How Low Do You Need to Go? Exploring the Variability in the Effects of Keto‐induction and Individual Outcomes From Low‐carbohydrate Diets

Harvey, Cliff
Schofield, Grant
Zinn, Caryn
Thornley, Simon
Item type
Degree name
Doctor of Philosophy
Journal Title
Journal ISSN
Volume Title
Auckland University of Technology

Low-carbohydrate diets (LCDs) and very low-carbohydrate ketogenic diets (VLCKDs) are increasingly used for the management of a range of health conditions and in the general population for weight-loss and maintenance. However, there is little evidence for the superiority of greater carbohydrate restriction compared to more moderate restriction for most outcome measures, but potential benefits for some, like glucose, triglycerides (TG), and high-density lipoprotein cholesterol (HDL-c). There are also specific benefits from VLCKDs and the resultant ketonaemia, including reduced inflammation, inhibited tumour growth, amelioration of neurodegeneration, and increased metabolic flexibility in athletes. Despite the popularity and benefits of VLCKDs, there is little consensus on what constitutes nutritional ketosis (NK) for clinical purposes, and there is an almost complete lack of research on the time taken to achieve NK of ≥ 0.5 mmol/L beta-hydroxybutyrate (BOHB) and the symptoms of carbohydrate withdrawal commonly described in mainstream media as ‘keto-flu’ that occurs during keto-induction. Dietary supplements and methods to improve ketonaemia, time-to-NK, and symptoms of carbohydrate withdrawal and mood during keto-induction are similarly not well understood. This thesis begins with a narrative review of supplementary methods purported to increase ketonaemia and improve time-to-NK and symptoms of carbohydrate withdrawal. It provided, for the first time, a synthesis of research related to the time it takes for people to achieve NK and highlighted that there were no studies that had specifically evaluated adverse effects specifically during keto-induction. This review showed that there is a clear ketogenic effect of supplemental medium-chain triglycerides (MCTs) and a possible greater effect resulting from shorter-chain fatty acids such as butyric acid. However, the ketogenic effect of other supplements was unclear. To understand the effect of increased ketonaemia on time-to-NK, symptoms of carbohydrate withdrawal, and mood, a randomised controlled trial comparing the use of MCTs in a ‘classic’ ketogenic diet providing a 4:1 lipid to non-lipid ratio of total energy vs a control oil rich in long-chain triglycerides (LCT) was then conducted. MCT resulted in higher BOHB at all time-points, and faster time-to-NK, but these results failed to reach significance. The magnitude of symptoms of keto-induction were greater in the LCT control group, except for abdominal pain, which occurred with greater frequency and severity in the MCT-supplemented diet. There was a possibly beneficial effect on symptoms by MCT but the effect on mood was unclear. Based on these results, there was a clear effect of MCTs on ketonaemia compared with LCT and a likely reduction in symptoms of carbohydrate withdrawal. Little research has been conducted on the qualitative experience of diet, and to deepen the understanding of the effects of carbohydrate restriction on individuals’ mood and experiences, daily diary entries and focus group findings from the previous study were qualitatively analysed to illustrate the ‘lived experience’ of a following a VLCKD. Despite some challenges, especially gastrointestinal effects, the overall perception of the diet was positive, and benefits for wellbeing, mood, sleep, and sugar cravings were reported, with negative experiences decreasing as participants adapted to the VLCKD. These findings suggested that the overall experience of a VLCKD is positive but varies markedly between individuals. The preceding study outcomes suggested that increased ketonaemia might positively affect symptoms of carbohydrate withdrawal during keto-induction, and mood, but it is unclear whether diets differing in carbohydrate content and resulting in differing levels of ketonaemia would elicit similar effects. The final study of this collective body of work was a randomised clinical trial comparing a VLCKD, LCD, and moderate-low-carbohydrate diet (MCD) consisting of 5%, 15%, and 25% of total energy (TE) from carbohydrate respectively, over 12 weeks. The first three weeks of this study was used to compare ketonaemia, symptoms of carbohydrate withdrawal, and mood between the dietary intervention groups. In 75 of 77 initial participants included for analysis, mean serum levels of BOHB were increased by 0.27 ± 0.32, 0.41 ± 0.38, and 0.62 ± 0.49 mmol/L for the MCD, LCD, and VLCKD respectively (p = 0.013). The achievement of NK was consistent for both VLCKD and LCD groups and sporadic for the MCD group. The overall mean change in symptoms