Sample size calculation

The use of exercise to decrease stress in emotionally impulsive participants is not widely developed (Mücke et al. 2018). However, Rimmele et al. (2007, 2009) have shown that healthy trained athletes have a significantly lower change in saliva cortisol levels compared to healthy untrained persons (large effect sizes) after a psychological stress task (i.e. TSST). To the best of our knowledge, only one randomised controlled trial has investigated the effect of a 12-week physical exercise intervention on stress reactivity after the TSST (Klaperski et al. 2014). Here, Klaperski et al. have shown that endurance exercise has better effects on stress reactivity (evaluated via salivary cortisol) than relaxation or passive waiting (moderate main effect size, ƞ² = 0.075) in healthy men. These results are very encouraging, and HIIT has been shown to likely be more efficient than endurance training in inducing metabolic changes (Ketelhut et al. 2016; Nunes et al. 2019; Paahoo et al. 2021) likely to modulate the endocrine stress response. Nevertheless, the effect of two covariates (i.e. self-efficacy and baseline cortisol level) has to be considered. Thus, we will use a covariates-adjusted effect size of 50% inferior to the one reported by Klaperski (ƞ² = 0.0375—small to moderate effect) to determine our sample size. The TSST test–retest reliability (indexed by the area under the curve with respect to the ground [AUC_G]) has been determined with a coefficient of correlation of 0.61 for healthy subjects (Kexel et al. 2021). In addition, based on an a-priori power analysis (G*power v3.1.9.2) for repeated-measures ANOVA (within-between interaction, alpha = 0.05; power = 0.90, repeated-measures correlation = 0.61), 56 participants were required. We considered 15% of dropouts, thus, we aimed to recruit 65 participants.