In a 2018 trial run out of the University of Toronto, 101 trained male athletes drank 0, 2, or 4 mg/kg of caffeine 60 minutes before a 10 km cycling time trial.1 The headline was not the average effect. It was the split. Athletes who carried two copies of the AA variant in the CYP1A2 gene rode 4.8 percent faster on 2 mg/kg and 6.8 percent faster on 4 mg/kg. Athletes with the AC heterozygous genotype showed no meaningful change at either dose. Athletes with the CC homozygous variant rode 13.7 percent slower on 4 mg/kg than on placebo. Same dose. Same sport. Three different outcomes.
That study is the cleanest available illustration of a problem most caffeine writing avoids. The supplement is real, the average performance effect is reproducible, and the variance is wide enough that a single number cannot describe what the molecule does to the person taking it. Two genes, one liver enzyme and one adenosine receptor, do most of the work explaining the spread.
This piece is the practical layer on top of the caffeine glossary entry. The glossary covers dose ranges and the safety ceiling. What follows is the decision framework underneath them: how to read your own response, what to dose for each kind of session, when to lean on coffee versus gum, what habituation actually costs, and where the sleep tradeoff turns the supplement into a net negative.
01CYP1A2 and ADORA2A genotypes drive response variance
CYP1A2 is the hepatic enzyme that handles about 95 percent of caffeine clearance. The rs762551 polymorphism splits adults into rapid metabolizers (AA), intermediate (AC), and slow (CC). The AA genotype is present in roughly 40 to 45 percent of the population in most studied cohorts, AC in about 40 percent, and CC in 15 to 20 percent. Womack and colleagues ran a 40 km cycling time trial in 2012 and found that AA cyclists improved by 4.9 percent on 6 mg/kg while C-allele carriers improved by 1.8 percent.2 Pataky and colleagues used 6 mg/kg caffeine in a 3 km cycling trial and found that genotype, time of day, and ingestion route changed the performance response.3
The mechanism shows up through exposure. Caffeine clears slowest in C-allele carriers, which lengthens systemic exposure. That can look useful on paper, but the cardiovascular side effects scale with serum caffeine. Slow metabolizers run higher peak concentrations, sit in the elevated-heart-rate zone longer, and report jitters, reflux, and sleep disruption at doses that AA carriers tolerate without comment. The Cornelis line of work tied slow metabolizer status to higher myocardial infarction risk at higher habitual coffee intakes, with the clearest elevated risk at 4 or more cups per day.4
ADORA2A is the second axis. The rs5751876 polymorphism affects the adenosine A2A receptor that caffeine binds. Human caffeine-anxiety studies have linked the TT variant with higher caffeine-induced anxiety in some cohorts.5 The receptor axis is less well characterized than the enzyme axis, but the practical implication is the same. The same milligram per kilogram does different things in different bodies.
You do not need a genetic test to act on this. The phenotype shows up within two weeks of deliberate dosing. The 14-day protocol at the end of this piece is how to read it.
02Modality-specific ergogenic gains from targeted dosing
The International Society of Sports Nutrition 2021 position stand identified 3 to 6 mg/kg as the reliable ergogenic range for trained athletes.6 That range averages a 2 to 7 percent performance improvement across endurance, sprint, team sport, and resistance training. Meta-analyses by Grgic and colleagues report small positive effects for strength and power outcomes, with larger effects when the dose lands in the 3 to 6 mg/kg window and is taken 60 minutes pre-session.7
The benefit varies across sport. The dose-by-modality matrix below is built from the Wickham and Spriet 2018 review and the modality-specific meta-analyses that have followed it.8
| Session | Working dose | Window before start | Expected performance lift | Caveats |
|---|---|---|---|---|
| Endurance race 30 to 180 min | 3 to 6 mg/kg, capped at 400 mg/day | 45 to 60 min | 2 to 4 percent on time to complete | Plan for higher heart rate at given pace. Pair with race-week fueling and sodium plan. |
| Ultra-endurance over 3 hours | 1.5 to 3 mg/kg, then 1 mg/kg once or twice later | First dose 30 min in | 3 to 7 percent on late-race output | Keep total caffeine at or below 400 mg/day. Watch GI tolerance. |
| Sprint or team sport repeated sprints | 3 to 6 mg/kg, capped at 400 mg/day | 45 to 60 min | 2 to 4 percent on repeated-sprint output | Larger effect in fatigued states than in fresh sprints. |
| Strength training, top sets above 80 percent 1RM | 3 to 6 mg/kg, capped at 400 mg/day | 45 to 60 min | 1 to 3 percent on peak force, 3 to 7 percent on reps to failure | Effect is larger on volume work than on absolute 1RM tests. |