scores was trivial (0.81 ± 2.84, p < 0.001) and while symptoms were increased most in the VLCKD group (1.49 ± 2.47), compared to LCD (0.65 ± 2.70), and MCD (0.18 ± 3.3) the differences were small and did not reach the threshold for significance (p = 0.264). Only halitosis (p = 0.039) and muscle weakness (p = 0.005) differed significantly between the groups with the largest effects seen in the VLCKD group. Mood improved significantly from baseline overall, but there was no significant difference between groups (p = 0.181) Interestingly, although participants were instructed to maintain habitual energy intake, energy restriction did occur, and it was more strongly associated with the magnitude of carbohydrate withdrawal effects than any other factor. Therefore, these findings suggest that reduced carbohydrate diets should be prescribed according to the projected benefits to the individual, rather than the desire to mitigate symptoms of carbohydrate withdrawal and that energy sufficiency could be a more important consideration for the avoidance of adverse effects of carbohydrate-restricted diets. There were only small differences observed between symptoms of carbohydrate withdrawal and mood between the diets ranging from 5-25% TE from carbohydrate and outcome measures were also analysed to determine which diet was most effective overall. In completers of the 12-week study, significant reductions in TG, weight, and body mass index occurred, along with increases in HDL-c, low-density lipoprotein cholesterol (LDL-c) and total cholesterol concentrations. It was more difficult for those in the VLCKD group to achieve the carbohydrate allocation of 5% of TE (mean: 7.9%, SD = 4.9%), whereas the MCD (22.5%, SD = 4.5%) and LCD (14.1%, SD = 3.2%) groups adhered to the allocation. Despite this, the positive effect on markers of health trended towards greater improvement from greater carbohydrate restriction. The largest improvements in HDL-c and TG, and anthropometric changes occurred in the VLCKD group. However, between-group changes were not significant. Over the 12-week period, adherence to the prescribed carbohydrate intake was less than allocation for both the MCD and LCD groups and was higher than the allocation for the VLCKD group. A linear trend was observed for reduction in carbohydrate intake as a proportion of TE for MCD relative to week (Beta = -0.137, p = 0.24). Conversely, increased intake by week was observed for LCD (Beta = 0.096, p = 0.24), and VLCKD (Beta = 0.174, p = 0.15) but these trends were not significant within groups, or between group allocations (p = 0.108). There has been the suggestion that outcomes from lower- or higher-carbohydrate diets might be predicted by baseline cardiometabolic indicators such as insulin homeostasis but there is little consensus on the use of blood or anthropometric measures for dietary prescription. Adverse effects, mood, and outcome measures differed by only a small amount between the LCDs but there was considerable variation between individuals. Baseline cardiometabolic measures were compared to changes in these measures, relative to carbohydrate allocation. Participants with ‘poorer’ baseline measures benefitted most from greater carbohydrate restriction, with 7 of 11 measures improved most by a VLCKD, relative to baseline measurements. However, only HDL-c reached between-group significance, with every 1 mmol/L lower HDL-c at baseline associated with a 0.5 and 0.2 mmol/L improvement in HDL-c for the MCD and LCD groups respectively, compared to a 0.4 mmol/L decrement for VLCKD (p = 0.0006). This study was the first to evaluate the use of baseline measures as predictors of outcomes resulting from differing carbohydrate-restricted diet interventions. Although the effects were equivocal, the findings do suggest that those with poorer baseline measures of cardiometabolic health might benefit most from greater carbohydrate restriction. There are several novel findings from the studies presented in this thesis.

  1. Medium chain triglycerides resulted in fewer overall symptoms of carbohydrate withdrawal when compared to a substitution oil rich in long-chain fatty acids.
  2. The degree of dietary restriction of carbohydrate only had a trivial effect on increasing symptoms of carbohydrate withdrawal.
  3. Very-low-carbohydrate diets were typically tolerated well and resulted in a range of self-reported health benefits, but results varied considerably between individuals.
  4. There was a trend towards small improvements in overall benefits to cardiometabolic measures of health from a greater restriction of carbohydrate.
  5. The benefits from greater restrictions of carbohydrate appeared to be greatest for those with poorer baseline measures of cardiometabolic health.
Keto-flu , Ketosis , Low-carbohydrate diet , Ketogenic diet , Health , Nutrition
Publisher's version
Rights statement