Reply to Borszcz & de Lucas: Comment on: “Effects of Carbohydrate Mouth Rinse on Cycling Time Trial Performance: A Systematic Review and Meta-Analysis”

Abstract

We read with interest the letter by Borszcz and de Lucas [1] commenting on our recent systematic review and metaanalysis [2] and subsequent correspondence [3, 4]. While we think some of the points raised by the correspondents may be insightful for readers, we have some concerns about others. We conducted our systematic review and meta-analysis assuming that carryover effects would be negligible in carbohydrate (CHO) mouth rinse study designs given the nature of treatment and washout usually observed between trials, and that the use of the standardized mean difference (SMD) and standard error (SE) suggested for parallel group designs should be conservative when identifying eventual beneficial CHO effects on performance [5]. However, crossover designs have advantages because each participant works as his/her own control so that more power is obtained with the same number of participants when compared to parallel group studies [6]. Therefore, we agreed with Borszcz and de Lucas [1] that our original meta-analysis could have overlooked some CHO mouth rinse effects, and performed more appropriate SMD and SE calculations for crossover designs. Accordingly, we first reanalyzed our SMD and SE, and thereafter corrected them for overestimation bias derived from small sample sizes (Tables 1, 2), as recommended by Borenstein et al. [20]. Similar findings were obtained as from our initial conservative approach, with neither power output (Fig. 1) nor time (Fig. 2) being different from results presented in our original meta-analysis [2]. Even when using calculations recommended for crossover designs we observed that CHO mouth rinse has the potential to improve cycling mean power output, but time does not. Importantly, sensitivity analysis revealed that these results (not shown) were consistently observed in a number of intrastudy correlation coefficients likely observed in crossover design studies (i.e. r from 0.50 to 0.99). It is important to note that, in contrast to Borszcz and de Lucas [1], we maintained the same original dataset when calculating SMD and SE for performance outcomes, as we have some concerns about the estimation of time of exercise suggested for the Lane et al. [16] results. In their letter [1], the authors argued that: “Therefore, ∆% for power = (∆% for speed, distance or time)x, where for the ergometer used by Lane et al., x = 2.2 (https ://www.cycle ops.com/post/blog-15-cycle ops-scien ce-resis tance -curve s). For example, a change of 1% in speed/distance or − 1% in time is equivalent to a change of 2.2% in power”. They then concluded that the percentage of change in time in the Lane et al. [16] data could be estimated directly using a power–time relationship of 2.2 [1]. Two aspects should be pointed out. First, we tried to understand the meaning of applying this approach to Lane et al.’s [16] data, as that study utilized a cycling trial closed by time (60 min) and reported mean power output as the main performance outcome. Therefore, it is difficult to understand how a measure that has been used as a criterion to delimit the test termination could be used as a performance indicator itself (i.e. This letter refers to the original article available at https ://doi. org/10.1007/s4027 9-018-1029-7.