REF-1537 · Diaz-Lara et al., 2025 · Sports Medicine · Paper · PubMed

Does low carbohydrate lift PGC-1α?

From 19 randomised studies: metabolic genes yes, PGC-1α and TFAM no clear overall effect, PGC-1α up in a subgroup with a glycogen gap above 200 mmol/kg dw. mRNA only, no performance.

Main caveat: Read from the abstract. Included studies, participants, per-study sizes, sex and risk of bias are not printed.
fairly strong evidenceMeasured on a mixed group

The review searched three databases to November 2023 and pooled 19 randomised controlled studies comparing low-carbohydrate conditions with high-carbohydrate ones. Low carbohydrate went with more mRNA for three metabolic genes in early recovery. It showed no clear overall effect on PGC-1α or TFAM, the two genes tied here to mitochondrial biogenesis.

Glycogen was lower in the low-carbohydrate condition (SMD 3.69). When studies were grouped by a glycogen difference above 200 mmol/kg dw between conditions, PGC-1α did rise and the effect on PDK4 and GLUT4 was greater.

Our reading: the common signal from low carbohydrate is in metabolic genes, and PGC-1α appears to need a large glycogen difference; this is a pooled number where the base had only a tally of studies (REF-543).

Protocol

What it isolates: the fuel: low against high carbohydrate availability around exercise. It does not isolate a second session; low availability was produced before, during or after exercise.
Search: MEDLINE, Scopus and Web of Science, PRISMA 2020, end date November 2023.
Included: 19 randomised controlled studies. Outcomes: PGC-1α, TFAM, PDK4, UCP3 and GLUT4 mRNA, and muscle glycogen.
Quality: Cochrane Risk of Bias tool; its findings are not in the abstract.
Analysis: random-effects model, standardised mean difference by Hedges' g with 95% CI; a subgroup analysis by glycogen difference between conditions.
Funding and interests: the authors state that no funding was received and declare no conflicts of interest.

Key Results

Measure, low against high carbohydrateSMD95% CIOther
PDK4 mRNA, early recovery1.610.80-2.42
GLUT4 mRNA, early recovery1.380.46-2.30
UCP3 mRNA, early recovery2.050.40-3.69
PGC-1α mRNA, overallnot printedno significant effect
TFAM mRNA, overallnot printedno significant effect
Muscle glycogen3.692.82-5.09lowered by restriction with exercise
PGC-1α mRNA, subgroup with a glycogen gap above 200 mmol/kg dw2.080.64-3.52p = 0.005, I² = 75%
No clear overall effect, not "no effect". The overall PGC-1α and TFAM estimates are not printed in the abstract, so their width is unknown here.
SMDs are not percentages. They cannot be read as a fold change in gene expression (inference).
A printed oddity: the glycogen interval (2.82-5.09) is not centred on its estimate (3.69), unlike the others. Given as printed.

Mechanisms

Measured: mRNA content after exercise, and glycogen.
Authors' interpretation: restriction raises the metabolic genes but not PGC-1α or TFAM overall; a substantial glycogen difference adds to the metabolic genes and brings in PGC-1α.
Inference: a threshold for the glycogen gap comes from grouping studies, not from a trial that randomised the gap. A trial that graded glycogen below 300 mmol/kg dw, across a range of 190 mmol/kg dw (our arithmetic), just under this gap, reported no added signal (REF-1551).
Not pooled, per the abstract: protein, enzymes, mitochondrial content, training adaptation and performance. mRNA is an early signal, and this review does not show that it becomes adaptation.

Practical Application

Does low carbohydrate lift PGC-1α? From 19 randomised studies: metabolic genes yes, PGC-1α and TFAM no clear overall effect, PGC-1α up in a subgroup with a glycogen gap above 200 mmol/kg dw. mRNA only, no performance.
Good for: correcting "train-low boosts the mitochondrial signal" as a general statement. The pooled result is narrower than the claim.
Good for: context on doubles sold as a signal trick: if the second session's extra signal is meant to come from low fuel, a small glycogen difference may not deliver it (inference; doubles were not pooled separately in the abstract).
Not good for: any performance decision. For performance in trained athletes see REF-218.
Not good for: a glycogen target in training. Athletes do not know their glycogen (inference), the subgroup is heterogeneous, and the abstract gives no practice line. Not from this paper: the base's practice on fasted sessions is in SYNTH-83.

Limitations

Read from the abstract. Included studies, participants, per-study sizes, sex and risk of bias are not printed.
Serious heterogeneity in the one subgroup that carries the PGC-1α claim (I² = 75%), with a wide interval.
Subgroup, not a randomised contrast. The glycogen gap was not assigned across studies.
Timing bundled: the abstract groups restriction before, during and after exercise.
Markers only. Nothing on performance.
Overlap with this literature: one author (Bishop) is also an author of REF-1535, REF-1536 and REF-1549; whether those are among the 19 is not stated in the abstract.
How this entry was checked and changed
Why this level

(from the abstract), for the mRNA question only — a meta-analysis of controlled trials with serious heterogeneity in its key subgroup (I² = 75%). It carries no performance outcome, so this level does not extend to performance. Per-study sizes and the risk-of-bias findings are not in the abstract. Subscription article; no copy in PubMed Central or Europe PMC