Your caffeine boost may hinge on one gene

Caffeine is among the best-established performance aids in sport, yet almost every coach and athlete knows someone for whom it does nothing, or worse. One proposed explanation is CYP1A2, the liver enzyme gene that governs how quickly caffeine is metabolised, with fast and slow variants that could plausibly change how much drug reaches the point of action and for how long. Commercial genetic tests already sell recommendations on this basis. This review set out to gather the randomised evidence and ask whether genotype actually predicts who benefits.

The authors followed PRISMA reporting standards and searched eight databases up to May 2025, including PubMed, EMBASE, Scopus, Web of Science, the Cochrane Central Register of Controlled Trials, CINAHL, LILACS and SPORTDiscus. Only randomised clinical trials that examined both caffeine supplementation and the influence of CYP1A2 genotype on exercise performance were eligible. Nineteen studies published between 2012 and 2024 met the criteria, involving 732 participants who were athletes or physically active individuals. Caffeine dosages across the included trials ranged from 2 to 9 mg/kg. Risk of bias was assessed with the updated Cochrane RoB 2 tool.

There was no consensus across all the included studies, but a pattern emerged by genotype. Caffeine supplementation showed positive results for sports performance in individuals with the AA genotype across various sporting modalities and physical tests. Those with the AC genotype also showed positive results, but with smaller effect sizes, described by the authors as a marginal improvement. For individuals with the CC genotype, the effects were null or negative. The quality picture was mixed: 84% of the included studies were classified as raising 'some concerns' in the risk of bias analysis, with only two studies rated as low risk.

This organises a scattered and often contradictory literature into a coherent gradient — AA best, AC marginal, CC null or worse — without pooling the data statistically, so it describes a direction rather than quantifying an effect. It complicates the standard advice that caffeine works for almost everyone at the right dose, while stopping well short of validating consumer genetic tests, since the trials disagreed among themselves and nearly all carried bias concerns. Dose is another unresolved variable: with intakes spanning 2 to 9 mg/kg, some apparent genotype effects could reflect dosing rather than metabolism. Well-powered trials that randomise by genotype with pre-registered performance outcomes would settle it.

The defensible reading is that individual response to caffeine before training or competition varies for reasons that may be partly genetic, and that a poor or negative response is a real documented pattern rather than imagination. It is not yet a reason to buy a genetic test, given that most of the underlying trials had methodological concerns and did not agree. Personal experience across several sessions remains the most direct evidence anyone has about their own response. The doses studied covered a wide range, which is itself a reminder that how much is taken sits alongside genotype as an explanation for who benefits.