This study has attempted to evaluate isocaloric dietary patterns that are very low and high in carbohydrate. Although the diets were consumed under free-living conditions and nutrient intakes analyzed using food records, considerable effort was taken to ensure that this was achieved by the provision of key foods and providing very prescriptive diet information and menu plans. We noted that percentage weight loss was greater on the VLCARB diet compared to the VLF diet, providing possible evidence of a metabolic advantage. We have shown that the amount of weight loss on a VLCARB diet is greater than similarly energy restricted higher carbohydrate patterns as has been observed some decades ago in albeit lower kilojoule but isocaloric comparisons [18, 20]. Our observation that this difference was due primarily to loss of lean mass is consistent with the findings of Vasquez & Adibi  but no isocaloric dietary studies such as ours have been conducted to confirm these findings. As previously shown, the amount of fat loss was similar on all diets when the same energy restriction is applied. It is surprising that despite a higher kilojoule intake than prior studies with a consequently smaller energy restriction as well as a longer study duration, we still noted a greater effect on lean mass loss for the VLCARB pattern. Other authors have argued that, by reducing plasma insulin levels, a low-carbohydrate, ketogenic diet would spare body protein by minimizing the need for gluconeogenesis . Although we and others  did indeed observe greater reductions in both fasting and post prandial insulin responses on VLCARB, this was not associated with protein sparing. Volek et al  observed an increase in lean mass in a small study in normal weight men on a VLCARB in energy balance but a subsequent study by the same author in overweight subjects using VLCARB in energy restriction showed no greater lean mass preservation .
One of the key findings of this study was that fasting and post prandial insulin was lower on the VLCARB diet than the other two high carbohydrate diets. We believe that the provision of a glucose tolerance test as well as a "meal test" was a major strength of this study. The virtual flat line glucose and insulin response to a low-carbohydrate meal in the VLCARB group (Figure 5) is remarkable data that clearly shows how effective this dietary pattern is at stabilizing the metabolic and hormonal milieu that is the goal for people with insulin resistance and type II diabetes. The fact that the low-carbohydrate diet did not worsen, and even improved, the glucose and insulin response to 75 g of glucose further emphasizes the fact that carbohydrate-restricted diets improve insulin sensitivity provided they achieve weight loss. Dysregulation of insulin function and glucose metabolism is the hallmark of diabetes and the fact that a low-carbohydrate diet can significantly improve this aspect of metabolism is noteworthy.
Our study is unique in that, unlike previous studies using whole foods that have used an ad libitum approach in implementing the dietary strategy, we have attempted to control and match total kilojoule intake on all diets as well as introducing an energy balance period to separate the effects of diet composition and weight loss from energy restriction. Consequently we observed an increase in LDL cholesterol on VLCARB compared to a reduction on the two low saturated fat dietary patterns. This is in contrast to several previous studies [7, 13] who saw no increase in LDL cholesterol from baseline levels on VLCARB. This is most likely due to the lowering effect of weight loss on LDL cholesterol attenuating the expected rise from an increase in saturated fat intake. This effect of weight loss on LDL cholesterol has been estimated to be a reduction of 0.02 mmol/L per kilogram weight loss  whereas the increase in LDL cholesterol for every 1% energy increase in saturated fatty acids is estimated to be 0.03 mmol/L . Therefore the likely net effect on LDL cholesterol with weight loss on VLCARB will be a balance between how much weight is lost versus the increase in saturated fatty acids. For small weight losses the impact is likely to represent a net increase in LDL cholesterol whereas for moderate weight losses this effect may be neutral. However, the cardiovascular risk represented by these changes in LDL cholesterol are not clear as Sharman showed that more men with "pattern B" had switched to "pattern A" after 6 wk of intake of a very low-carbohydrate diet (75%) compared with a low-fat diet (42%) . An examination of the effect on apoB concentrations revealed no significant effect of diet composition on repeated measures ANOVA although both VLF and HUF resulted in a net lowering of apoB concentrations (P < 0.05 Students paired t test) whereas for VLCARB it remained unchanged from baseline despite a significant fall during active weight loss. This represents a balance between the rise in cholesterol-rich particles and the fall in triacylglycerol-rich particles with this diet. Volek et al have proposed a model of lipoprotein metabolism on VLCARB diets that is consistent with the observed decrease in triacylglycerols concentrations, increase in HDL-Cholesterol, and a redistribution of LDL to a larger particle size . We did observe a significant effect of gender to the apoB response to refeeding to weight maintenance (P < 0.05). This has not previously been described and suggests that in men, apoB may be more resistant to caloric flux than in women. ApoB may arguably be a better predictor of vascular risk  although there is some controversy in this area. When triacylglycerol is elevated, such as in people with type 2 diabetics, apoB (or non HDL cholesterol) is clearly superior but this may not be true in people with normal triacylglycerol levels. We also observed a diet effect on apoB48 which was unexpected and may be related to high fat diets increasing chylomicron clearance.
