Effect of a Ketogenic Diet on Submaximal Exercise Capacity and Efficiency in Runners

David M. Shaw; Fabrice Merien; Andrea Braakhuis; Ed Maunder; Deborah K. Dulson


Med Sci Sports Exerc. 2019;51(10):2135-2146. 

In This Article


Study Design

This study was conducted during the maintenance training phase for all participants. Participants underwent two 31-d experimental conditions (KD or their habitual mixed diet (HD)) with a testing block immediately before and during the final 3 d of the intervention in a randomized (www.randomizer.org), counterbalanced, crossover design with a 14- to 21-d washout period between dietary interventions. A schematic overview of the study design is shown in the Supplemental Content (Figure, Supplemental Digital Content 1, Overview of study design, http://links.lww.com/MSS/B596).


A sample size of 8–10 participants was established a priori based on previous investigations detecting physiological adaptations throughout a 2- to 4-wk dietary training intervention period.[11,18,19] Participants were required to have been 1) habitually consuming a mixed diet for >12 months, 2) weight stable for >1 month, 3) running >50 km·wk−1, and 4) able to run a marathon in <3.5 h. Participants were excluded if they 1) had a history of fat- or keto-adaptation, 2) previously ingested ketone supplements, 3) were currently or recently injured, 4) experienced moderate-to-severe gastrointestinal symptoms or illness within the previous 4 wk, 5) had a history of irritable bowel syndrome, 6) habitually smoked, or 7) had been ingesting dietary supplements or medications known to effect performance within the previous 2 wk, with the exception of caffeine, protein, and CHO supplements. Ten eligible healthy, trained, male endurance athletes (two marathoners, five ultramarathoners, and three long-distance triathletes) volunteered to participate in the study. Participants were comprehensively screened by an experienced registered dietitian (RD) after their expression of interest. Participants were fully informed of the rationale of the study and possible risks of the experimental procedures before providing their written consent; however, they were not informed of the potential performance effects of a KD and were requested to refrain from personal investigation to prevent biasing their results. The study was approved by the Auckland University of Technology Ethics Committee (Auckland, New Zealand).

Testing Block

Metabolic and body composition testing (days −2 and 29). Participants presented to the laboratory between 0600 and 0700 h having fasted from 2300 h the previous day and abstained from caffeine, alcohol, and strenuous exercise for the previous 24 h. Participants' body mass (shorts only), height, and sum of 8 (Σ8) skinfolds were measured by an accredited anthropometrist (ISAK, level 1, skinfold coefficient of variation of 3.8%). To determine V̇O2max, maximal aerobic EE, MFO, and the relative intensity at which MFO occurs (Fatmax), participants performed a graded metabolic test to volitional exhaustion on a motorized treadmill (h/p/cosmos, Nußdorf, Germany). Participants ran for 3-min stages at 7.5, 9, 10.5, 12, 13.5, and 15 km·h−1 against a 1% gradient to simulate the energetic cost of level-gradient outdoor running.[20] If RER ≥1.0 at 13.5 km·h−1, the 15-km·h−1 stage was excluded. Expired gas was collected and analyzed continuously using a computerized metabolic system with mixing chamber (Parvo Medics TrueOne 2400, Salt Lake City, UT), with the final 30 s of each stage averaged to calculate V̇O2 and V̇CO2. The second ventilatory threshold (VT2) was determined using the V-slope method.[21] RPE (Borg 6–20 scale[22]) and heart rate (HR) using short-range telemetry (Garmin Fenix 3; Garmin, Kansas City, KS) were recorded during the final 30 s of each stage. After the completion of the final stage, treadmill speed was reduced to 11 km·h−1 and subsequently increased by 0.5 km·h−1 every 30 s until the attainment of volition exhaustion. V̇O2max was determined by averaging the highest 30-s period and accepted if there was at least a 30-s plateau in V̇O2 and one of the following two criteria: 1) RER >1.10 or 3) HR ± 10 bpm age-predicted maximum (220 − age). Simple regression equations were used to estimate the speed required to elicit 70% V̇O2max for the use in the run-to-exhaustion trial, until which participants were requested to refrain from strenuous exercise (~48 h).

