Caffeine on Marathon Race Day: Dose, Timing, and Which Gels to Use

Caffeine is the most reliably ergogenic thing you can put in your body on race morning, and most marathoners get it wrong in one of two directions: a random amount of coffee at 5:00 AM, or nothing at all.

The number to work from is 3-6 mg per kilogram of body mass, taken roughly 60 minutes before the start. That is the range the International Society of Sports Nutrition landed on in its 2021 position stand (Guest et al., Journal of the International Society of Sports Nutrition) after reviewing the literature. More recent work has pushed the floor down: doses as low as 2 mg/kg, and in some trials below that, produce measurable gains without the jitters. For a 70 kg runner, the classic range is 210-420 mg. The low-dose range is closer to 100-200 mg.

Then top up once in the last third of the race, with a caffeinated gel somewhere around miles 16-20.

That is the whole strategy. The rest of this is how to execute it, and which gels actually carry the caffeine they claim.

What caffeine actually does in the last 10K

Caffeine is not energy. It supplies no meaningful carbohydrate and does nothing for the glycogen problem that ends most marathons. It works as an adenosine receptor antagonist. Adenosine builds up as you fatigue and binds to receptors that make you feel tired and heavy, and caffeine occupies those receptors instead. The output is that a given pace feels easier than it should.

That effect is measurable. A meta-analysis cited in the ISSN position stand found an average 5.6% reduction in rating of perceived exertion after caffeine ingestion, and calculated that the RPE reduction accounted for up to 29% of the performance improvement.

Note the direction of that. The pace does not get physiologically cheaper. It gets perceptually cheaper, and you hold it longer as a result.

Which tells you exactly when the dose matters. Perceived exertion is not the limiter at mile 4. It is the limiter at mile 22.

The dose math by body weight

Everything in the caffeine literature is expressed per kilogram, which means a 60 kg runner and an 80 kg runner should not be taking the same gel count. Here is what the ranges work out to:

Practical target is the total for race day, split between a pre-race dose and in-race gels.

The top of the ISSN range is a lot of caffeine. 480 mg is four or five cups of coffee, taken in one sitting, before running 26.2 miles. The position stand notes that very high doses around 9 mg/kg produce side effects without added benefit, and you hit diminishing returns well before that. For a marathon, the middle of the range is the useful zone.

Start at the low end if you have not tested this. Caffeine response varies widely between individuals, and the ISSN attributes part of that to genotype, including ADORA2A variants that predict how badly a dose wrecks your sleep. Your training partner’s 400 mg protocol is not transferable evidence.

Pre-race timing: coffee versus a gel at the start line

Caffeine peaks in the blood roughly 45-60 minutes after you take it, and the ISSN reports that 60 minutes pre-exercise is the most commonly used protocol in the research. That is the window you are aiming at.

Coffee works. The ISSN position stand says so directly, with plenty of supporting data. The problem is precision. Brew method, bean, and shop all move the number, so the dose you took before your best long run is not a dose you can reproduce. A caffeinated gel or a tablet gives you a printed milligram figure you can test in training and repeat on race day.

The real timing trap is wave starts. At the NYC Marathon, a late-wave runner clears security on Staten Island hours before the gun and can sit at Fort Wadsworth for four hours after breakfast. Coffee at 5:00 AM for an 11:00 AM start is a spent dose. Caffeine’s half-life runs roughly three to six hours in most people, so by the time you cross the Verrazzano you are well past peak and on the downslope.

At any race with a long wait between security and your corral, move the dose. Take a caffeinated gel or a tablet 45-60 minutes before your wave is called, not before you get on the bus.

In-race caffeine: the last third, not the first

There is direct evidence for dosing late rather than early. Cox and colleagues (Journal of Applied Physiology, 2002) had trained cyclists take caffeine in the final stages of a two-hour ride before a time trial. Even at low doses of around 1.5 mg/kg, the late dose improved time trial performance by 2.2%, and the cola trial came in at 3.3%.

