Caffeine trimmed cycling time trials, short and long

Caffeine is the most widely used legal ergogenic aid in endurance sport, but the size and location of its benefit still get argued over. A time trial lasting five minutes and one lasting half an hour sit in very different places on the power-duration curve: the short one is dominated by anaerobic capacity above critical power, the long one by sustainable aerobic output and pacing. If caffeine works through different routes at different durations, that has practical consequences for who bothers with it. This review used the critical power framework to test whether the effect really does depend on event length.

Following PRISMA 2020 guidelines and with prospective PROSPERO registration, the authors searched five databases from inception to 14 January 2026 for randomised, double-blind, placebo-controlled trials of acute pre-exercise caffeine in healthy cyclists. Twenty-three trials met inclusion. Random-effects meta-analyses were run on standardised mean differences using Hedges' g for two outcomes: time-trial completion time and mean power output. Crucially, short-duration trials of 10 minutes or less and long-duration trials of 20 minutes or more were prespecified as subgroups rather than defined after the fact, and sensitivity analyses were performed.

Caffeine reduced time-trial completion time (SMD −0.32, 95% CI −0.49 to −0.16) and increased mean power output (SMD 0.20, 95% CI 0.02 to 0.38), with I² of 0% for both — unusually consistent across trials. Completion time improved in both short-duration (SMD −0.30) and long-duration (SMD −0.37) events, with no significant subgroup difference. Mean power output rose significantly in short trials (SMD 0.27) but did not reach significance in long ones (SMD 0.07), although that between-subgroup difference was also not significant. Sensitivity analyses were consistent throughout.

This confirms caffeine's ergogenic effect on cycling time trials while adding a wrinkle about mechanism. The finish-time benefit held across durations, but average mechanical power only clearly increased in short efforts, which the authors read as evidence that longer time-trial gains involve additional factors beyond raw mean power. Since the subgroup comparisons were not statistically significant, that mechanistic split is a hypothesis rather than a demonstrated difference. The zero heterogeneity strengthens confidence in the pooled estimate; questions remain about dose, habituation, individual response and whether the same pattern holds in less trained riders.

The realistic expectation from caffeine before a cycling time trial is a small but genuine improvement in finish time, not a transformation — and that expectation appears to hold whether the effort is a few minutes or half an hour. In shorter efforts, the improvement seems to show up as more average power; in longer ones, something less easily measured may be doing the work. All of these trials were in cyclists, and none of this speaks to how caffeine interacts with your sleep or your habitual intake. It is a reason to treat caffeine as a marginal gain that reliably shows up in the data.