Ninety subjects with at least one cardiovascular risk factor in addition to a BMI>28 were recruited by public advertisement to participate in a clinical trial of 12 weeks duration. Seven subjects withdrew before randomization and a further 16 withdrew during the study leaving a total of sixty seven subjects that completed the study (Figure 1). Subjects were matched on the basis of age, gender, BMI and randomly allocated to one of 3 dietary intervention groups (Table 1). There was an intensive weight loss period of 8 weeks and a weight maintenance period of 4 weeks duration. The protocol and potential risks and benefits of the study were fully explained to each subject before they provided written informed consent. A schematic representation of study design is shown in Figure 2. All experimental procedures were approved by the Human Ethics Committees of the Commonwealth Scientific Industrial Research Organisation (CSIRO).

The diets were designed to be isocaloric and 30% energy restricted (~6000 kJ) for 8 weeks, followed by 4 weeks on the same macronutrient proportions but maintaining energy balance. Energy requirements of individuals were calculated and 30% energy restriction calculated accordingly The planned macronutrient profiles of the treatment diets were as follows: Very Low Fat (VLF) (CHO:Fat:Protein; %SF = 70:10:20; 3%), High Unsaturated Fat (HUF) = (50:30:20; 6%), VLCARB (4:61:35; 20%). Templates for the dietary patterns were developed for 5.5 MJ, 6 MJ, 7 MJ and 8 MJ and these served as a basis for minor individual adjustments.

Key foods for each diet (2200 kilojoules or 36% of total energy) representative of the diets' macronutrient profile were supplied every 2 weeks for the 12 wk study. Foods provided were uncooked but pre-weighed to facilitate compliance. The dietary pattern was structured to include specific daily quantities of foods to ensure the correct macronutrient and energy requirements (Table 2A). These foods were listed in a checklist which subjects completed on a daily basis. Detailed dietary instruction, meal planning and recipe information was provided at baseline and every 2 weeks by a qualified dietitian. Checklists of prescribed foods and additional foods were checked, and 3 d weighed food records were collected every 2 wks to assess dietary compliance up to the 12 wk time point. At week 8, adjustment of energy intake was titrated upwards to restore the daily energy deficit based on the calculated daily deficit assessed from the previous 4 wks weight loss. Standard meal plans for each diet were developed for a range of energy levels in 1000 kJ increments which served as the basis for the revised energy balance prescription. Subjects were counselled by a dietitian on the dietary protocol and on how to keep dietary-intake checklists for all foods consumed each day over the study duration. The subjects' body weight and dietary-intake checklist were monitored every 2 weeks and dietary adjustments were made if necessary. Three consecutive days (one weekend and 2 weekdays) of the checklists from each 2-week period were analyzed by "Diet/1 Nutrient Calculation" software (Xyris Software 1998, Highgate Hill, Australia), a computerized database of Australian foods. Recipes were entered as proportions of the original ingredients. The database had been extensively modified by our group to add new foods and recipes. A questionnaire with numerical rating scale was used to assess diet acceptability on a range of parameters including palatability, ease of adherence, effect on hunger, fullness, nausea and desire to eat, and cravings.

Blood samples were taken according to the schedule as per Figure 2. Venous blood samples were taken in the morning after an overnight fast for determination of plasma glucose, insulin, ketones and lipid concentrations. Fasting serum was collected, and stored at -20°C until the end of the study. All lipid assays were performed in a single run at the end of the 12-week study on a Roche Cobas-Bio centrifugal Analyser using standard Roche enzymatic kits. HDL cholesterol was measured after PEG 6000 precipitation of apoB containing lipoproteins. Fasting lipids were taken on two consecutive days and then values averaged at each of the timepoints. Coefficients of variance (CV) for all assays were less than 5% with the exception of insulin for which the CV was 9.75%. Total apoB (B100 and B48) was measured by immunoturbidimetry using Roche antisera. Apo B48 levels were determined directly in serum using a Western Blotting procedure as previously described by James et al 15. Following visualisation using enhanced chemiluminescence (Amersham, Little Chalfont, UK) apo B48 bands were identified and quantified by densitometry against purified apo B48 protein of known mass using NIH Image (version 1.6.3). The mean intra- and inter-assay CV for apo B48 were each less than 4% Plasma levels of ketones were assayed using an enzymatic method using the principle that β hydroxybutyrate in the presence of NAD is converted to acetoacetate and NADH by β-hydroxybutyrate dehydrogenase. The NADH produced was quantified spectrophotometrically on a Cobas-Bio centrifugal Analyser measured by absorbance at 340 nm.