| Hypertrophy session, 8 to 20 rep sets | 2 to 4 mg/kg, capped at 400 mg/day | 45 to 60 min | 5 to 10 percent on volume completed | Lower-dose works because session is endurance-flavored at the muscle level. Pair with leucine-anchored meals. |
| Short HIIT 6 to 30 min | 3 to 6 mg/kg, capped at 400 mg/day | 45 to 60 min | 2 to 5 percent on total work | Effect plateaus above 6 mg/kg. |
| Combat sport rounds | 3 mg/kg | 45 to 60 min | Improved late-round output | Higher doses raise anxiety, heart-rate, GI, and sleep costs. |
| Cognitive work block | 1 to 2 mg/kg | 15 to 30 min | Lower perceived effort and steadier vigilance | The dose that helps a desk session is often too small for a hard workout. |
The table has no row recommending 8 or 10 mg/kg. The dose-response above 6 mg/kg flattens hard, side effects scale linearly, and the ISSN position stand reports that very high doses such as 9 mg/kg are associated with a higher incidence of side effects and are not required for ergogenic benefit.6 When 4 mg/kg falls short, more caffeine rarely produces more performance. The useful next lever is usually a different one entirely. More carbohydrate, better pacing, more sleep, or a different supplement like sodium bicarbonate for anaerobic events.
03Caffeine formulations, onset kinetics, and in-event redosing limits
Caffeine in capsule form peaks at about 45 to 75 minutes. Black coffee peaks slightly later because the matrix slows gastric emptying, with the peak landing closer to 60 to 90 minutes in most studies. Sugar-free pre-workout drinks behave more like capsules than coffee because the matrix is mostly water. None of those three onset profiles works for an athlete who needs the dose to land 15 minutes before a sprint start or 90 minutes into a marathon when the legs start to slow.
Caffeine gum is the practical answer to the onset problem. Ryan and colleagues showed in 2013 that caffeinated gum improved cycling performance most when taken immediately before exercise.9 Gum can deliver caffeine faster than capsules because part of the dose can absorb through the buccal mucosa while the rest is swallowed. Two pieces of typical caffeine gum (about 200 mg) delivers a useful sub-maximal dose with a peak around 15 to 30 minutes after chewing starts.
The practical decision matrix below is what most athletes actually need.
| Scenario | Best form | Timing | Why |
|---|---|---|---|
| Pre-race, 60 min available | Capsule or coffee | 45 to 60 min pre | Stable peak that holds across the first hour of effort. |
| Pre-race, less than 30 min available | Gum | Chew 10 to 15 min before start | Faster absorption can help the dose land closer to the gun. |
| Mid-race redose at hour 2 or 3 | Gum or caffeinated gel | At the moment the legs start to dim | Slow GI absorption defeats the purpose of a mid-race redose. |
| Morning lifting session | Capsule, coffee, or pre-workout | 30 to 45 min pre | Capsule and coffee timed before the warm-up land peak at top sets. See fueling early morning training. |
| Late afternoon or evening session | Lowest effective dose, capsule | 60 min pre, ≥6 h before bed | Half-life is the constraint, not absorption speed. |
| Combat sport before a 5 round fight | Capsule, low dose | 45 min pre | Avoid pre-workout proprietary formulas that stack 4 stimulants. |
| Field sport with 90 min warm-up | Half dose pre warm-up + half dose at kickoff via gum | Staggered | Avoids peak landing in the wrong period of the match. |
Pre-workout proprietary formulas introduce a separate problem. The caffeine dose is often disclosed but the other stimulants (theobromine, theacrine, synephrine, yohimbine, dynamine, dimethylethylamine analogs) stack additively on heart rate and arousal. Many of those compounds extend caffeine's apparent duration and shift the safe ceiling downward. If your goal is replicable performance dosing, single-ingredient caffeine wins on transparency and predictability. The pre-workout matrix is fine for general training. It is a poor tool for a calibrated competition dose.
04Habituation patterns and their partial impact on ergogenesis
The most common claim about caffeine is that daily use destroys the performance effect. The evidence points to tolerance that depends on dose history and study length. Acute performance benefits can remain during shorter daily-use periods, but longer loading can erase the measured benefit.
Beaumont and colleagues ran a 28-day caffeine-loading study in trained men in 2017 published in the Journal of Sports Sciences.10 Participants escalated daily caffeine intake from 1.5 mg/kg/day in week 1 to 3.0 mg/kg/day in weeks 3 and 4. Cycling time-to-exhaustion at 80 percent VO2 peak after a 3 mg/kg acute dose improved early in the loading period, then the performance benefit was no longer apparent after 4 weeks of daily caffeine.