Run-to-exhaustion trial (days 0 and 31). Participants presented to the laboratory between 0600 and 0700 h having abstained from caffeine and alcohol for the previous 24 h. After the metabolic test on either day −2 or 29, participants were requested to ingest a high-CHO diet and consume adequate fluid to replicate their typical race preparation. Participants ingested a prescribed breakfast containing 2 g·kg−1 of CHO 90 min before arrival to replicate typical race-day nutrition. In the post-KD trial, participants continued ingesting a KD after the metabolic test and were prescribed an isoenergetic LCHF breakfast (<10 g CHO). Participants collected their first morning void to measure hydration status via urine specific gravity with a preexercise hypohydration threshold set at >1.025. Soon after arrival, an indwelling intravenous teflon catheter (18G; Terumo, Tokyo, Japan) was inserted into the antecubital vein for serial venous blood sampling. After 10 min of rest, body mass (shorts only), and venous and capillary blood samples were collected for determination of serum glucose, and blood D-βHB and lactate concentration. Participants then commenced running at the speed estimated to elicit 70% V̇O2max until volitional exhaustion (120–135 min after breakfast). Treadmill speed was matched during the prediet and postdiet trials. Participants were prescribed 4 mL·kg−1 of a 7.2% CHO-electrolyte drink (4:1 glucose-to-fructose ratio; Replace, Horleys, New Zealand) every 20 min, which was adjusted based on each participant's tolerance and replicated in their postdiet trial. During the post-KD trial, participants received a combination of artificially sweetened fluids (electrolyte drink and cordial), water, and coconut oil (100% energy derived from fat; Blue Coconut Oil, Blue Coconut, New Zealand) at a rate reciprocating the fluid and energy ingested during the pre-KD trial. Expired gas was collected for 4–5 min every 30 min and at exhaustion, with the final 1 min averaged to calculate V̇O2 and V̇CO2, alongside RPE[22] and HR (Garmin Fenix 3; Garmin). Venous and capillary blood samples were collected at 60 and 120 min to determine serum glucose, and blood D-βHB and lactate concentration. The cannula was flushed with 3–4 mL of saline every 30 min to maintain patency. On attainment of volition exhaustion, treadmill speed was reduced to 4.4 km·h−1 for 2 min, then restored to the speed eliciting 70% V̇O2max until the participant indicated volitional exhaustion. This process was repeated, so at the third attainment of volitional exhaustion, the test was terminated. The walking time was excluded from the total TTE. This protocol has a lower coefficient of variation for measuring exercise capacity compared with traditional single exhaustion protocols (5.4%, 1.4%–9.6% (95% confidence intervals, or CI)).[23] Stimulatory aids (television, music, and conversation) were provided until the end of the first exhaustion phase to reduce boredom. Immediately after exercise cessation, venous and capillary blood samples were collected to determine serum glucose, and blood D-βHB and lactate concentration. Participants removed wet clothing and towel dried themselves before measuring their body mass. All trials were conducted by the same researcher, and standardized encouragement was provided. To ensure maximal effort in each trial, a substantial monetary incentive was awarded to the participant who accumulated the highest TTE after completion of the four trials. Participants were blinded to elapsed time during each trial and were not notified of their results until study completion. Participants were provided with 5 mL·kg−1 of water during the first hour of recovery, after which venous and capillary blood samples were collected to determine serum blood glucose and blood D-βHB concentration.

Dietary Intervention and Monitoring

Participants commenced their dietary allocation (KD or HD) immediately upon completion of their initial run-to-exhaustion trial (day 0), as theoretically the ability to restore depleted muscle and hepatic glycogen led to the rapid differentiation between the two dietary conditions. To ensure immediate and ongoing compliance to the KD, participants undertook a comprehensive education session with an RD after the metabolic test to provide sufficient preparation time. The education session included the following: 1) provision of a KD information booklet specifically developed for this study; 2) 3-d menu plan specific to the participant's energy requirements, dietary preferences, and tolerances as determined by prior dietary review; 3) 7-d example menu plan for additional meal ideas; 4) extensive list of snack ideas; and 5) lifestyle, dining-out, shopping, cooking, and budgetary advice. To further support dietary compliance, participants were required to have daily contact with an RD, which included unlimited daily access (phone and e-mail). The prescribed KD contained ≤50 g·d−1 CHO, 15%–20% EI from protein, and 75%–80% EI from fat. Each participant was provided with coconut oil, extra virgin olive oil, LCHF cereal, and (for both dietary conditions) discounted fruit and vegetables. Participants were requested to refrain from alcohol and dietary supplement use for the duration of the study.

To monitor compliance to each dietary intervention, participants were trained in dietary reporting and provided an image-assisted (alongside a fiducial marker) weighed dietary record reported remotely in real-time to an RD via a mobile phone application (WhatsApp, Facebook, San Francisco, CA) for the 5 d preceding each run-to-exhaustion trial, a minimum of two nonconsecutive days between days 1–7, 8–14, and 15–21, and the morning of the trial to confirm the ingestion of the prescribed breakfast. Where underreporting was suspected, participants were required to provide a 24-h dietary recall or repeat the dietary report the subsequent day. Each dietary record was coded (FoodWorks Professional Edition, Version 8; Xyris Software, Queensland, Australia), with images validating the reported intakes, by an RD and checked for accuracy by a second RD. To help maintain energy balance, participants reported their morning body mass daily and were advised to prevent a >2% fluctuation. Verification of compliance to the KD was via daily self-measurement of urinary acetoacetate (AcAc) concentrations with a semiquantitative (color range) strip (Ketostix, Bayer) and capillary blood D-βHB concentrations on days 3, 7, 14, 21, and 28 before ingesting breakfast and exercising. Capillary blood D-βHB concentrations were also measured immediately before the post-KD metabolic test by the primary researcher. Images of the results were immediately sent to the primary researcher for quantification of ketosis. Color comparisons were made using the urinary AcAc strips in a subgroup (n = 4) of participants when ingesting an HD to dismiss the potential of false positives.

