Keto for Endurance: Better Fat Burning, Tougher Race Pace

Low-carbohydrate and ketogenic diets can teach trained endurance athletes to burn much more fat during exercise. They have not shown that this skill makes athletes faster when racing hard. That distinction matters: a marathon, triathlon, or cycling race is not a laboratory VO2max test, and a 1–3% loss in pace can decide a result.

The most useful reading of the evidence is neither "carbs are obsolete" nor "fat adaptation is fake." Fat adaptation is real. Chronic full-time keto remains a poor bet for athletes whose events demand sustained work near or above lactate threshold.

What the newest review actually found

A 2026 systematic review and meta-analysis by Gawelczyk et al. examined 33 studies in trained endurance athletes. Its most consistent finding was metabolic: all 30 studies that measured fat oxidation reported increases, ranging from 28% to 200%.

Maximal aerobic outcomes looked less uniform. VO2max was preserved in 50.0% of studies and improved in 11.1%; time to exhaustion was maintained in 69.2%. Studies assessing athletes within seven days of starting a low-carbohydrate or ketogenic diet uniformly reported impaired performance, while studies lasting longer than seven days more often reported maintained or improved maximal aerobic outcomes.

That pattern supports a familiar physiological point: the first days of carbohydrate restriction are rough. Athletes need time to alter fuel use. It does not prove that a week of adaptation restores race performance. The review also reported impaired submaximal exercise economy in 50.0% of studies after the acute phase. Economy is the oxygen required to hold a given pace or power. For an athlete racing near threshold, that can matter more than a preserved VO2max.

There is an important reporting limitation. The available publication summary does not provide forest-plot effect sizes, confidence intervals, heterogeneity statistics, publication-bias testing, or sport-specific subgroup analyses. The reported percentages describe how individual studies classified results; they do not establish that low-carb and high-carb diets are clinically equivalent for competitive performance.

More fat oxidation does not erase the oxygen bill

The strongest challenge to a performance-neutral interpretation comes from a controlled intervention in trained cyclists. In Burke et al.'s 2021 study, adaptation to a low-carbohydrate, high-fat diet rapidly increased fat oxidation, yet endurance performance remained worse than under high-carbohydrate conditions. The impairment persisted after researchers restored glycogen availability.

That result points away from a simple "the athletes needed more glycogen" explanation. Fat oxidation produces less ATP per unit of oxygen than glucose oxidation: the reported P/O ratios are about 2.3 for fat and 3.17 for glucose. At a fixed oxygen uptake, relying more heavily on fat can therefore reduce the power or pace an athlete can sustain. Estimates in this literature put the oxygen cost of fat oxidation roughly 5–15% above that of carbohydrate oxidation per unit of ATP produced.

This is not an argument that fat is a bad fuel. At lower intensities, especially during long events, abundant fat stores are useful. But as intensity rises, oxygen becomes a tighter constraint. A metabolic adaptation can be impressive and still be poorly matched to a hard race.

The reviews agree on metabolism, not on winning

An earlier 2021 meta-analysis by Cao et al. reached a more cautious conclusion. Ketogenic low-carbohydrate, high-fat diets produced mixed effects on aerobic capacity and did not demonstrate a clear advantage over high-carbohydrate conditions for time-trial performance or VO2peak.

The two reviews agree on the finding that has held up best: trained athletes burn more fat after carbohydrate restriction. They differ in how reassuring the performance data appear. Differences in diet duration, athlete populations, restriction depth, outcome tests, and small trial sizes make a simple average misleading. A statistically non-significant difference in a small sports-nutrition trial can mean genuine equivalence, or it can miss the small decrement that matters in competition.

VO2max also answers a narrower question than many athletes assume. It measures maximal aerobic capacity; it does not directly measure the oxygen cost of holding race pace, surging on a climb, or closing the final kilometers. "Maintained VO2max" should not become "maintained race pace" by editorial magic.

Periodization makes more sense than permanent keto

The practical compromise is to use low-carbohydrate exposure as a training tool rather than a permanent race diet. The 2026 review endorses a train-low, compete-high strategy, and a 2025 controlled intervention by Soegaard et al. supplies a useful proof of concept. Restricting carbohydrate during recovery from high-intensity interval training increased fat oxidation during a subsequent exercise session without compromising performance when caffeine was co-administered.

The caffeine condition is part of the result, not a footnote. The study did not show that carbohydrate restriction alone preserved performance in every training context. It showed that this specific recovery strategy, paired with caffeine, could increase fat oxidation without reducing the measured performance outcome.

Nor does the evidence yet provide a competition-week recipe. It does not establish the ideal carbohydrate dose, the number of days needed for reintroduction, or whether an athlete can fully preserve a fat-adaptation advantage after carbohydrate loading. In a 2025 controlled study in 40 moderately to well-trained men, only three days of low-carbohydrate, high-fat or high-carbohydrate, low-fat eating shifted peak fat oxidation and Fatmax in relation to muscle glycogen availability. Fuel-use adaptations can move quickly in both directions.

Do not confuse keto with underfueling

Energy availability is a major confounder. In a randomized crossover trial of trained competitive cyclists and triathletes, Graybeal et al. found that ketogenic-diet conditions produced markers consistent with low energy availability. Those markers do not prove that keto caused a performance problem, but they make interpretation harder. Restricting carbohydrate can also lead athletes to eat too little overall.

That risk has direct experimental support. A 2024 randomized, blinded crossover trial in 12 endurance-trained females compared 14 days of low versus optimal energy availability while participants maintained about eight hours of weekly training. Low energy availability impaired substrate utilization, reduced insulin sensitivity, and compromised skeletal-muscle mitochondrial oxidative capacity. Those effects belong to inadequate energy intake, not necessarily to carbohydrate restriction itself.

Athletes trying low-carbohydrate training should therefore monitor total energy intake, recovery, and menstrual or hormonal health where relevant. Macronutrient ratios do not protect someone from underfueling.

What about long-term keto?

Human evidence beyond roughly 12 weeks remains thin. A 2025 review by Wang et al. identified animal-model findings of reduced mitochondrial biogenesis and myocardial pathology with long-term ketogenic diets. Animal findings cannot establish harm in human athletes, but they justify caution rather than year-round confidence.

For competitive athletes, low-carb training may build fat-oxidation capacity. Current evidence does not show that chronic keto improves race-day performance, and controlled cycling data point to an economy penalty when intensity rises.

Ultra-endurance athletes racing for more than three hours at lower, sub-threshold intensities may have the strongest theoretical case for fat adaptation. The available review does not provide the subgroup data needed to confirm that advantage. For now, the defensible approach is modest: use carbohydrate availability strategically in training, restore carbohydrate for demanding competition, and treat any precise timing protocol as an experiment rather than established science.