Lara and colleagues published a parallel finding in 2019. After 20 days of 3 mg/kg/day caffeine intake in 11 active participants, an acute 3 mg/kg dose still increased peak cycling power during the Wingate test by about 4.9 percent on several test days.11 The improvement was smaller than the naïve-user response but it was still there.
The practical implication is that daily caffeine intake should stay below the performance dose when the race-day effect matters. What helps is restraint at the high end. A 3 mg/kg race dose has more room to work in someone who has held habitual intake below 2 mg/kg/day. For a major race, dropping habitual intake by half for the preceding 7 to 10 days and then redosing on race day is a defensible approach, and one with fewer side effects than the multi-week withdrawal sometimes prescribed.
The metabolic adaptation still matters. Dropping caffeine cold during a hard training block usually backfires because withdrawal symptoms can disrupt training quality. Taper instead of quitting. Run the taper during the back end of a low-volume week rather than during a peak training block.
05Caffeine's dose-dependent effects on sleep architecture
The single most underweighted variable in caffeine prescribing is sleep. The dose-by-modality table assumes the athlete is sleeping well enough to absorb a training stimulus. When caffeine is what makes that sleep insufficient, the performance ledger inverts inside a week.
Drake and colleagues are the canonical reference. Their 2013 trial gave 400 mg of caffeine at three timing points: 0, 3, and 6 hours before bed.12 All three reduced total sleep time. The dose taken 6 hours before bed cut sleep by 1 hour and 4 minutes compared to placebo, and the subjects did not perceive the disruption. That last part is what makes caffeine sleep cost so insidious. People who feel like they slept fine after a 4 PM cappuccino may still lose meaningful sleep time without noticing.
Gardiner and colleagues published a 2023 systematic review of 24 caffeine-and-sleep trials.13 Across the pooled caffeine trials, total sleep time fell by an average of 45 minutes, slow-wave sleep fell by about 11 minutes, and sleep latency increased by about 9 minutes. The same review estimated that a 107 mg coffee needs at least 8.8 hours before bedtime to avoid reducing total sleep time, while a 217.5 mg pre-workout serving needs about 13.2 hours.
The decision rule that follows is simple. If you are dosing caffeine for performance, you need to know the time between the dose and your bedtime, and you need to know whether the dose is coffee-sized or performance-sized. The defaults below use Gardiner's modeled cutoffs as the floor.
| Bedtime | Coffee-sized dose near 100 mg | Performance dose 200 to 400 mg |
|---|---|---|
| 10:00 PM | 1:00 PM | 9:00 AM or earlier |
| 11:00 PM | 2:00 PM | 10:00 AM or earlier |
| 12:00 AM | 3:00 PM | 11:00 AM or earlier |
Known CYP1A2 slow metabolizers should move the performance-dose cutoff 2 to 4 hours earlier or avoid the dose. Intermediate metabolizers should add 1 to 2 hours. The Drake numbers come from a 400 mg dose, which is at the top of the practical performance range, so the strict end of the rule belongs to larger capsules, gels, and pre-workout servings rather than a small morning coffee.
Late-day racing creates the worst dose conflict. A 6 PM team-sport game might call for 3 mg/kg, and that dose at 5 PM sits inside the sleep-risk window for a 10 PM bedtime. The pragmatic answer in those scenarios is to use the lower bound of the dose range, take it as gum to land peak earlier, and accept that the night after the game is part of the cost of the dose. Stacking caffeine with poor sleep across a multi-day tournament is where the supplement turns into a problem.
06Sex-specific pharmacokinetics across cycle phases, contraceptives, and pregnancy
The caffeine-PK literature historically excluded women, which is partly why so much performance dosing reads as body-mass-normalized despite limited population normalization. The female-specific data that exists points to several practical effects.
Lane and colleagues showed in 1992 that systemic caffeine clearance is slower during the luteal phase than the follicular phase, although caffeine half-life did not differ.14 That points to slightly greater accumulation with repeated daily intake near menses, rather than a reliable single-dose half-life shift. For lifters and runners tracking their cycle-synced fueling, this matters most when afternoon caffeine is repeated across several late-luteal days.
Oral contraceptives extend caffeine half-life further. In one low-dose oral contraceptive study, mean caffeine half-life increased from 5.37 to 7.88 hours.15 That is a large enough effect that women on combined oral contraceptives should treat themselves as more caffeine-sensitive for sleep-timing purposes.