The greater triacylglycerol reduction on VLCARB is in keeping with what is anticipated on isocaloric lower carbohydrate patterns and also consistent with what has been observed in longer term ad libitum studies when adjusted for weight loss [11, 13, 26, 30]. Low HDL-C and hypertriglyceridemia have been shown to be independently related to the risk of myocardial infarction [31, 33]. In the Veterans Affairs High – Density Lipoprotein Intervention Trial  which used gemfibrozil, it was observed that for every 1% increase in HDL-C, there was a 3% reduction in death or myocardial infarction although not all of this effect was attributed to the effect on HDL-C. It is therefore, possible that weight loss on dietary patterns that are very low in carbohydrate and which improve these risk factors may be therapeutic for subjects with this pattern of dyslipidaemia despite much of the fat being saturated. However there is currently minimal evidence that increasing HDL cholesterol with fat is protective.
CRP is an independent risk factor for cardiovascular disease and abdominal obesity is associated with elevations of CRP . In the present study weight loss per se resulted in a reduction in CRP irrespective of dietary macronutrient composition. Reduction in CRP with weight loss has been observed previously by us in a study using very-low-fat diets and in a recent study when CRP fell irrespective of dietary macronutrient composition [35, 36]. Others al have also reported reductions in CRP on VLCARB [37, 38]. We also observed a beneficial effect on blood pressure with weight loss which was independent of diet composition. Other authors have reported blood pressure reductions with modest weight loss . There is known to be a strong positive association between systolic blood pressure and increasing risk of stroke and cardiovascular disease, and reductions in systolic blood pressure contributes to overall CVD risk reduction .
The very low insulin response observed following the VLCARB test meal was unexpected given that protein is well known to stimulate insulin secretion. However it is likely that the large amount of fat in the test meal markedly delayed gastric emptying of the protein and blunted the rise in insulin . Although apparent glucose tolerance did not change with weight loss on VLCARB, the insulin response to both glucose and the test meals was lowered suggesting improvements in insulin sensitivity.
The other cardiovascular risk factor which worsened on VLCARB was plasma homocysteine which increased by 6.6% despite no differences in plasma folate. This small increase in homocysteine may or may not have clinical significance as homocysteine levels after the VLCARB diet were low at 7.76 umol/L, which are below values that are indicative of higher risk for cardiovascular disease. Hyperhomocysteinemia was found to be an independent risk factor for cardiovascular disease in a prospective study of plasma homocysteine and risk of myocardial infarction in US physicians  although in men free of coronary disease high circulating homocysteine concentrations were not a risk factor for acute coronary events. The same study showed that folate concentrations are inversely associated with the incidence of acute coronary events. In our study, folate concentrations were measured during energy balance during which time absolute carbohydrate intake and vegetable intake were higher than during the energy restriction period. Hence it is possible that plasma folate levels were lower during the energy restriction period suggesting a need for folate supplementation on VLCARB dietary patterns. Although mandatory folate fortification does not occur in Australia, it is likely that such fortification which does occur in countries such as the US may minimize these effects.
We noted an increase in calcium excretion on VLCARB. This is in contrast to a reported previous report from our group which showed that while weight loss was associated with increased bone resorption on a higher protein diet (34% energy from protein) calcium excretion decreased . Metabolism of dietary protein (particularly fish, meat, and cheese) is associated with acid generation, which can reduce blood pH and cause obligatory calcium losses whereas metabolism of fruit, and vegetables (both of which were low in VLCARB) produces alkali, which can partially ameliorate the effect of this acid [45, 46]. On the other hand, a protein intake greater than 87 g/day is related to improved lower limb bone mass in elderly women. Calcium intake on VLCARB was significantly higher than on VLF and not different to HUF. Hence the possible adverse effects of long term use of VLCARB dietary patterns on bone mass remains speculative.