The reliability of power in tests of physical performance affects the precision of assessment of athletes, patients, clients and study participants. In this meta-analytic review we identify the most reliable measures of power and the factors affecting reliability. Our measures of reliability were the typical (standard) error of measurement expressed as a coefficient of variation (CV) and the percent change in the mean between trials. We meta-analysed these measures for power or work from 101 studies of healthy adults. Measures and tests with the smallest CV in exercise of a given duration include field tests of sprint running (approximately 0.9%), peak power in an incremental test on a treadmill or cycle ergometer (approximately 0.9%), equivalent mean power in a constant-power test lasting 1 minute to 3 hours on a treadmill or cycle ergometer (0.9 to 2.0%), lactate-threshold power (approximately 1.5%), and jump height or distance (approximately 2.0%). The CV for mean power on isokinetic ergometers was relatively large (> 4%). CV were larger for nonathletes versus athletes (1.3 x), female versus male nonathletes (1.4 x), shorter (approximately 1-second) and longer (approximately 1-hour) versus 1-minute tests (< or = 1.6 x), and respiratory- versus ergometer-based measures of power (1.4 to 1.6 x). There was no clear-cut effect of time between trials. The importance of a practice trial was evident in studies with > 2 trials: the CV between the first 2 trials was 1.3 times the CV between subsequent trials; performance also improved by 1.2% between the first 2 trials but by only 0.2% between subsequent trials. These findings should help exercise practitioners and researchers select or design good measures and protocols for tests of physical performance.

This review suggests that there is little or no effect of elevating pre-exercise muscle glycogen contents above normal resting values on a single exhaustive bout of high-intensity exercise lasting less than 5 minutes. Nor is there any benefit of increasing starting muscle glycogen content on moderate-intensity running or cycling lasting 60 to 90 minutes. In such exercise substantial quantities of glycogen remain in the working muscles at the end of exercise. However, elevated starting muscle glycogen content will postpone fatigue by approximately equal to 20% in endurance events lasting more than 90 minutes. During this type of exercise, exhaustion usually coincides with critically low (25 mmol/kg wet weight) muscle glycogen contents, suggesting the supply of energy from glycogen utilisation cannot be replaced by an increased oxidation of blood glucose. Glycogen supercompensation may also improve endurance performance in which a set distance is covered as quickly as possible. In such exercise, high carbohydrate diets have been reported to improve performance by 2 to 3%.

The results of this study show that race time for top triathletes competing over the Olympic distance can be accurately predicted from the results of maximal and submaximal laboratory measures.

We examined neuromuscular activity during stochastic (variable intensity) 100-km cycling time trials (TT) and the effect of dietary carbohydrate manipulation. Seven endurance-trained cyclists performed two 100-km TT that included five 1-km and four 4-km high-intensity epochs (HIE) during which power output, electromyogram (EMG), and muscle glycogen data were analyzed. The mean power output of the 4-km HIE decreased significantly throughout the trial from 319 +/- 48 W for the first 4-km HIE to 278 +/- 39 W for the last 4-km HIE (P < 0.01). The mean integrated EMG (IEMG) activity during the first 4-km HIE was 16.4 +/- 9.8% of the value attained during the pretrial maximal voluntary contraction (MVC). IEMG decreased significantly throughout the trial, reaching 11.1 +/- 5.6% during the last 4-km HIE (P < 0.01). The study establishes that neuromuscular activity in peripheral skeletal muscle falls parallel with reduction in power output during bouts of high-intensity exercise. These changes occurred when <20% of available muscle was recruited and suggest the presence of a central neural governor that reduces the active muscle recruited during prolonged exercise.