Medical nutrition therapy is an integral component of diabetes management education to facilitate optimal glycemic control and prevention of complications [1]. Nutrition education for this population includes education on carbohydrate estimation as well as recommendations for general healthful eating [1,2]. Nevertheless, the diets of youth with type 1 diabetes (T1D) are characterized by patterns known to increase risk for certain chronic diseases [3]. Intake of fruits, vegetables, and whole grains are far below dietary recommendations [4,5]. Intake of total and saturated fat is above recommendations [6,7], and a substantial proportion of daily energy intake is obtained from refined grains and discretionary foods such as chips and sweets [5].
Previous research indicates the critical role of diet in promoting long-term health among persons with type 1 diabetes, including reducing risk of cardiovascular disease. Cardiovascular disease is more common, occurs earlier, and is the primary cause of premature mortality in persons with T1D [8,9]. This increased risk begins early in the disease process, with children and adolescents with T1D demonstrating subclinical cardiovascular abnormalities [9]. In observational studies among persons with T1D, better diet quality is associated with lower blood pressure [10], more optimal LDL/HDL ratio [11], and lower CVD risk profile including lower arterial stiffness [12]. In a study of youth with T1D in Italy, implementation of a Mediterranean-style diet led to improved lipid profiles [13]. Considering the high prevalence of cardiovascular risk factors observed in youth with T1D [14-18], optimal dietary intake is critical for improving long-term health outcomes among this population. The effect of diet quality on glycemic control, however, is not well-established. In an observational study of youth with T1D, better diet quality was associated with lower A1c [19], and in short-term feeding studies, better diet quality has been shown to improve glycemic control [20,21].
Despite suboptimal diet quality among youth with type 1 diabetes and the potential long-term health benefits of improving dietary intake, little previous research has addressed strategies for improving dietary intake in this population. Limited research has tested the efficacy of specific dietary recommendations on health outcomes, including evaluating the effect of a Mediterranean-style diet on lipid profile [13], the effect a low glycemic index diet on glycemic control [22], and the effect of an optimized mixed diet on dietary nutrient composition [23]. These studies utilized educational guidance only; however, it is well-established that optimal methods for achieving dietary change incorporate behavioral strategies along with educational guidance [24]. Behavioral strategies such as self-monitoring, goal-setting, problem-solving, contracting, and motivational interviewing have demonstrated effectiveness in achieving healthful dietary change in youth in the general population [24]. Additionally, achieving dietary change among youth must consider the key role played by parents, who influence youth’s dietary behavior through behaviors such as modeling eating habits and determining what foods are available and accessible in the home [25]. To date, no randomized trial of a behavioral intervention to improve dietary intake among youth with type 1 diabetes has been published. The purpose of this study was to evaluate the efficacy of a family-based behavioral intervention that integrated motivational interviewing, active learning, and applied problem-solving to increase intake of whole plant foods (fruit, vegetables, whole grains, legumes, nuts and seeds) among youth with type 1 diabetes. We hypothesized that the intervention would improve youth diet quality and glycemic control relative to the control condition.
Subjects and methods
Design and participants
This was a parallel-group study with equal randomization conducted at an outpatient, free-standing, multidisciplinary tertiary diabetes center in Boston, Massachusetts. Eligibility criteria included age 8.0 to 16.9 years, diagnosis of type 1 diabetes ≥ 1 year, daily insulin dose ≥0.5 units per kilogram, most recent HbA1c ≥6.5% and ≤10.0%, intensive insulin therapy with either an insulin regimen of ≥3 injections daily or insulin pump, at least one clinic visit in the past year, and ability to communicate in English. Exclusion criteria included daily use of premixed insulin, transition to insulin pump therapy in the last three months, real-time continuous glucose monitoring use in the last three months, participation in another intervention study in the last six months, and presence of gastrointestinal disease such as celiac disease, multiple food allergies, use of medications that interfere significantly with glucose metabolism, or significant mental illness.
Procedures
The study was conducted from August 2010 through May 2013. Medical record data were screened to identify eligible patients; recruitment was implemented by trained research staff at regular clinic visits. All youth provided assent; parents and youth turning 18 years old during the 18-month trial provided written informed consent. Randomization was stratified by age (<13 years and ≥13 years), HbA1c (<8.5% and ≥8.5%), and insulin regimen (injection and insulin pump), with a permuted block randomization scheme. Randomization was conducted by the data coordinating center; group assignment was indicated to the site research assistant by an online data management system, and families were informed of their group assignment at the second study visit.