Pregnancy increases half-life dramatically. By the third trimester, caffeine half-life can reach 15 hours.16 EFSA says caffeine intake up to 200 mg/day from all sources does not raise safety concerns for pregnant women in the general population.17 Pregnant athletes should rely on obstetric guidance rather than performance dosing protocols.
Postpartum and lactating athletes return toward normal caffeine clearance over the first few months. A useful default is to add 3 to 4 hours to the sleep cutoffs above during pregnancy and for the first 8 weeks postpartum.
07Fourteen-day individual calibration protocol
You will calibrate faster than you think. The protocol below collapses several months of trial-and-error into two structured weeks.
Days 1 to 3 (baseline). Drop habitual caffeine to a single 100 to 150 mg dose with breakfast. Log sleep duration, resting heart rate, and a perceived energy score from 1 to 10 each evening. Expect a 1 to 2 day headache if your habitual intake was above 300 mg/day. This is the lowest your baseline can reasonably go without disrupting work.
Days 4 to 6 (small-dose probe). Take 2 mg/kg of caffeine in capsule form 60 minutes before your hardest training session. Note RPE, total work, and how you sleep that night. The AA-genotype response is usually a noticeably easier session with no sleep cost. The AC response is a slightly easier session with a 15 to 30 minute sleep cost. The CC response is usually flat session quality plus elevated resting heart rate and a measurable bedtime delay.
Days 7 to 9 (mid-dose). Repeat at 3 mg/kg. For most athletes, this is where the ergogenic effect becomes obvious if your genotype responds at all. If 3 mg/kg produces no perceptible benefit and costs you sleep, you are probably AC or CC and a higher dose will make the tradeoff worse. If 3 mg/kg produces an obvious benefit, you may already have the working dose.
Days 10 to 12 (high-dose probe). Repeat at 5 mg/kg, capped at 400 mg total caffeine from all sources for the day, only if days 7 to 9 felt productive. This dose is at the upper bound of the ISSN range. Pay attention to GI symptoms, post-session heart rate recovery, and sleep onset. Many athletes find their personal ceiling between 3 and 4 mg/kg and never benefit from higher doses.
Days 13 to 14 (form check). Take your working dose half as capsule, half as gum. The split lets you compare onset and peak feel without losing the calibration on absolute dose. Many lifters and team sport athletes find that the gum-half version feels sharper at the right time without the long tail.
At the end of 14 days, you have a working dose, a working form, a working time-to-event, and a personal cutoff. That is what the supplement is supposed to deliver. The number on the label was never the point.
08Conditions where caffeine impairs rather than enhances performance
A few situations call for not dosing at all.
A short race in the morning following a poor sleep night. Caffeine will mask the perceived fatigue but the autonomic load is still there, and the dose will block the recovery nap that would actually help. Skip the dose, accept the 1 to 3 percent performance hit, and protect the next 48 hours of sleep instead.
A high-anxiety taper week. The pre-race nerves that come with a goal event interact with caffeine arousal more than people expect. Caffeine-sensitive athletes will feel the interaction most. A taper week is the wrong time to push caffeine to the top of the dose range. Stick with habitual intake until the morning of the race.
A hard return-to-training block after illness. Heart rate variability is depressed for 7 to 14 days after most respiratory viral infections. Caffeine raises resting heart rate and lowers HRV further. Pull caffeine back to habitual intake or below until the HRV signal returns to baseline.
A pregnancy or active reflux flare. The EFSA pregnancy ceiling sits below most performance doses, and reflux is one of the symptom categories that caffeine can reliably worsen.
A late-evening cardiac symptom workup. Palpitations or unexplained tachycardia warrant medical evaluation before any further stimulant use. Most arrhythmias have causes beyond caffeine, and caffeine can unmask them. The right diagnostic path is medical evaluation before any higher-dose tolerance experiment.
09Expected outcomes following the fourteen-day calibration
The athlete who completes the 14-day calibration usually arrives at a single dose-form-timing combination that they keep for years. The dose is almost always lower than they were taking before, usually by 100 to 200 mg per session. The form mix is almost always two-form: capsule or coffee for sessions with 60 minutes of warning, gum for short-notice or in-event use. The cutoff time is almost always earlier than they wanted it to be.
The performance lift is real. The supplement is well placed in any toolkit that already includes creatine, sodium bicarbonate, and a properly sized carb intake for the event. What changes after the calibration is the confidence interval. You stop wondering whether the dose will work and start treating caffeine the way you treat your warm-up, your fueling timing, and your gear check. A repeatable input that delivers a repeatable, small, useful effect, with a known cost on the back end of the day.
