Reviews/Caffeine and Performance/CYP1A2 genotype moderates the effect
Individual · CAFF-IND-01 · v2
Studies that split athletes by their CYP1A2 gene type disagree on whether the gene changes how much caffeine helps performance. A large cycling study [1] found a strong effect, and two smaller trials [8][11] also saw bigger gains in one gene group, but most careful trials that formally tested for an interaction found none [3][5][6][9][12]. A recurring pattern is that the AA group tends to benefit most and the CC group least, yet in the largest crossover this ordering did not reach statistical significance [2]. The strongest positive study used a design that compares different people rather than the same people on and off caffeine, which can be misled by baseline differences, and the other positives were small. Because formal positive and null results roughly balance, the panel placed the moderation claim in the middle ground rather than confirming or rejecting it.
Across twelve primary trials genotyped for CYP1A2 rs762551, a significant caffeine-by-genotype interaction appears in [1] (P<0.0001, with a 13.7% ergolysis in CC carriers), in [8] (4.9% vs 1.8% improvement, p<0.05) and in [11] (peak power p=.041, only AA improving). Formal interaction tests are null in [3] (p-values 0.094 to 0.994), [5] (no caffeine-gene interaction in time to completion despite higher plasma caffeine in AA), [6], [9] (polymorphism "minimally altered" ergogenicity) and [12] (improvement independent of genotype). The largest crossover [2] found no change to the overall effect, with a supplement-by-genotype interaction only at 90% 1RM, yet still reported a directional gradient (AA 4%-12%, AC 3%-9%, CC ≤4%). Caffeine itself improved performance in nearly every trial; the division is strictly about whether genotype modifies that effect. The contested point: The division is whether a recurring descriptive AA≥AC≥CC magnitude gradient — visible even where the formal interaction is null, as in [2] — counts as moderation or as noise. The positive formal tests [1][8][11] come from either a confounded between-subjects design [1] or small underpowered samples, while the better-powered blinded crossovers [3][5][6][9] return nulls.
All 13 papers behind this claim with their DOIs and what each does to it, the panel’s arbitration, the evidence gap map and the open research questions. Free, and sent to your inbox.
We’ll also send you other claims from Caffeine and Performance, one at a time, and tell you when a release or a pivotal trial changes a grade. Unsubscribe by replying.
One reviewer's only job is to attack the finding. The challenges below were weighed before the grade was set.
The strongest positive, [1], is a between-subjects split-plot design in which the dramatic CC ergolysis has not replicated in any placebo-controlled crossover in the corpus ([3], [5], [6], [9]); baseline differences between genotype groups confound a between-subjects interaction.
Most positive trials are underpowered to detect an interaction term (n=16 in [11], n=18 in [4], n=19 in [9], n=22 in [3]); [11] itself reports observed power of only 0.552, so a significant interaction is at high risk of being a false positive.
Where it remains uncertain. Most adequately blinded crossovers testing the interaction formally found none [3][5][6][9][12], and the strongest positive [1] uses a between-subjects design that has not replicated in any crossover here.
Aleatory
Interaction terms estimated from samples as small as n=16 [11], n=18 [4], n=19 [9] and n=22 [3] are dominated by chance, so both positive and null results are unstable.
Epistemic
The large CC ergolysis in [1] has not been reproduced in any placebo-controlled crossover in the corpus ([3],[5],[6],[9]), so its true size is unknown.
Inferential
A non-significant AA>AC>CC ordering, explicit in [2], is read by some as moderation though no formal interaction reached significance there.
The load-bearing papers from a corpus of 13, each classified by what it does to the claim. Every DOI is taken from the corpus the claim was graded on.
Strongest positive: a significant genotype interaction with AA benefit and CC ergolysis in endurance cycling.
“However, a significant (P <0.0001) caffeine-gene interaction was observed. Among those with the AA genotype, cycling time decreased by 4.8% at 2 mg·kg (17.0 ± 0.3 vs 17.8 ± 0.4 min, P = 0.0005) and by 6.8% at 4 mg·kg (16.6 ± 0.3 vs 17.8 ± 0.4 min, P < 0.0001).”
Panel auditInternal validity: high · scope of inference: Competitive young male athletes performing a 10-km cycling time trial after acute caffeine.Key limitations: male-only · single endurance modality · CC subgroup likely small
Largest crossover: no overall moderation yet a consistent magnitude gradient by genotype.
“Although CYP1A2 genotype did not alter the overall effect, benefits were greatest in AA, intermediate in AC, and minimal in CC carriers.”
Panel auditInternal validity: high · scope of inference: Resistance-trained men and women performing bench press and back squat velocity and endurance tests.Key limitations: one isolated interaction at 90% 1RM only · small CC subgroup (n=11) · directional gradient reported without overall interaction
Null interaction in endurance despite genotype-dependent pharmacokinetics.
“No caffeine-gene interaction effects were observed in TTC, HR or plasma concentrations of paraxanthine and theophylline.”
Panel auditInternal validity: medium · scope of inference: Recreationally active males completing a 10-km run or 40-km cycle time trial; only AA vs AC contrast available.Key limitations: no CC genotype · pooled across two different exercise modalities · male-only
Null interaction across resistance, jump and sprint outcomes.
“We did not find a significant genotype × caffeine interaction effect (p-values ranged from 0.094 to 0.994) in any of the analyzed performance outcomes.”
Panel auditInternal validity: low · scope of inference: Resistance-trained men performing bench press, countermovement jump and Wingate tasks after acute caffeine.Key limitations: very small sample · male-only · underpowered for interaction
4 of 13 papers shown. The full review lists all of them, with the panel’s arbitration →
Claims are graded at each quarterly release, and re-graded between releases only when a pivotal trial lands. Every change is a new version of the claim.
Positive and null formal interaction tests roughly balance, with the strongest positive confounded by design and the positives underpowered, placing the moderation claim in the Uncertain band.
Release 2026-q4 re-graded every claim from its corpus under protocol v0.19. An earlier release's grade is shown for the record; it came from an earlier protocol and is not a movement.
Assumptions that carry weight but are not themselves established.
The panel records the gap that would most move the claim, and the study that would close it.
Adequately powered crossover for the interaction term. Existing crossovers [3][5][6][9][11] are sized for main effects, not the genotype interaction, so nulls and positives alike are underpowered. The panel rates this the highest-importance open gap for the claim, and names the study design that would close it: A placebo-controlled crossover powered a priori on the caffeine-by-genotype interaction, with pre-registered genotype allocation targets.
Open question: Does a caffeine-by-CYP1A2 interaction survive when a crossover is powered specifically for the interaction term? Why it is still open: Every crossover in the corpus ([3][5][6][9][11]) is sized for main effects, and [11] reports observed power of only 0.552.
Priors is a sceptical first reader, not peer review, and not medical advice.
Release 2026-q4 · graded 21 Oct 2026 · protocol v0.19