At weeks 0 and 12, a single venous blood sample was taken for the determination of homocysteine, folate and B12 which were measured in a certified commercial laboratory (Institute of Medical and Veterinary Science, Adelaide, South Australia). Serum CRP concentrations (CV 3.5%) were measured in duplicate at baseline and at week 12 with an ultrasensitive ELISA (Alpha Diagnostica). Serum insulin was measured by radioimmunoassay (Pharmacia & Upjohn Diagnostics AB, Uppsala, Sweden) while CRP was measured by immunoturbidimetry using Roche antisera.

A 24-hour urine sample was collected for the assessment of the urea/creatinine ratio, calcium, sodium and potassium excretion, as well as deoxy-pyridinoline/creatinine and pyridinoline/creatinine ratios (biomarkers of bone turnover) were also assessed from the 24-hour urine sample at weeks 0 and 12. Urine samples were frozen, and urea and creatinine was measured in one run at the end of the study using a Hitachi auto analyzer (Roche, Indianapolis, USA). Urinary pyridinium crosslinks (markers of bone turnover) were measured using HPLC. Urine samples were measured at the Institute of Medical and Veterinary Science, Adelaide, South Australia) for calcium, phosphate and sodium using proprietary techniques on the Olympus AU5400 chemistry analyzer (Japan).

Also at weeks 0 and 12, measurements of body composition and a 75 g oral glucose tolerance test (GTT) and on a subsequent day a 3-hour meal tolerance test (MTT) using meals that were representative of the diet to which the subjects were assigned (Table 2B), were performed. Venous blood samples for the determination of glucose, insulin, free fatty acid concentrations were taken prior to consuming the test meals as well as at 30, 60, 120 and 180 minutes after the meal.

The homeostatic model assessment (HOMA) was used as a surrogate measure of insulin sensitivity and was calculated as [fasting serum insulin (mU/L) × fasting plasma glucose (mmol/L)/22.5] (16). Total glucose, insulin and area under the curve during the 3-hour GTT and MTT was calculated geometrically using the trapezoidal rule 17.

Body composition was determined by whole body DEXA using a Norland Densitometer XR36; (Norland Medical Systems, Fort Atkinson, Wisconsin, USA; CV of 2.3 ± 0.7% for total body fat mass and 2.1 ± 0.4% for lean mass) at baseline (prior to commencement) and at 12 weeks. Blood pressure was measured using an HDI/Pulsewave™ instrument (Hypertension Diagnostics inc. Minnesota, USA).

Statistical analysis was performed using SPSS for Windows 10.0 software (SPSS Inc, Chicago, USA). Baseline measurements were assessed using two-factor ANOVA with diet and gender as the fixed factors. The effect of the diet intervention was assessed using repeated-measures ANOVA; for each dependent variable, the measurements at weeks 0, 4, 8, and 12 are the within-subject factor (i.e. time) and diet and gender are the between-subject factors. Week 0 and 12 response curves following the GTT and test meals were compared using repeated measures ANOVA with week and blood sampling time (or AUC) as the within-subject factors and diet as the between-subject factors. When significant time-by-diet effects were found, post hoc sub-group analysis was performed using Tukey's test. The study had 80% power (α = 0.05) to detect differences between dietary groups of 3.6 kg in body weight, 0.9 kg in lean and fat mass, 3 mU/L in fasting insulin and 0.2 mmol/L in LDL-cholesterol. Significance was set at P < 0.05. All data except baseline characteristics are presented as means ± SEM, unless stated otherwise.

Reported dietary intake was consistent with the prescribed dietary treatments (Table 3). Compliance to dietary treatment was also confirmed by a change in plasma ketones between diets. VLCARB produced higher plasma levels of ketones (β hydroxybutyrate and acetoacetate) than the VLF or HUF diet treatments (P < 0.01), indicating adherence to a very low carbohydrate intake during the study (Figure 3). Despite continued apparent compliance to the diet plasma ketones declined with time.

Each treatment group reduced weight over the 8 wk energy restriction period and maintained this weight during the subsequent 4 wk period (Figure 4). There were no significant differences in absolute weight loss by diet composition, with a net weight loss of 8.0 ± 0.6 kg (n = 24), 6.7 ± 0.7 kg (n = 22) and 6.4 ± 0.6 kg (n = 21) on the VLCARB, VLF and HUF diets respectively (P = 0.18). However, percentage change in weight from baseline differed significantly by diet (P = 0.034) with the VLCARB diet resulting in a greater weight loss of 9.2% compared to the VLF (7.3%) and HUF (7.0%). After applying Tukey's post hoc test however, only HUF remained significantly different to VLCARB (P = 0.044).

