A non-linear association between perceived stress and vagal-related HRV during elite women’s World Tour stage races

INTRODUCTION: Heart-rate variability (HRV) reflects cardiac autonomic modulation and is widely used to monitor training stress and recovery in athletes [1]. The root mean square of successive differences (rMSSD) is particularly sensitive to perturbations in vagal-related activity during short, intensive periods [2]. However, in elite athletes, rMSSD may show a weak or non-linear association with external training load, partly due to complex autonomic regulation and potential "vagal saturation" [2]. Accordingly, brief psychometric questionnaires are gaining popularity as complementary, non-invasive monitoring tools [3][4]. This study aimed to examine the relationship between rMSSD and the Short Recovery and Stress Scale (SRSS) before and during world-level stage races in elite women cyclists. METHODS: Twenty-five elite women cyclists (twenty-one competing in the Giro d'Italia Women 2025 and seven in Le Tour de France Femmes 2025, with some athletes participating in both events) recorded morning HRV immediately upon waking using HRV4Training [5] and completed the SRSS [3] via mobile devices. Missing HRV metrics or the questionnaire data were excluded, yielding a final dataset of 308 complete entries. Associations between log-transformed rMSSD (lnrMSSD) and subjective measures were analyzed using generalized additive mixed models (GAMMs) to test for non-linear effects (edf>1.5). Subsequently, if GAMMs found a nonlinear relationship, the quadratic relationship was tested using linear mixed models (LMMs) with random slopes to account for inter-individual variability and provide a parametric interpretation. RESULTS: GAMMs identified a significant nonlinear association between perceived stress and lnrMSSD (edf=2.68, p<0.05), whereas no association was observed between perceived recovery and lnrMSSD (p=0.18). LMMs with random slopes by athletes supported a quadratic effect (inverted-U) of perceived stress and lnrMSSD (linear and quadratic terms p<0.05), providing parametric confirmation of the non-linear pattern. Conditional R2 indicated that 70.5% of variability in lnrMSSD was explained by between-athlete differences, while marginal R2 for fixed effects was low (1.9%) highlighting pronounced individual specificity in autonomic responses. Perceived stress was the only subjective measure associated with the non-linear lnrMSSD response, whereas recovery scores were not associated with HRV. CONCLUSION: Perceived stress was non-linearly related to lnrMSSD, exhibiting an inverted-U pattern with lower lnrMSSD values at both low and high levels of perceived stress. These findings emphasize the need for individual and non-linear approaches to athlete monitoring, particularly during elite races. 1. Lundstrom et al., 2023 https://doi.org/10.1055/a-1864-9726 2. Schmitt et al., 2015 https://doi.org/10.3389/fphys.2015.00343 3. Kölling et al., 2020 https://doi.org/10.1080/02640414.2019.1684790 4. Esco et al., 2026 https://doi.org/10.3390/s26010003 5. Altini & Amft, 2016 https://doi.org/10.1109/EMBC.2016.7591265