Reviews/Caffeine and Performance/Ergogenic effect is dose-dependent
Dosing · CAFF-DOS-01 · v2
Across ten independent randomised trials, how much caffeine a person takes clearly changes whether and how much their performance improves. Very low doses such as 1 mg/kg often do nothing [6], a moderate dose around 6 mg/kg tends to work, and pushing higher frequently adds nothing or even makes performance worse [5][8][11][12]. So the amount of caffeine genuinely matters, which is what the claim says in its loosest sense. But the relationship is not a simple 'more is better' line: it levels off and sometimes reverses, the best dose depends on the task and load [2], and one trial built specifically to test this found no difference at any dose [4]. The panel judged the claim well supported in broad outline but tempered by its vague wording, the mixed curve shapes, small samples and one null.
The corpus repeatedly shows that caffeine dose modulates the ergogenic response rather than acting all-or-nothing. Sub-threshold doses commonly fail (1 mg/kg in [6]; 3 mg/kg against high loads in [2] and for upper-body throws in [9]) while ~6 mg/kg is effective ([1][8]). The response is frequently non-monotonic: it plateaus with 6 mg/kg equal to 9 mg/kg ([5][8]) and follows an inverted-U with performance beginning to deteriorate at 7.5 mg/kg in fencers [11], while in one cognitive study the low 3 mg/kg dose outperformed moderate and high doses [12]. The required dose is load-dependent, with 3 mg/kg sufficient for low loads but 9 mg/kg needed for the heaviest load [2]. The contested point: The division is whether the claim's 'dose-response' is genuinely supported given that the most on-point designs do not show a clean graded curve: [4], built to test dose-response in rowing, found no difference across any dose or placebo (P = 0.249), and [3] found the dose-dependent effect statistically nonsignificant. The rest of the corpus supports dose-dependency but in plateaued or inverted-U form rather than the monotonic gradient the wording suggests.
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 most on-point designs partly contradict the claim: [4], a trial built to test dose-response in rowing, found no difference across any dose or placebo (P=0.249), and [3] found the dose-dependent effect statistically nonsignificant.
'Dose-response relationship' implies a graded curve, but the corpus repeatedly shows a plateau (6 = 9 mg/kg in [5][8]) and an inverted-U with decline at high dose ([11][12], low > high in [12]) — real dose-dependency, but not the monotonic gradient the wording suggests.
Where it remains uncertain. The trial explicitly designed to test dose-response found no difference at any dose (P = 0.249, [4]) and another found the dose effect nonsignificant [3].
Aleatory
[4] reports considerable interindividual variation in plasma caffeine concentrations, so a given dose produces different exposures and responses between people.
Epistemic
Plateau ([5][8]), inverted-U with deterioration at 7.5 mg/kg [11] and an inverse low-dose advantage [12] coexist; the true functional form across doses is not pinned down.
Inferential
Combining physical performance, reaction time/cognition and physiological surrogates (HR/BP in [6]) into one 'ergogenic' dose-response risks inferring a single curve from distinct response systems.
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.
Dose requirement varies with load — direct evidence that effect depends on dose.
“The ergogenic dose of caffeine required to enhance neuromuscular performance during a single all-out contraction depends on the magnitude of load used.”
Panel auditInternal validity: medium · scope of inference: Acute muscle strength and power during single all-out contractions in resistance-trained men.Key limitations: small sample · trained men only · side effects high at 9 mg/kg
Counter-evidence: a direct dose-response trial returning no effect at any dose.
“Time trial performance was not significantly different across the three caffeine doses or placebo (P = 0.249).”
Panel auditInternal validity: medium · scope of inference: 2000-m rowing ergometer time-trial performance in competitive male rowers.Key limitations: very small sample · large interindividual plasma variation · preexercise feeding may have blunted response
Dose-dependent improvement with optimum at moderate dose.
“Total sprint number was greater in MOD and HIGH compared to PLA (20 ± 7 and 18 ± 8 vs. 13 ± 4; all P < 0.05), with MOD also higher than LOW (15 ± 6; P = 0.02).”
Panel auditInternal validity: medium · scope of inference: Repeated cycling sprint ability to exhaustion in active males under normobaric hypoxia.Key limitations: small sample · plateau at highest dose (no 9>6) · hypoxic-specific protocol
Threshold dose effect on reaction time (6 but not 3 mg/kg).
“Reaction time only improved following 6 mg/kg of caffeine intake”
Panel auditInternal validity: medium · scope of inference: Acute cognitive performance (reaction time, attention, memory) in healthy physically active young adults at 3 and 6 mg/kg.Key limitations: cognitive not physical performance · only two active doses · most outcomes unaffected at either dose
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.
Dose clearly shapes the effect across many independent RCTs, but non-monotonic curves, one null and one nonsignificant direct trial, small samples and the claim's loose wording preclude the top tier.
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.
Functional form of the dose-response curve. Plateau, inverted-U and inverse-dose patterns coexist ([5][8][11][12]); the actual shape is not established. The panel rates this the highest-importance open gap for the claim, and names the study design that would close it: A large multi-dose within-subject trial densely sampling 0 to ~9 mg/kg in a single task, powered to estimate curvature and inflection.
Open question: What is the true shape of caffeine's dose-response curve within a single exercise task across a dense dose range? Why it is still open: The corpus shows plateau ([5][8]), inverted-U ([11]) and inverse ([12]) patterns in different tasks, with no single design resolving the form.
Priors is a sceptical first reader, not peer review, and not medical advice.
Release 2026-q4 · graded 21 Oct 2026 · protocol v0.19