DEXA data indicated both the VLCARB and VLF diets resulted in significantly more lean mass loss as a proportion of weight loss (32% and 31%) compared to the HUF diet (21%) (P < 0.05) whereas the proportion of fat loss did not differ between diets (Table 4). DEXA data also indicated no significant differences in regional fat or lean mass loss between diets.

There was a significant main effect of diet on LDL-C with a net increase of 0.18 ± 0.18 mmol/L on the VLCARB diet, but net decreases of 0.40 ± 0.11 mmol/L on the VLF and 0.34 ± 0.14 mmol/L on HUF (p = 0.008 unadjusted and p = 0.006 adjusted for weight loss) (Table 5). However, the effect of diet composition on apoB concentrations was not significant (p = 0.418) although concentrations declined with weight loss overall (p = 0.011). ApoB was unchanged in the VLCARB group. Diet significantly affected apoB48 (p = 0.05) but not after adjusting for weight loss (p = 0.11), increasing on VLF and decreasing on VLCARB and HUF (data not shown).

Diet composition significantly affected the change in HDL-C with or without correcting for weight loss (p = 0.023 unadjusted and p = 0.029 adjusted for weight loss) with an increase on VLCARB (0.06 ± 0.03 mmol/L) whereas the other diets resulted in similar net decreases of 0.06 ± 0.03 mmol/L.

Similarly diet had a significant lowering effect on TG (p = 0.001 unadjusted and p = 0.002 adjusted for weight loss) with VLCARB having the greatest TG reduction (-0.73 ± 0.12 mmol/L) followed by the HUF diet (-0.15 ± 0.07 mmol/L) and VLF the least change (-0.06 ± 0.13 mmol/L).

Five subjects had C reactive protein (CRP) >15 mg/L at baseline or at the completion of the study and were excluded from the analysis. All diets resulted in a significant decrease in CRP with weight loss, independently of diet (p = 0.037).

The main effect of weight loss or diet composition on changes in plasma folate failed to reach statistical significance (p = 0.106 and p = 0.09 respectively) whereas plasma homocysteine increased 6.6% on VLCARB, decreased 6.8% on the VLF and remained unchanged on the HUF diet (P = 0.026 for diet effect) (Table 5). Increases in homocysteine concentrations were observed in 16/24 subjects on VLC, 10/22 on VLF and 10/21 on HUF. There were no changes in plasma B12 levels over the course of the study and levels remained static at 266 ± 13 pmol/L.

There was no significant effect of diet composition on blood pressure changes with weight loss resulting in a net reduction in blood pressure of 7 ± 2 mmHg systolic and 3 ± 1 mmHg diastolic blood pressure.

Fasting glucose decreased with weight loss by 2% (p = 0.036) independently of diet composition (p = 0.10) (Table 5). However, diet composition significantly affected fasting insulin levels (main effect of diet p = 0.004 unadjusted and p = 0.006 adjusted for weight loss) with the VLCARB diet lowering fasting insulin concentrations by 33%, HUF by 19% whereas it increased 15% on VLF (Table 5).

At baseline, glucose tolerance (as assessed by the total area under the curve) was not different according to diet allocation (P = 0.552) but the insulin response to glucose was significantly different (P = 0.038) with a greater insulin AUC in the VLCARB group compared to the other two groups (P < 0.05). There was a significant effect of diet on the test meal glucose response (P = 0.016) (Figure 5) with the VLCARB meal provoking a lower glucose response than the VLF meal (P = 0.014 post hoc analysis) and the HUF meal (P = 0.054 post hoc analysis). This effect was strengthened if adjustment was made for the differences in baseline insulin AUC as a covariate (P = 0.005). The VLCARB meal also induced an insulin response that was substantially lower compared to HUF and VLF meals (both P < 0.001 on post hoc analysis). Weight loss on the diets resulted in improvements in glucose tolerance in subjects allocated the VLF and HUF diets (P < 0.05) whereas no changes in glucose tolerance were observed in the VLCARB group (Figure 6). However, the insulin response to the glucose challenge was significantly lowered after weight loss on VLCARB (P = 0.016) but the small reductions observed on VLF and HUF diets were not statistically significant. Weight loss provoked a lower insulin response to the test meals (P < 0.05).

Calcium excretion increased 25% on the VLCARB diet yet decreased 12–16% on both the VLF and HUF (Table 6). Sodium excretion was not significantly different by time or diet whereas potassium excretion increased only on VLF (P < 0.001). Markers of bone turnover Dpr:Cr and Pyr:Cr increased significantly overall by 15% with weight loss (P < 0.001), but no specific diet composition effects were observed.