We investigated the acceptability of power variation during a cycling time trial (TT) with simulated uphill and downhill sections. Seven cyclists first completed an 800-kJ self-paced TT on a simulated flat course. An 800-kJ TT course with four sections of uphill/downhill was then modeled. Each section involved 100 kJ of cycling up a simulated gradient of 5 % followed by 100 kJ of riding down a simulated gradient of - 5 %. Participants were required to complete this simulated course using two pacing strategies; (i) at a constant power equivalent to the mean power achieved during the initial TT, and (ii) increasing power by 5 % of mean power when traveling uphill (mean duration of each climb = 714 s) and decreasing power in the downhill sections (mean duration of each descent = 190 s), so that overall mean power was equivalent to that in (i). All participants maintained this variable power strategy during the first half of the TT, but two riders could not adhere to the power variations during the final 400 kJ. Nevertheless, mean ± SD finish time for the variable power trial (3670 ± 589 s) was significantly faster than that for the constant power TT (3758 ± 645 s), the 95 % confidence interval for the percentage improvement being 0.4 to 4.3%. Heart rate and lactate responses were highest in the initial self-paced TT and did not differ between the subsequent constant and variable power trials. Ratings of perceived exertion were also similar between trials. In our externally-valid TT, we found that some cyclists cannot fully adhere to a pacing strategy involving an approximate ± 5 % variation in mean power in parallel with gradient variation. Nevertheless, an important time saving can still result even if a variable pacing strategy is only partially adopted during a hilly time trial, so that no additional physiological strain is incurred.