Translated to a marathon, the caffeinated gel goes in around miles 16-20, which is roughly two to two and a half hours into a 3:00-3:30 race. Take it on flat ground with water. It is fully working from around mile 19 to the finish, which is precisely the stretch where perceived exertion starts dictating your pace.

The obvious next thought is to caffeinate every gel. Don’t.

Run the arithmetic first. Six Maurten Caf 100 gels is 600 mg, or about 8.5 mg/kg for a 70 kg runner, which puts you in the territory the ISSN describes as side effects without added benefit. You are also stacking a stimulant on top of a gut already processing 80-100g of carbs per hour. The marathon is not the place to introduce a second variable to a digestive system running at its ceiling.

A workable split for a 70 kg runner targeting about 300 mg:

150-200 mg 45-60 minutes before the gun. One caffeinated gel at mile 16-18. Optionally one more at mile 21-22 if you have trained it. Everything else caffeine-free.

If you are fueling at 90g/hr, that is six to eight gels over the race, of which one or two carry caffeine. Work out how many gels your carb target actually requires in the gel count breakdown, then build the timing in the fueling calculator.

Habitual users, abstainers, and the taper question

This is where the internet is more confident than the evidence is.

The case for tapering: research has found that habitual caffeine consumption reduces the ergogenic effect of an acute dose in both aerobic and anaerobic trials, and at least one study found that short-term withdrawal restored some of the effect on 3 km running performance.

The case against: the ISSN position stand calls the findings equivocal and cites a well-designed study that found no influence of habitual intake at all. A four-day withdrawal period failed to improve caffeine ergogenicity in trained cyclists. The position stand also flags a methodological problem that undercuts the whole taper argument: short abstinence periods of 12-48 hours may confound results, because what looks like an ergogenic gain is partly the reversal of withdrawal rather than a true performance effect.

The honest summary is that nobody has settled this, and the evidence base is small relative to how often the advice gets repeated.

The practical read: a multi-day caffeine taper buys you headaches and bad sleep in the exact week you need both, in exchange for a benefit that may not exist. If you want to test it, test it before a tune-up race, not before the marathon. What is worth doing is not escalating your habitual intake during the final weeks of a block, when you are tired and reaching for another cup.

GI and sleep: the two costs

GI. Caffeine on its own is not a common cause of race-day stomach failure. The risk is additive. At 90-100g of carbs per hour you are already asking a lot of the gut, and a few hundred milligrams of a stimulant is a new input on top of that. The answer is the same as it is for every other fueling variable: test the full protocol, at race pace, at race carb rate, in a long run. Not the caffeine alone and the carbs alone. Both together, at effort.

Sleep. Race morning caffeine is rarely the problem. The taper week is. With a half-life of roughly three to six hours, afternoon caffeine in the week before a marathon is still circulating at bedtime, and the ISSN notes that sleep sensitivity to caffeine is partly genetic via ADORA2A variants. Some people can drink an espresso at 6:00 PM and sleep fine. Some cannot. If you are in the second group, the week before your goal race is when it costs you the most.

Caffeinated gels: what is actually in them

Caffeine content varies more than any other spec in this category, and it varies by flavor within a single product line. Every figure below is from the manufacturer’s own product page:

Caffeine varies by flavor within GU, Huma, and Roctane lines. Check the packet, not the brand.

Two things fall out of that table. First, the spread is enormous: one SiS Beta Fuel + Nootropics gel carries the caffeine of ten standard caffeinated GU gels, and SiS itself recommends a limit of one serving per day. Second, the caffeinated version usually carries the same carb load as the caffeine-free version of the same product, so swapping one gel in your plan for its caffeinated twin does not disturb your carb math at all.

That is the cleanest way to build this. Pick your gel from the gel rankings on carbs, price, and how it sits in your stomach. Then swap the one or two packets that land in miles 16-22 for the caffeinated version of the same gel. Same brand, same texture, same carb count, no new variable except the one you meant to add.

FAQ