Effects of the ketogenic diet on performance and body composition in athletes and trained adults: a systematic review and Bayesian multivariate multilevel meta-analysis and meta-regression

Abstract

This systematic review with meta-analysis aimed to determine the effects of the ketogenic diet (KD) against carbohydrate (CHO)-rich diets on physical performance and body composition in trained individuals. The MEDLINE, EMBASE, CINAHL, SPORTDiscus, and The Cochrane Library were searched. Randomized and non-randomized controlled trials in athletes/trained adults were included. Meta-analytic models were carried out using Bayesian multilevel models. Eighteen studies were included providing estimates on cyclic exercise modes and strength one-maximum repetition (1-RM) performances and for total, fat, and free-fat masses. There were more favorable effects for CHO-rich than KD on time-trial performance (mode [95% credible interval]; −3.3% [−8.5%, 1.7%]), 1-RM (−5.7% [−14.9%, 2.6%]), and free-fat mass (−0.8 [−3.4, 1.9] kg); effects were more favorable to KD on total (−2.4 [−6.2, 1.8] kg) and fat mass losses (−2.4 [−5.4, 0.2] kg). Likely modifying effects on cyclic performance were the subject’s sex and VO₂max, intervention and performance durations, and mode of exercise. The intervention duration and subjects’ sex were likely to modify effects on total body mass. KD can be a useful strategy for total and fat body losses, but a small negative effect on free-fat mass was observed. KD was not suitable for enhancing strength 1-RM or high-intensity cyclic performances.

Keywords:

• Athlete
• body composition
• ketogenic diet
• endurance
• performance
• strength

Acknowledgment

The authors thank the Post-Graduate Program in Nutrition of the Federal University of Santa Catarina, Brazil.

Disclosure statement

The authors declare no conflict of interest.

Funding

This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001.

Data availability

All data collected from the studies included in the review and the syntaxes of the analysis and figures from software R are available on Open Science Framework (https://osf.io/5nce2/.

Notes

1 For consistency we used the terms “population” and “group” level effects also known as “fixed” and “random” effects, respectively.

2 Wroble et al. (2019) did not evaluate the subjects’ V̇O₂max, so the data was imputed during the model fitting, using procedures detailed elsewhere (Bürkner 2021). Details are provided in Supplementary Material 2.