Carbohydrate, not just calories, drove the hormone drop of underfuelling in 16 adults
A tightly controlled crossover trial separated low energy availability from low carbohydrate availability for four days. One hormone moved clearly; most did not, and performance did not move at all.
Low energy availability is the exposure at the centre of relative energy deficiency in sport, and almost every study of it has a design problem: athletes who eat too little for their training load are usually also eating too little carbohydrate. The two exposures travel together, so their effects cannot be told apart.
A randomised crossover trial published in the Journal of Applied Physiology pulls them apart under laboratory control, and finds that carbohydrate accounts for part — but only part — of the endocrine response [s1].
The design
Sixteen recreationally active adults, nine of them female, completed two four-day conditions of low energy availability set at 25 kcal per kg of fat-free mass per day, with low and high carbohydrate availability of 1.5 versus 4.0 g per kg per day [s1]. Diets were fully controlled, and participants performed daily exercise expending 20 kcal per kg of fat-free mass per day [s1].
The study defines carbohydrate availability as the difference between carbohydrate intake and oxidation during exercise [s1] — a definition that matters, because it makes the exposure a balance rather than an intake target.
Holding energy availability constant while varying carbohydrate is the whole point of the design. Both arms were underfuelled to the same degree; only the macronutrient composition differed.
What moved
Independent of condition, low energy availability produced reductions in body mass, with a mean relative change of −1.3% (P<0.001) [s1]. That confirms the exposure worked.
The clearest condition effect was on insulin-like growth factor 1. High carbohydrate availability significantly attenuated the low-energy-availability-induced decline in IGF-1: −22.9 ± 29.6 ng/mL against −48.5 ± 26.2 ng/mL in the low-carbohydrate arm (P=0.018) [s1]. Roughly half the fall in IGF-1 was attributable to carbohydrate rather than to energy deficit as such.
Insulin, leptin and testosterone also showed greater decreases under low versus high carbohydrate availability, but the condition × time interactions did not achieve significance [s1]. This is the part most likely to be over-read. Three hormones moving in the same direction as IGF-1 is suggestive, but the trial did not detect those differences, and reporting them as effects would misstate what a study of sixteen people can resolve.
Metabolically, low energy availability induced a shift toward greater fat oxidation at rest and during exercise, which was not modulated by carbohydrate availability but by sex [s1]. High carbohydrate availability did mitigate reductions in submaximal and maximal blood lactate concentrations, which the authors read as partial preservation of glycolytic function [s1].
Performance did not move. Maximal performance indices — peak oxygen uptake and maximal power output — were neither affected by low energy availability nor by condition [s1].
What sixteen people over four days can support
The sample is small and mixed-sex, with nine women among sixteen participants [s1], which leaves the sex-specific substrate finding underpowered as a standalone result even though the trial reports it. The participants were recreationally active rather than competitive athletes, so the training loads and baseline energy flux differ from the population the concept was built around.
The duration is the sharper limit. Four days of low energy availability is enough to move hormones and body mass [s1]; the clinical syndrome it models develops over months to years. That performance was unaffected [s1] is therefore not reassurance about chronic underfuelling — it is what you would expect over four days, and the trial cannot speak to longer exposures.
Crossover designs also carry order and carryover risks. The abstract does not report washout details, and with a four-day intervention repeated in the same individuals, that is a question a reader should hold open.
Where it sits in the literature
A systematic review published online in the British Journal of Sports Medicine in October 2025 set out what is known about how low energy availability affects adaptation to training, and its central result is disagreement [s2].
The review searched six databases in November 2024, screening 3,388 articles after removing duplicates from an initial 6,399, and included 21 studies with 536 total participants — 44% male, 56% female — at a mean PEDro methodological quality score of 5 ± 2 [s2]. Inconsistencies in low energy availability thresholds and criteria, exercise exposures and follow-up timeframes precluded pooled analyses [s2].
Every included study examined changes in lean muscle or fat-free mass. Ten identified impaired lean mass or fat-free mass responses in those with low energy availability, ranging from −1% to −5%; ten found no substantial changes, under 1% [s2]. Six of the seven studies that examined functional or sport-specific performance identified impairments over time, ranging from −4% to −10% [s2]. All remaining primary adaptations examined — strength, limb circumference, cellular-level and subjective measures — were impaired with low energy availability [s2].
An even split on the most-measured outcome, in a literature that cannot be pooled because its definitions do not agree [s2], is the context the crossover trial enters. Its contribution is not a large effect; it is a controlled separation of two exposures that the rest of the field has been unable to distinguish.
What to watch
The obvious next study is the same manipulation run for weeks rather than days, in trained athletes, with bone turnover and menstrual function measured alongside the endocrine panel. Until that exists, the defensible claim from this trial is narrow: over four days of matched energy restriction, carbohydrate availability accounted for roughly half the IGF-1 decline and some preservation of glycolytic capacity, and for nothing measurable in maximal performance [s1].
Sources
- Increased carbohydrate availability partially attenuates endocrine suppression in response to low-energy availability, Journal of Applied Physiology, published online 16 July 2026
- Impaired neuromusculoskeletal response to training stimuli associated with low energy availability: a systematic review, British Journal of Sports Medicine, published online 13 October 2025 (January 2026 issue)
Sources
- Increased carbohydrate availability partially attenuates endocrine suppression in response to low-energy availability — Journal of Applied Physiology , July 16, 2026
- Impaired neuromusculoskeletal response to training stimuli associated with low energy availability: a systematic review — British Journal of Sports Medicine , October 13, 2025
More on
In 43% of energy-availability studies, menstrual status is reported too vaguely to grade
The same audit found high-performance female athletes well represented in the field — and that the studies generating prevalence figures lean on self-reported food records and MET lookup tables.
A DXA body-fat Z-score flagged twice as many athletes with absent periods as BMI did
In 388 female athletes, a per cent fat Z-score below −1.0 more than doubled sensitivity for amenorrhoea against standard BMI thresholds. It gave up specificity to do it.
A review of low energy availability in team sport found prevalence between 0% and 80%
Fourteen studies covering 265 elite athletes used six different methods to identify relative energy deficiency. The spread in the answers is mostly a spread in the definitions.
A worrying ECG pattern was rare in screened high-schoolers — but not meaningless
Across 16,893 high school athletes screened over 15 years, 0.36% had T-wave inversions. Four were later found to have hypertrophic cardiomyopathy, a leading cause of sudden cardiac death in the young.