The GIOCAR (contegGIO-CARboidrati) was designed as a prospective, randomized, controlled, open-label clinical trial with a duration of 24 weeks. The study was approved by the Ethics Committee of the San Raffaele Scientific Institute in Milan and was registered at ClinicalTrials.gov (no. NCT01173991). After having received detailed information about the study and before any study procedure, participants signed a written informed consent.
We recruited adult patients with type 1 diabetes treated with CSII and followed at the CSII Outpatient Clinic of the San Raffaele Scientific Institute in Milan. We included patients with type 1 diabetes, aged 18–65 years, who had been treated with CSII for >3 months. Exclusion criteria were serum creatinine >124 μmol/L in women and >150 μmol/L in men, previous training in carbohydrate counting, celiac disease, pregnancy, severe comorbidities, and any disability preventing compliance with study procedures. Three diabetologists with experience in managing patients on CSII and trained in carbohydrate counting and one dietitian certified in carbohydrate counting conducted the study. We used the Complete Guide to Carb Counting
(2nd ed.) (14
) as a reference for carbohydrate counting. Learning carbohydrate counting involves several steps. The first step is keeping a food diary: complete food records include day of the week, meal time, amounts of food, carbohydrate grams for each food, total carbohydrate grams for the meal or snack, preprandial and postprandial (2 h after the start of the meal) blood glucose, short-acting insulin dose, and physical activity. The I:CHO tells how much insulin is needed to “cover” the amount of carbohydrates eaten and bring blood glucose level back to pre-meal target (14
). This ratio is calculated on the basis of individual recorded diary data by dividing the total grams of carbohydrates of a meal by the number of units of short-acting insulin that were able to hold post-meal glucose excursions within 1.6 mmol/L. The sensitivity factor or correction factor is calculated by dividing 1,800 by the total daily insulin requirement (14
) and corresponds to the glucose lowering obtained with one unit of short-acting insulin. By combining the I:CHO and sensitivity factor, patients are instructed to estimate the preprandial insulin dose, taking into consideration preprandial blood glucose and the amount of carbohydrates they plan to eat.
The primary outcome of the study was the change in HbA1c
at week 24. Secondary outcomes were the changes of the following variables at week 24: quality of life, assessed with the Diabetes-Specific Quality-of-Life Scale (DSQOLS) questionnaire (16
), BMI and waist circumference, hypoglycemic events (capillary glucose 2.8 mmol/L), hypoglycemia and hyperglycemia risk indexes (low blood glucose index [LBGI] and high blood glucose index [HBGI]) (17
), total daily insulin dose, and fasting plasma glucose. Participants were randomly assigned to two groups (group 1 intervention, group 2 control subjects) with a 1:1 ratio. An investigator without contact with study participants generated the treatment allocation sequence using a computerized random number generator (Stata, version 10.0; Stata Corp, College Station, TX). Because of the type of intervention, blinding was not possible. Patients were given the same glucose meter (OneTouch Ultra2; LifeScan Inc., Milpitas, CA) for self-monitoring of blood glucose during the study period and were asked to measure capillary glucose six times per day, according to American Diabetes Association Standards of Medical Care (4
). Before randomization, all participants attended a group lesson with the dietitian about the recommended diet for patients with diabetes. After randomization, patients in group 1 (intervention) were trained on carbohydrate counting and bolus calculation in the first 12 weeks using the I:CHO and sensitivity factor during four to five individual sessions with the dietitian and a diabetologist, whereas patients in group 2 (control subjects) continued estimating their pre-meal insulin dose in an empirical way. HbA1c
and fasting plasma glucose were measured at baseline and after 12 and 24 weeks. At baseline and after 24 weeks, we measured BMI and waist circumference, recorded total daily insulin dose, and asked patients to complete a validated instrument for assessing diabetes-specific quality of life (DSQOLS) (16
). Capillary glucose measurements were downloaded from the memory of glucose meters at 12 and 24 weeks at the time of the outpatient visits, and LBGI and HBGI were calculated as reported (17
). Study data were recorded on a paper Case Report Form and then entered in a dedicated database maintained in Microsoft Office Access (Microsoft Corp., Redmond, WA), and de-identified datasets were extracted for statistical analyses. HbA1c
was measured using ion-exchange high-performance liquid chromatography (DCCT-certified method) (18
), with a normal range of 3.5–6.0%.
The intention-to-treat (ITT) analysis included all randomized patients who concluded the trial, i.e., 56 patients (28 patients per group). The per-protocol (PP) analysis included 20 patients in the carbohydrate counting group and 27 patients in the control group. For this analysis, we excluded nine patients because of discontinuous use of carbohydrate counting (<75% of the meals) (six participants) or shift from CSII to multiple insulin injections for >7 consecutive days (two participants in the carbohydrate counting group, one participant in the control group).
Baseline characteristics of study participants in the two groups were compared using the χ2 test, unpaired, two-tailed t test, or Mann–Whitney two-sample statistic as appropriate. Changes from baseline of DSQOLS scores, BMI and waist circumference, total daily insulin dose, fasting plasma glucose, LBGI, and HBGI in the two groups were compared using the unpaired, two-tailed t test or the Mann-Whitney two-sample statistic, as appropriate. HbA1c levels and hypoglycemic events during the study in the two groups of participants were analyzed using mixed-effects models.