Training Monitoring

Participants designed their own 28-d training schedule and were asked to replicate this during each dietary adaptation period. This included a combination of running and cycling. Participants reported their resting morning HR and training data for each session, which included session duration (in minutes), average HR, and RPE. These variables were used to calculate Banister training impulse (TRIMP) and session RPE.[24] Because the intervention was applied under free-living conditions, all other lifestyle choices were allowed to vary naturally.

Blood Sampling and Analysis

Venous blood samples were collected into 8-mL serum vacutainers (Becton Dickinson and Co, Franklin Lakes, NJ) with the participants seated in an upright position. Each serum vacutainer was left to clot for 30 min at room temperature before centrifugation at 1500g for 10 min at 4°C and separation into two 1.5-mL aliquots to be stored at −80°C before the analysis of glucose concentration (Cobas Modular P800 Analyser; Roche Diagnostics, Auckland, New Zealand). Capillary blood D-βHB (Freestyle Optium Neo; Abbott Diabetes Care, Victoria, Australia) and lactate (Lactate 2 Pro; Akray, Kyoto, Japan) concentration was measured from a fingertip blood sample using standardized techniques.

Calculation of Whole Body Rates of Substrate Oxidation and Energy Expenditure

Whole body rates of CHO oxidation, fat oxidation, and EE were calculated from steady-state V̇O2 and V̇CO2 using nonprotein RER values.[25] During the KD trials, no corrections to the equations were implemented, as the oxidation of fat-derived KB does not alter the stoichiometry.[25] To ascertain maximal aerobic EE (EEaero-max), V̇O2max was multiplied by 21.745 J·mL−1 O2, assuming that glucose was the only substrate oxidized.[26] Fat oxidation rates obtained during the metabolic test were depicted graphically as a function of exercise intensity (% V̇O2max), and a third-order polynomial curve with intersection in (0,0) was constructed to determine MFO and Fatmax. A third-order polynomial curve with intersection was chosen based on the best-curve fit and with the assumption that if V̇O2 was zero, no fat oxidation would be observed.

To compare exercise efficiency between the pre- and post-KD conditions, calculations were adapted from previously published sources[25] to predict V̇O2 based on shifts in RER with unchanged efficiency (see the equation hereinafter). The derivation of this equation can be referred to in the Supplemental Content (Supplemental Digital Content 2, Derivation of the equation to calculate predicted oxygen uptake, http://links.lww.com/MSS/B597). Discrepancies between predicted and measured V̇O2 were used to quantify unaccounted oxygen uptake (i.e., oxygen uptake that cannot be explained by shifts in RER alone).

Predicted VO2 = ([0:55 × Pre-KD VCO2] + [4.471 × Pre-KD VO2])/([0.55 × Post-KD RER] + 4.471) Energy conversion: 1 kcal = 4.18 kJ

Data Analysis

All data are expressed as mean (±SD) unless otherwise stated. The mean differences (Δ; ±90% confidence limits (CL)) between interventions and preintervention to postintervention are also expressed. Data were checked for normal distribution using the Shapiro–Wilk tests, and where appropriate, statistical analysis was performed on the logarithmic transformation of the data. For cardiorespiratory data, values were averaged for each hour and only used in analysis if n = 100% for all preintervention and postintervention trials. For urinary AcAc concentrations, weekly values were averaged for each participant. A three-way (diet–adaptation–intensity/time for cardiorespiratory and metabolic variables) or two-way (diet–adaptation for exercise capacity and body composition variables or adaptation–intensity/time for predicted vs measured oxygen uptake) repeated-measures ANOVA was performed, and if Mauchly's test of sphericity was violated, adjustments to the degrees of freedom were made for the ANOVA using Greenhouse–Geisser ε (IBM SPSS Statistics software, version 21; IBM Corp). Where a significant effect was observed, post hoc analysis was conducted using Holm–Bonferroni adjustments for multiple comparisons. Within-group changes from preintervention to postintervention were examined using adjusted Student's paired t-tests for dependent variables. Significance level was accepted at an α of P < 0.05. To interpret the magnitude of effect and to identify trends within nonsignificant data, Cohen d effect sizes (±90% CL) were estimated using a purpose-built spreadsheet,[27] with ES thresholds set at <0.2, >0.2, >0.6, >1.2, >2.0, and >4.0 for trivial, small, moderate, large, very large, and extremely large effects, respectively. If the 90% CL overlapped 0, the magnitude of effect was deemed unclear.