Umbral
← Back
Energy & metabolismCardio fitness

Caffeine

Anhydrous caffeine capsules for dose precision, or coffee (~95 mg per 240 mL cup); caffeinated gum absorbs faster and needs less lead time

StrongLikely visible in your data
5
sources
58
trials
340
participants (largest study)
3
study designs
The verdict

Ergogenic across every modality tested (aerobic endurance ES 0.22–0.63; peak power ES 0.27, 95% CI 0.08 to 0.47) but costs 45 min of total sleep (95% CI −61.5 to −29.0), 11 min of deep sleep (95% CI −18.5 to −4.3) and 9 min of extra sleep latency (95% CI 3.8 to 14.4) when taken too late.

Why this grade
Emerging
Preliminary or mechanistic
Moderate
Some RCTs, limited replication
Strong
Multiple RCTs or meta-analyses, consistent direction

Both directions are strongly evidenced: an umbrella review of 21 meta-analyses for performance, and a 24-study meta-analysis for the sleep cost. The one honest gap — a 2022 meta-analysis of 5 studies found caffeine did NOT significantly alter post-exercise HRV recovery (RMSSD 95% CI −0.45 to 0.29), so any HRV drift Umbral shows should be attributed to lost sleep, not to a direct autonomic effect.

How it works

Antagonises adenosine A1 and A2A receptors, blocking accumulated sleep pressure and lowering perceived effort while raising sympathetic drive. The same receptor blockade that delivers the performance benefit is what suppresses slow-wave sleep and sustains sympathetic tone into the night — the trade-off is one mechanism, not two.

Who it helps, and who it doesn't

CYP1A2 fast metabolisers get more performance benefit and clear it before bedtime; slow metabolisers pay a disproportionate sleep cost from the same dose. ADORA2A genotype predicts the anxiety and insomnia response. Habitual users retain the ergogenic effect but their sleep cost persists.

Measured effects (absolute, from the trials)

Every estimate, against a line of no effect

Diamonds are point estimates; the hairline is the 95% confidence interval where the trial reported one.

  • Total sleep time
    minutes · Gardiner et al., Sleep Medicine Reviews 2023; noise_sd = estimated within-person night-to-night SD of wearable TST
    −45.3
    −61.5 … −29
  • Deep sleep (N3)
    minutes · Gardiner et al., Sleep Medicine Reviews 2023; noise_sd = estimated within-person night-to-night SD of wearable deep sleep
    −11.4
    −18.5 … −4.30
  • Sleep onset latency
    minutes · Gardiner et al., Sleep Medicine Reviews 2023; noise_sd = estimated within-person night-to-night SD of wearable sleep latency
    +9.10
    +3.80 … +14.4
−50−250NO EFFECT
Separate scale · SMD (null)
  • Post-exercise HRV recovery (RMSSD)
    SMD (null) · Porto et al., Nutr Metab Cardiovasc Dis 2022 — no significant effect; noise_sd = estimated within-person night-to-night SD of wearable RMSSD
    0.00
    −0.45 … +0.29
−0.5−0.30+0.3NO EFFECT

Units are never mixed on one line: each scale gets its own axis. Where a trial reported no confidence interval we leave it blank rather than inventing one.

Will you see it in your wearable?

We compare each effect to the typical night-to-night noise of that metric. Effects smaller than the noise are real but invisible in a week of data.

  • Total sleep timeLikely visible
    effect: 45.3 minutesnightly noise: ±45 minutes
  • Deep sleep (N3)Visible over weeks
    effect: 11.4 minutesnightly noise: ±18 minutes
  • Sleep onset latencyLikely visible
    effect: 9.1 minutesnightly noise: ±9 minutes
  • Post-exercise HRV recovery (RMSSD)Below noise
    effect: 0 SMD (null)nightly noise: ±9 SMD (null)
Studies over time

5 studies across 6 years

2017 — 2023
20172019202020222023
Meta-analysis or systematic reviewRandomized or crossover trialObservational or preliminaryHover a marker for detail
Dose
3–6 mg/kg body mass; as little as 2 mg/kg is ergogenic for some, and 9 mg/kg adds side effects without adding performance
Timing
60 min before exercise (less for gum). The cut-off matters more than the dose: a 107 mg coffee needs ≥8.8 h before bedtime and a 217 mg pre-workout ≥13.2 h — meaning a pre-workout at 6 pm is still costing sleep at midnight.
Safety flags
  • · Sleep disruption is the dominant documented cost: −45.3 min total sleep time, −7.0% sleep efficiency, −11.4 min deep sleep
  • · Anxiety, jitteriness, tachycardia and palpitations, strongly dose-dependent and modulated by ADORA2A genotype
  • · Acutely raises blood pressure and delays post-exercise heart rate and blood pressure recovery
Where to buyRecord what I'm taking

References

  1. 1.
    Wake up and smell the coffee: caffeine supplementation and exercise performance—an umbrella review of 21 published meta-analyses
    British Journal of Sports Medicine · 2020 · systematic review · 21 trials

    Aerobic endurance ES 0.22–0.63; muscle endurance 0.28–0.38; strength 0.16–0.20; peak power 0.27 (0.08–0.47); mean power 0.18 (0.05–0.31); vertical jump 0.17 (0.00–0.34). Most common dose 6 mg/kg, 60 min pre-exercise.

  2. 2.
    The effect of caffeine on subsequent sleep: A systematic review and meta-analysis
    Sleep Medicine Reviews · 2023 · meta-analysis · 24 trials · n = 340

    Total sleep time −45.3 min (−61.5 to −29.0), sleep efficiency −7.0% (−10.0 to −4.0), sleep onset latency +9.1 min (3.8 to 14.4), N3 deep sleep −11.4 min (−18.5 to −4.3). Cut-off: coffee (107 mg) ≥8.8 h and pre-workout (217.5 mg) ≥13.2 h before bed.

  3. 3.
    International society of sports nutrition position stand: caffeine and exercise performance
    Journal of the International Society of Sports Nutrition · 2021 · systematic review

    3–6 mg/kg improves performance consistently, 60 min pre-exercise most common; 9 mg/kg adds side effects without benefit. Inter-individual differences in performance and in sleep/anxiety adverse effects attributed to CYP1A2 and adenosine receptor genotype.

  4. 4.
    Caffeine intake and its influences on heart rate variability recovery in healthy active adults after exercise: A systematic review and meta-analysis
    Nutrition, Metabolism and Cardiovascular Diseases · 2022 · meta-analysis · 12 trials

    Pooled across 5 studies, caffeine (100–400 mg or 2–6 mg/kg) did not significantly affect post-exercise HRV recovery: RMSSD 95% CI −0.45 to 0.29 (p=0.67), HF 95% CI −0.34 to 0.30 (p=0.22).

  5. 5.
    Caffeine affects autonomic control of heart rate and blood pressure recovery after aerobic exercise in young adults: a crossover study
    Scientific Reports · 2017 · crossover · 1 trials · n = 32

    300 mg caffeine delayed parasympathetic reactivation — RMSSD and SD1 recovered at 15–20 min vs 5–10 min with placebo; systolic BP recovery delayed from 5 to 7 min.

Umbral is a wellness product, not a medical device. It does not diagnose, treat, cure or prevent any disease. Supplement information is educational; talk to a clinician before starting anything, especially if you are pregnant, under 18, managing a condition or taking medication.