed for 8 out of 32 foods in the quiz, with cereal, French fries, and soda underestimated by more than 5 g. Overall, carbohydrate accuracy for an average breakfast was +7.2 ± 15.1 g; lunch was –7.9 ± 16.7 g;dinner was +3.8 ± 17.3 g, and accuracy for average daily snacks was+0.8 ± 13.6 g.

As shown in Table 3, only 11 of 48 (23%) adolescents estimated daily carbohydrates within 10 g of the true amount despite selection of common meals, and only 15 (31%) estimated accurately within 20 g/day. For dinner meals, individuals with accurate estimation of carbohydrate grams had the lowest A1C values (7.7 ± 0.8%)with A1C levels of 8.5 ± 1.2% and 7.9 ± 1.0% for those in the overestimated and underestimated groups, respectively (P = 0.04 comparing across three categories).

The findings in this pilot study indicate that adolescents with type 1 diabetes do not accurately count carbohydrates and commonly either over- or underestimate carbohydrate grams in a given meal. Adolescents with accurate(within 10 g) carbohydrate assessment had the lowest A1C values, and this was statistically significant for the dinner meals. Adolescents who overestimated the carbohydrate grams at dinnertime had the highest A1C values.

Although this initially seems counterintuitive, one might speculate that routine excess insulin bolusing because of overestimation of anticipated carbohydrate ingestion may lead to hypoglycemia, with subsequent overtreatment of low blood glucose leading to chronically higher glucose levels. Or, from a clinical perspective, if an adolescent underestimates the carbohydrate content in a given meal by 10 g, and he or she is using an I:C of 1:5, the prandial insulin dose administered would be 2 units less than actually needed, likely resulting in postprandial hyperglycemia.

From an efficacy perspective, the Diabetes Control and Complications Trial results demonstrated that adjusting insulin for meal size and content using meal-planning strategies (including carbohydrate counting, the exchange system, weighing and measuring food, and estimating portion sizes) in the intensively managed group was associated with improved glycemic control.⁸  The Dose Adjustment for Normal Eating study group in the United Kingdom found that adult patients with type 1 diabetes who were taught how to use flexible intensive insulin regimens with I:C ratios improved their A1C values by 1%after 6 months and reported improved quality of life.¹⁴  Adjusting insulin doses to match carbohydrate intake can enhance adolescents'independence and feelings of normalcy and increase communication and trust with their health care team.¹⁵ 

One of the only studies assessing carbohydrate counting accuracy in type 1 diabetic patients was reported by Rabasa-Lhoret et al.⁶  and only included nine adult patients. The investigators presented correlation data between patient and nutritional software–assisted evaluation of carbohydrate content of meals and found 85.2% of participants' carbohydrate evaluation was within 15% of the computer-assisted carbohydrate assessment.⁶ 

Despite some concern for multiple comparisons, because this was a feasibility pilot study and the project was not formally powered to test hypotheses related to accuracy of carbohydrate counting, we considered a P value of < 0.05 as statistically significant. Also, the use of a combination of foods with real foods in standard serving sizes, real foods as self-serve, and food models may have introduced some bias into the study. Additionally, food models are in a standard serving size that might be easier for participants to estimate; however, there was no difference found in accuracy between food models, real foods in standard serving sizes, and self-served foods.

No data are available assessing the accuracy of carbohydrate counting in adolescents, although this task is part of their daily diabetes self-care. The pilot study reported here includes adolescents specifically and suggests that in a larger sample of patients with type 1 diabetes, carbohydrate counting is less accurate when compared to true carbohydrate content of food.

From a practical perspective, preliminary findings from this pilot project suggest that adolescents seen for diabetes care in a specialty clinic do not count carbohydrates of commonly consumed foods with acceptable accuracy. Given the need for accurate insulin bolusing to achieve optimal glycemic control,this represents an understudied aspect of diabetes care with the potential to improve patient outcomes.

Future studies will need to address many limitations of this initial pilot project; include a larger sample size; collect data on participants' actual insulin bolusing, their postprandial glucose levels, and their daily glucose variability; and include consideration of other factors important to glucose control, such as physical activity and dietary factors other than carbohydrate intake. This study was cross-sectional and only looked at a one-time measurement of carbohydrate counting accuracy and A1C, both of which can change over time due to multiple factors. Better understanding of these relationships is needed to improve clinical care of youths with type 1 diabetes. This work will help RDs, diabetes educators, and patients self-managing their diabetes using either MDI or CSII regimens by providing data to make more informed bolusing decisions that optimally incorporate both diet and physical activity behaviors.

The authors thank the patients and their families who participated in this pilot study and all clinic staff for their help with this project. This study was supported by a pilot and feasibility grant from the Diabetes and Endocrinology Research Core P30 DK57516 and K12 DK063722-05.