Families were enrolled in the study for 18 months. Study visits were completed in the clinic; diet records were completed in the home following assessment visits. Youth and parents each received a total of $380 compensation for completion of all study visits and reimbursement for parking costs. Study procedures followed were approved by the Eunice Kennedy Shriver National Institute of Child Health and Human Development Institutional Review Board and the Joslin Diabetes Center Committee on Human Subjects.
Treatment conditions
The intervention content and process was guided by self-regulation perspective [26], social cognitive theory [27], and self-determination theory [28]. Each session integrated a motivational interviewing style of interaction designed to increase internal motivation for healthful eating [29,30], active learning for youth and parents to facilitate skill-building and engagement with the educational information, and applied problem-solving to facilitate goal-directed behavior and self-regulation skills. The intervention was delivered by research assistants who received training in pediatric T1D, intervention procedures, and motivational interviewing. In addition, study investigators provided feedback on audiotaped role-play practice sessions prior to intervention delivery and on a random sample of audiotaped intervention sessions.
Families in the intervention condition received six “core” sessions during the first seven months of the study period. An initial overview session addressed key principles of healthy eating, with a focus on increasing intake of whole plant foods, defined as whole fruits, vegetables, whole grains, legumes, nuts, and seeds. These food groups were emphasized due to their importance in the diet for disease prevention and consistent findings of low intake relative to dietary guidelines [31-33]. Monitoring of carbohydrate intake is central in T1D management [1,2]; the focus on these food groups encourages families to also consider the quality of their sources of carbohydrate. The next five sessions addressed application of these principles to specific eating contexts – breakfast, lunch, dinner, snacks, and eating out (one context per session). Each session included interactive education, learning activities, goal-setting, and application of the problem-solving process to increase intake of fruits, vegetables, whole grains, and/or legumes/nuts/seeds at that eating occasion. Children and parents set goals for increasing intake of two selected food groups, considered barriers, chose strategies, and developed a specific action plan for increasing their intake of the target foods. At each session, families reviewed their progress on the previous session’s goal, allowing previous efforts to inform subsequent problem-solving. Three “booster” sessions delivered during months nine to fifteen dealt with overcoming challenges associated with social eating, meal planning, and the food environment. Families were provided with a book of approximately 300 recipes highlighting the target food groups and providing detailed nutrition information to assist with insulin dosing. Recipes were selected based on consideration for ease of preparation, acceptability and familiarity. Intervention materials are summarized in the Additional file 1: Table S1 and are available upon request from the corresponding author.
Participants in both groups received intermittent, masked continuous glucose monitoring (CGM) for three consecutive days six times across the study duration, paired with completion of diet records. Following completion of each monitoring period, all subjects received individualized feedback on their CGM results with a diabetes nurse educator or certified diabetes educator. For intervention families, CGM feedback reinforced session content by addressing how glycemic patterns were associated with quality of food ingested, highlighting the effect of food choices on blood glucose levels.
The control condition was designed to match on potentially important aspects of research contact that may impact health outcomes but were not the focus of the behavioral intervention. Participants in the control condition received equal frequency of contacts with research staff, focused on case management (scheduling, confirming, and documenting medical follow-up) within the diabetes health care system in a “care ambassador” model [34], and equal frequency of three-day masked CGM use. Participants in the control condition received no additional dietary advice beyond that provided as part of standard type 1 diabetes care. Scales, measuring cups, and spoons were also provided to all participants to facilitate portion size estimation.
Measures
The child’s usual dietary intake was estimated using three-day food records. Children and parents were instructed on accurately measuring and reporting food and beverage intake and given a sample diet record. Families were instructed to keep records beginning at the time of CGM insertion and continuing for the next three consecutive days. Families were asked to use measuring utensils when at home, and if away from home, to provide their best estimate of portion size. Families were reminded to provide all specific details for each food item, including names of brands or restaurants and specific item labeling (e.g., low fat, 1% milk), and to leave no blank fields on the form. Research staff reviewed the completed records upon receipt from the family to ensure completeness, and solicited missing information (e.g., brand names) from the family as needed. For visits in which a family did not complete a diet record, 2 non-consecutive 24-hour dietary recalls were obtained by a registered dietician (1.7% of dietary assessments). Diet records were entered by two registered dietitians and verified for consistency and accuracy. Nutrition Data System for Research software (NDSR 2012; Nutrition Coordinating Center, University of Minnesota, Minneapolis, MN) was used to analyze the records and assess nutrient intake and food group servings.
