REF-1621 · Kilding, Overton & Gleave, 2012 · International Journal of Sport Nutrition and Exercise Metabolism, Vol. · Paper · PubMed

Caffeine and sodium bicarbonate did not add up: in 10 well-trained cyclists, each raised mean power in a 3-km time trial over placebo (caffeine 3 mg/kg +2.4%, bicarbonate 0.3 g/kg +2.6%), and taking both gave +2.7%, with no harm found from the combination

Main caveat: One study of 10 men; small effects judged against a ~1% worthwhile difference, without significance testing in the abstract.
limited evidenceMeasured on trained to highly trained athletesRead from the abstract only

Kilding, Overton and Gleave asked whether caffeine and sodium bicarbonate, two supplements that act through different mechanisms, add to each other's effect on a short, hard effort.

Ten well-trained cyclists rode four 3-km time trials on a laboratory ergometer on separate days, double-blind and in crossover order, after caffeine (3 mg/kg), sodium bicarbonate (0.3 g/kg), both, or a placebo. Bicarbonate raised blood pH and bicarbonate before the trial, alone and in the combination.

Against placebo, mean power rose by 2.4% with caffeine, 2.6% with bicarbonate and 2.7% with both, and times were 0.9%, 1.2% and 1.2% faster; all three effects were small (effect sizes 0.21-0.24). Differences in perceived exertion and gastric discomfort between conditions were mostly unclear.

The authors conclude that each supplement helps on its own, that their effects were not additive, and that taking both did no harm.

Participants

n = 10 well-trained cyclists, age 24.2 ± 5.4 y; men by the PubMed indexing (the abstract does not state sex); training volume, VO2max and competition level not in the abstract.

Protocol

Four 3-km time trials on a laboratory cycle ergometer, on separate days, in crossover order, double-blind.
Conditions: caffeine 3 mg/kg body mass; sodium bicarbonate 0.3 g/kg; both at those doses; placebo.
Measures: pre-trial blood pH and bicarbonate, perceived exertion, gastric discomfort, mean power and time.
Analysis against a minimal worthwhile difference of about 1%, taken from earlier research.
Not in the abstract: when each supplement was taken, how bicarbonate was split, the warm-up, and the riders' training and fitness.

Key Results

Condition vs placeboMean powerTime
Caffeine 3 mg/kg+2.4%−0.9%
Sodium bicarbonate 0.3 g/kg+2.6%−1.2%
Both+2.7%−1.2%
Effect sizes for all three: small, 0.21-0.24.
Pre-trial pH and bicarbonate: higher in the two bicarbonate conditions than with caffeine or placebo.
Perceived exertion and gastric discomfort: differences mostly unclear.

Practical Application

For a short, maximal effort, the two did not stack. Caffeine alone and bicarbonate alone each gave about 2.5% more power in a 3-km time trial (2.4% and 2.6%); both together gave 2.7%, about what either gave alone.
No penalty from the pair either. The combination did not lower performance, and gastric discomfort did not clearly differ between conditions in this sample.
Inference, not tested here: choose one and rehearse it. If one buys about as much as both, the one that sits better with the rider's gut and race routine is the sensible pick; how the base introduces each, one at a time, is in SYNTH-67.
Not settled beyond this design. One small crossover, one 3-km effort, one dose of each; longer races, higher caffeine doses and riders who respond to only one of the two were not studied, and the bicarbonate position stand calls caffeine with bicarbonate unclear (REF-354).

Limitations

One study of 10 men; small effects judged against a ~1% worthwhile difference, without significance testing in the abstract.
One 3-km laboratory time trial per condition; the effect on longer or outdoor efforts is unknown.
Timing and splitting of the doses are not in the abstract.
A crossover of 10 cannot rule out a small extra gain from the combination; it shows only that none was seen.
Written from the abstract and the PubMed record; full text not read.
How this entry was checked and changed
Why this level

a crossover with fewer than 15 participants is B (rubric; n = 10); effects judged against a ~1% minimal worthwhile difference rather than by significance tests

Content correction 2026-11-06
2026-11-06: added from the abstract and the PubMed record; level B (crossover, n = 10). Found through a newsletter summary, which is not a source here; no number in this entry comes from it.