Improper Assessment of the Effect of Ad Libitum Drinking on Cycling Performance
In a recent article in Medicine & Science in Sports & Exercise, Bardis et al. (1) evaluated the effect of prescribed drinking (PD) versus ad libitum drinking (ALD) during what they called a simulated 30-km criterium-like performance in the heat. On the basis of their data, we believe that the authors incorrectly concluded that PD sufficient to match sweat losses provides a performance advantage during cycling in the heat compared with ALD.
Bardis et al. (1) used a performance test consisting of three sets of a 5-km bout at 50% of maximal power output, followed by a 5-km all-out hill climb at 3% grade. The authors analyzed and interpreted their results by isolating and dichotomizing the performance times achieved during each hill climb without consideration of the fact that the overall performance during a cycling competition is a representation of the sum of each race segment, and not of any section in particular without taking into account the others.
The authors report that cycling speeds were not significantly different between ALD and PD during the first (30.3 ± 2.3 vs 29.2 ± 2.7 km·h−1) and second (29.8 ± 2.1 vs 29.2 ± 2.4 km·h−1) 5-km hill climb. However, the subjects completed the third 5-km hill climb faster with PD than ALD (30.2 ± 2.4 vs 28.8 ± 2.6 km·h−1, P < 0.05). The conclusions of Bardis et al. were based on those observations.
A computation taking into account all three performance bouts reveals that the ALD group completed the simulated race at a faster mean speed (29.6 vs 29.5 km·h−1) and time (30.4 vs 30.5 min) than the PD group. More striking is to consider the fact that at the start of the third hill climb, the ALD group was 282 m ahead of the PD group, and that by the time the ALD group had reached the finish line, the PD group was trailing by an impressive and practically relevant distance of 50 m. Clearly, the conclusion of Bardis et al. is not only erroneous but also misleading. Likely, the appropriate statistical analysis taking into account the overall mean speed will show that there was no performance difference between drinking strategies, which would be in line with results of Dugas et al. (3) who demonstrated no difference in performance between PD and ALD during an 80-km cycling time trial in the heat.
Recently, Bardis et al. (2) evaluated the effect of mild dehydration on cycling performance using a similar research protocol consisting of three consecutive sets of a low-intensity, fixed-power output 5-km ride, an all-out 5-km hill climb, and a 5-min rest period. Ironically, in that work, the authors report the analysis of mean speed maintained by their subjects throughout the testing protocol. We find it interesting that the authors reported that analysis in their earlier work but not in the present paper and would appreciate comment on this change in methods.