Participants with type 2 diabetes were recruited by study personnel from a diabetes clinic and from local community advertisements. Participants were aged 30–75 years, with a BMI of 27–40 kg/m2 and HbA1c levels of 6.5–10%, and were excluded if they had significant heart disease (unstable angina, cardiac failure, or recent myocardial infarction or coronary intervention), stroke within the previous 3 months, renal disease (proteinuria or serum creatinine >0.13 mmol/l), liver disease, or malignancy. Informed consent was obtained from all participants and the study was approved by the International Diabetes Institute’s Human Ethics committee.
This single-centre, randomised controlled trial was conducted at the Baker IDI Heart and Diabetes Institute (Melbourne, Australia) between May 2005 and December 2006. Participants were randomised in a 1:1 ratio to either a high-protein (HP group) or high-carbohydrate (HC group) diet for a period of 12 months. Randomisation was carried out by a third party using computer-generated random numbers (using block randomisation and random block sizes) and stratified according to diabetes treatment (diet alone, oral medication, or insulin; see Table 1). Study personnel who enrolled participants were blinded to the sequence allocation. Dietary assignment was performed by a third party on the day of the initial dietary counselling visit.
The trial involved visits every 3 months to assess study outcomes. The primary endpoint of the study was change in HbA1c. Predefined secondary endpoints included changes in weight, waist circumference, lipid profile (total cholesterol, LDL-cholesterol, HDL-cholesterol and triacylglycerol), blood pressure (systolic and diastolic), renal function (estimated GFR [eGFR] and albumin excretion rate) and calcium loss. It was calculated that 46 participants per group would provide 80% power (at the two-sided 5% level) to detect a difference of 0.5% in the HbA1c levels between groups, assuming a standard deviation of 0.85%. To compensate for participant withdrawal, 108 participants were recruited.
This study was carried out using an intention-to-treat model following delivery of dietary advice. Participants who were randomised but withdrew consent before receiving dietary instruction were not included in the intention-to-treat analysis. To maintain the intention-to-treat integrity of the study design and to minimise the amount of missing data, study staff encouraged all participants to return for follow-up assessments of primary and secondary outcomes, regardless of dietary adherence.
A qualified dietitian administered the diet-specific advice to each study participant. The study consisted of two dietary periods: a 3 month energy restrictive period (∼6,400 kJ/day or 30% energy restriction), followed by 9 months of energy balance. The HP diet consisted of 30% energy intake from protein (a combination of lean meat, chicken and fish) and 40% energy intake from carbohydrate. The HC diet consisted of 15% energy intake from protein and 55% energy intake from carbohydrate. The diets were matched for total fat (30% of energy) and the fatty acid profile (7% saturated fat, 10% polyunsaturated fat, 13% monounsaturated fat). Consistent with current guidelines , both diets recommended carbohydrates of low glycaemic index. In addition to regular face-to-face dietary counselling appointments, written materials were supplied to both groups containing information on the key nutrition intervention messages, prescriptive fixed menu plans and food choice lists.
Dietary intakes were calculated from weighed/measured food records collected at baseline (5 days) and every 3 months during the intervention period (1 day/month) using Australia-specific dietary analysis software (Foodworks; Xyris Software, Highgate Hill, QLD, Australia). Dietary compliance was monitored by self-reported food intakes and 24-h urine samples for an assessment of urea excretion as a marker of protein intake. At 12 months, participants were asked to rate their dietary self-management on a Likert scale. This was the first section of a validated diabetes-specific questionnaire relating to dietary satisfaction and quality of life. Details regarding the questionnaire are described elsewhere .
Behavioural therapy and physical activity recommendations
Both groups were offered the same level of behavioural therapy and physical activity recommendations. The behavioural therapy consisted of individual appointments that were designed to monitor the participant’s progress and provide individualised feedback to the participants. This consisted of four visits during the 3 month energy restrictive period (totalling 2.5 contact hours), and five visits during the 9 months of energy balance (totalling 2.5 contact hours). In addition, group meetings were held every 3 months to help reinforce the principles of the dietary programmes (total contact time 3.3 h). Group topic modules included healthy cooking, goal setting and problem solving, physical activity, and supportive counselling. Physical activity was encouraged as a strategy to increase energy expenditure for those without limitations or complications, and the recommendations were consistent with public health guidelines. Physical activity was measured using the validated Active Australia survey .
At all visits, study personnel recorded information regarding concomitant medications (drug name, frequency and dosage). All medication changes were made independently of the study personnel by the treating physician.
Baseline assessments of study outcomes were done at the initial informed consent appointment, approximately 1 week before randomisation/dietary counselling. Weight and waist circumference were measured every 3 months following dietary counselling. All participants were weighed in light clothes and weight was used to calculate BMI. At 6 and 12 months, a trained assessor, who was blinded to the dietary assignment of the individual, measured and recorded weight and waist circumference. Blood pressure was measured at baseline, 3, 6, 9 and 12 months with a mercury sphygmomanometer.
Fasting blood samples and 24-h urine specimens were collected at baseline, 3, 6, 9 and 12 months for the assessment of HbA1c, lipids, eGFR, urinary albumin and urinary calcium. All biochemical analyses were performed by an independent laboratory. HbA1c was determined using immunoturbidimetric spectrophotometry with the Roche Integra 800 analyser (Roche Diagnostics, Castle Hill, NSW, Australia). Urinary albumin was measured by nephelometry with the Beckman Coulter Image system (Beckman Coulter Instruments, Gladsville, NSW, Australia). Serum creatinine (for eGFR), urinary calcium and lipids were measured by spectrophotometric methods on the Roche Modular Analyser (Roche Diagnostics).
This study used the single imputation method of last measurement carried forward for missing data for primary and secondary outcomes. This method was chosen because of the high rate of follow-up at each time point (see Fig. 1) and as this method was thought to provide a close approximation to the unobserved value.
All statistical analyses were performed using SPSS 17.0 for Windows. Dietary data and study outcomes were analysed by repeated measures ANOVA incorporating data from all time points to explore the effects of time, dietary group allocation and the time course between groups (time × group interaction). Triacylglycerol, HbA1c, HDL-cholesterol and urinary albumin were loge-transformed as the data were loge-normally distributed. Blood pressure and medication changes were analysed using non-parametric tests to test for differences between groups. Post hoc analyses (multifactorial repeated measures ANOVA and repeated measures analysis of covariance [ANCOVA]) were performed to also account for differences between sexes and medication changes. To ensure that any observed changes were not the result of multiple comparisons, exploratory within- and between-group comparisons were performed at the end of the weight loss and weight maintenance phases and interpreted with the use of the Bonferroni adjustment. A critical α value of p < 0.0025 was used for the evaluation of the 20 dietary comparisons (Table 2) and p < 0.0019 for the 26 study outcome comparisons (Tables 3 and 4).
As this study involved changes to a number of dietary variables (i.e. intakes of calories, protein and carbohydrate), subsidiary correlation analyses were performed to identify whether study endpoints were a function of the change in specific dietary variables. The regression analysis was performed for the per protocol population after pooling data from both groups. These tests were interpreted marginally as there was no formal adjustment of the overall type 1 error rate and the p values serve principally to generate hypotheses for validation in future studies.