Footnotes
Guest, N., Corey, P., Vescovi, J., & El-Sohemy, A. (2018). Caffeine, CYP1A2 genotype, and endurance performance in athletes. Medicine and Science in Sports and Exercise, 50(8), 1570-1578.
↩Womack, C. J., Saunders, M. J., Bechtel, M. K., Bolton, D. J., Martin, M., Luden, N. D., Dunham, W., & Hancock, M. (2012). The influence of a CYP1A2 polymorphism on the ergogenic effects of caffeine. Journal of the International Society of Sports Nutrition, 9(1), 7.
↩Pataky, M. W., Womack, C. J., Saunders, M. J., Goffe, J. L., D'Lugos, A. C., El-Sohemy, A., & Luden, N. D. (2016). Caffeine and 3-km cycling performance: Effects of mouth rinsing, genotype, and time of day. Scandinavian Journal of Medicine and Science in Sports, 26(6), 613-619.
↩Cornelis, M. C., El-Sohemy, A., Kabagambe, E. K., & Campos, H. (2006). Coffee, CYP1A2 genotype, and risk of myocardial infarction. JAMA, 295(10), 1135-1141.
↩Childs, E., Hohoff, C., Deckert, J., Xu, K., Badner, J., & de Wit, H. (2008). Association between ADORA2A and DRD2 polymorphisms and caffeine-induced anxiety. Neuropsychopharmacology, 33, 2791-2800.
↩Guest, N. S., VanDusseldorp, T. A., Nelson, M. T., Grgic, J., Schoenfeld, B. J., Jenkins, N. D. M., Arent, S. M., Antonio, J., Stout, J. R., Trexler, E. T., Smith-Ryan, A. E., Goldstein, E. R., Kalman, D. S., & Campbell, B. I. (2021). International Society of Sports Nutrition position stand: caffeine and exercise performance. Journal of the International Society of Sports Nutrition, 18(1), 1.
↩Grgic, J., Trexler, E. T., Lazinica, B., & Pedisic, Z. (2018). Effects of caffeine intake on muscle strength and power: a systematic review and meta-analysis. Journal of the International Society of Sports Nutrition, 15, 11.
↩Wickham, K. A., & Spriet, L. L. (2018). Administration of caffeine in alternate forms. Sports Medicine, 48(Suppl 1), 79-91.
↩Ryan, E. J., Kim, C. H., Fickes, E. J., Williamson, M., Muller, M. D., Barkley, J. E., Gunstad, J., & Glickman, E. L. (2013). Caffeine gum and cycling performance: a timing study. Journal of Strength and Conditioning Research, 27(1), 259-264.
↩Beaumont, R., Cordery, P., Funnell, M., Mears, S., James, L., & Watson, P. (2017). Chronic ingestion of a low dose of caffeine induces tolerance to the performance benefits of caffeine. Journal of Sports Sciences, 35(19), 1920-1927.
↩Lara, B., Ruiz-Moreno, C., Salinero, J. J., & Del Coso, J. (2019). Time course of tolerance to the performance benefits of caffeine. PLOS ONE, 14(1), e0210275.
↩Drake, C., Roehrs, T., Shambroom, J., & Roth, T. (2013). Caffeine effects on sleep taken 0, 3, or 6 hours before going to bed. Journal of Clinical Sleep Medicine, 9(11), 1195-1200.
↩Gardiner, C., Weakley, J., Burke, L. M., Roach, G. D., Sargent, C., Maniar, N., Townshend, A., & Halson, S. L. (2023). The effect of caffeine on subsequent sleep: A systematic review and meta-analysis. Sleep Medicine Reviews, 69, 101764.
↩Lane, J. D., Steege, J. F., Rupp, S. L., & Kuhn, C. M. (1992). Menstrual cycle effects on caffeine elimination in the human female. European Journal of Clinical Pharmacology, 43(5), 543-546.
↩Abernethy, D. R., & Todd, E. L. (1985). Impairment of caffeine clearance by chronic use of low-dose oestrogen-containing oral contraceptives. European Journal of Clinical Pharmacology, 28(4), 425-428.
↩Knutti, R., Rothweiler, H., & Schlatter, C. (1981). Effect of pregnancy on the pharmacokinetics of caffeine. European Journal of Clinical Pharmacology, 21(2), 121-126.
↩EFSA Panel on Dietetic Products, Nutrition and Allergies. (2015). Scientific Opinion on the safety of caffeine. EFSA Journal, 13(5), 4102.
↩