Hemoglobin A1c (HbA1c) was measured using a laboratory assay standardized to the Diabetes Control and Complications Trial (reference range, 4%-6%, [20–42 mmol/mol]). Initial A1c assays were performed with the Tosoh (Tosoh Medics, South San Francisco, CA, USA) followed by the Roche Cobas Integra (Indianapolis, IN). All values obtained with the Tosoh were standardized to the Roche assay. Height, weight, insulin regimen, and frequency of blood glucose monitoring were extracted from the medical records. Demographic characteristics were assessed by parent self-report. The poverty income ratio was calculated as the ratio of reported household income divided by the 2008 US Census poverty threshold for household size and composition adjusted for inflation [35]. This measure accounts for household size when evaluating income, with a higher value indicating greater income.
Primary outcomes and power
Primary study outcomes were diet quality and glycemic control. Two indicators of overall diet quality were evaluated. The Healthy Eating Index 2005 (HEI2005) score measures conformance to the 2005 Dietary Guidelines for Americans, and is comprised of 12 component scores corresponding to dietary guidelines for intake of total fruit, whole fruit, total vegetables, dark green/orange vegetables and legumes, total grains, whole grains, milk, meat and beans, oils, saturated fat, sodium and energy from solid fat, alcohol and added sugars [36]. The maximum component score is achieved if intake meets recommended intake levels, with truncation for intakes exceeding recommendations. Recommendations and scores are expressed on a per-1000 kilocalorie basis to enable comparability and applicability to individuals regardless of total energy requirements. Component scores are summed to obtain the total score, with possible values ranging from 0–100; a score of 100 indicates meeting intake recommendations for all dietary components. Whole Plant Food Density (WPFD) is a continuous measure that represents the proportion of the diet allocated to whole grains, whole fruit, vegetables, legumes, nuts, and seeds; calculated as the total number of cup or ounce equivalents of these foods consumed per 1000 kilocalorie total intake [37]. WPFD was developed by two of the investigators to provide a measure that directly corresponds to the target food groups of the intervention.
A target sample size of 160 participants was selected based on detecting meaningful differences between intervention and control conditions in dietary intake and HbA1c at 18-month follow-up. At the time of the study development, there were no published data quantifying HEI2005 scores in a cohort of youth with T1D, and the WPFD had not yet been developed. Based on available data from 67 youth age 2 to 12 years receiving care at the same source population (mean ± SD HEI2005 57.6 ± 6.5), the sample size provided 97% power to detect a 4 point difference in the HEI2005. Power for detecting treatment effect on HbA1c was based on electronic medical record data from 560 patients at the recruiting clinic site, ages 8–16 years, with HbA1c between 7.5 and 9.5%, inclusive (mean ± SD HbA1c 8.4 ± 0.6); the target sample provided 88% power to detect a 0.3% difference in HbA1c. Given the small sample size used for power analyses of dietary outcomes, and the later development of the WPFD, power analyses were subsequently recalculated using data from a larger cross-sectional study of subjects from the same source population [5]. Using a two-sample t-test with a two-sided 5% significance level, the power at the achieved sample size of 136 was 83% to detect a difference between groups of 5.5 in HEI2005, 86% to detect a difference between groups of 0.7 in WPFD, and 93% to detect a difference between groups of 0.5% in HbA1c, assuming a common standard deviation across groups of 10.95 for HEI2005, 1.33 for WPFD, and 0.84 for HbA1c.
Analysis
Baseline demographic and disease-related characteristics of the study participants were summarized with means and standard deviations for continuous variables and frequencies for categorical/ordinal variables. Comparison of these variables between intervention and control groups was done using independent t-tests for continuous variables or Pearson chi-square for categorical variables.
Mean values for each dietary outcome variable (overall diet quality indicators and individual food groups) at each visit for each treatment condition were estimated by population ratios, which take the ratio of total food group intake to total energy intake at the population (treatment group) level; this method reduces bias in estimates of usual intakes from limited dietary assessment data [38]. The standard errors of each study outcome were estimated using bootstrap with 5000 samples with replacement. Between-group comparison of each outcome was conducted using permutation test. Five-thousand permutations of the combined samples were generated to obtain the permuted p-value, that is, the proportion of the permuted samples that yielded a more extreme difference than that observed between intervention and control conditions. Here the difference was defined as the Euclidean distance in the vector of visit-specific population ratios between intervention and control groups; the resulting p-value indicates significance of between-group differences across the study duration. Intent-to-treat analyses were conducted using multiple imputation for missing data, including missing due to subject withdrawal. Analyses were applied to ten complete data sets obtained by replacing the missing outcome values with imputation. Estimates from the imputed samples were then combined to generate a single estimate and p-value [39]. A p-value of less than 0.05 was considered statistically significant. All analyses were performed using either SAS version 9.3 (SAS Institute, Cary, NC) or R version 2.15.1 (The R Foundation for Statistical Computing).
