Stress and recovery in sports: Effects on heart rate variability, cortisol, and subjective experience

Athletes face increasing training and competition demands that can disrupt the stress–recovery balance and lead to non-functional overreaching (NFOR). Monitoring tools targeting psychological states (RESTQ-Sport), hypothalamic–pituitary–adrenal (HPA) activity (salivary cortisol), sympathetic–adrenal–medullary (SAM) activity (salivary alpha-amylase; sAA), and autonomic regulation (heart rate variability; HRV) enable multidisciplinary assessment of allostasis. Prior work indicates cortisol awakening response (CAR) stability and the utility of sAA, though findings are inconsistent regarding long-term training effects on sAA and cortisol. HRV, particularly vagal-related indices (RMSSD), reflects adaptation status. Recovery–stress and emotional states may follow different temporal trajectories than HRV during functional periodization, but most associations with training load have been examined between subjects, risking biased results. This study examined, during a 12-week ecological training periodization in swimmers, (1) time variations in performance, recovery–stress, emotions, HPA/SAM, and HRV, and (2) repeated-measures intra-individual associations between training load and these markers. The hypotheses were that performance, recovery–stress, and emotions would show quadratic trajectories, HRV would show linear trajectories, and salivary markers would have no specific hypothesized change; training load would correlate with recovery–stress, emotions, parasympathetic markers, and the AOC ratio.

Allostatic responses involve SAM (catecholamines) and HPA (cortisol). Salivary cortisol and CAR provide noninvasive HPA assessment with high within-person stability; cortisol is also a training biomarker. sAA has been proposed as a marker of autonomic activity, yet long-term training effects are mixed: some report increases with chronic stress, others no changes, and recent small studies show training-related variations. Given HPA–SAM coordination, the sAA/cortisol ratio (AOC) may better reflect chronic stress system dysregulations than either marker alone. HRV is widely used to monitor ANS status; increases in vagal indices (e.g., RMSSD) are linked to positive training adaptation, while reductions can indicate maladaptation. Prior research showed differing trajectories (quadratic vs. linear) among psychological and physiological markers across periodized training. However, most studies used between-subject designs; within-subject repeated-measures associations remain underexplored, which may lead to Simpson’s paradox when interpreting longitudinal training data.

Design: Ecological longitudinal study over 12 weeks with three test periods: REST (baseline after 1 week off), PREP (after 6 weeks of intensive preparation), and TAPE (after 6 weeks of tapering). Participants: Fifteen national-level swimmers (9 males, 6 females; mean age 17.8 ± 1.1 years; height 1.76 ± 0.06 m; BMI 20.70 ± 1.15 kg·m⁻²), training 6 days/week (~14.0 ± 1.8 h/week plus 3.5 h/week of physical training). Inclusion criteria ensured no smoking, medication, supplements, or contraceptives; mixed-sex sample justified by lack of sAA sex differences. Ethical approval obtained; informed consent provided. Training load: Internal subjective TL (TLin; arbitrary units) calculated via session-RPE (CR-10) × session duration; daily TLin summed; weekly totals computed. External objective TL (TLex; kilometers) computed from session volume. Between REST and PREP: TLin increased +1233%, TLex +725%; between PREP and TAPE: TLin −78%, TLex −60%. Protocol per test day: (1) Saliva collection; (2) Psychological questionnaires; (3) Submaximal 5′ run + 5′ passive recovery (5′-5′); (4) Performance test (race-specific maximal effort; speed in m·s⁻¹). Salivary assessment: Four samples per test day—upon waking (7:00 a.m.), +30 min, +60 min, and 8:00 p.m. (pre-evening meal). Salivettes used; passive drool for 3 min; pre-sampling restrictions (no alcohol, caffeine, fruit juice 60 min; no brushing 30 min). Samples stored at −20 °C; centrifuged 10 min at 3000 rpm; flow rate computed (valid if ≥0.1 ml·min⁻¹). Cortisol assayed via Salimetrics EIA (sensitivity 0.82 nmol·l⁻¹; intra-assay CV 3.7%; inter-assay CV 6.4%). sAA assayed via kinetic reaction (Salimetrics; detection 0.078 U·ml⁻¹; intra-assay CV 6.7%; inter-assay CV 5.8%). All samples run in duplicate in the same series; sAA not corrected for flow or protein concentration. Psychological measures: RESTQ-36-R-Sport (French version) assessing general/sport-specific/total stress and recovery (Likert 0–6; reliability α 0.82–0.92). Sport Emotion Questionnaire (SEQ; French translation) assessing anxiety, dejection, anger, excitement, happiness (0–4; reliability α 0.72–0.87). Heart rate and HRV: Standardized 5 min submaximal run at 9 km·h⁻¹ on pool deck, followed by 5 min passive recovery. HR/HRV recorded via Suunto t6 Memory Belt. Outcomes: exercise heart rate (HRex; last minute of run), post-exercise HR recovery (nHRR60s = (HRex − HR60)/HRex), MeanRR and RMSSD computed from last 2 min of the 5 min resting period using Kubios; RMSSD log-transformed (lnRMSSD). Artifact correction applied. Performance: Race fitting specialty; maximal effort; performance expressed as swimming speed (m·s⁻¹). Statistics: Normality (Shapiro–Wilk) and homoscedasticity (Levene) tested. Repeated-measures ANOVA with a priori intra-individual contrasts for linear and quadratic time effects; Greenhouse–Geisser corrections when sphericity violated; Bonferroni-adjusted pairwise t-tests. Balance checks via dependent t-tests for stress–recovery and pleasant–unpleasant emotions within each period. Repeated measures correlations (rmcorr package in R) computed for within-subject associations between TL markers and outcomes, reporting rrm, 95% CIs, and p-values; magnitude interpreted as trivial to almost perfect per Hopkins et al. Alpha = 0.05.

Sample characteristics and compliance were adequate; no sex differences or time×sex interactions for psychological, physiological, or salivary markers. Time trajectories: - Performance: Significant quadratic trajectory (F(1,14) = 26.86, p < 0.001, η² = 0.67); speed decreased from REST to PREP (t(14) = 3.37, p = 0.005, d = 0.87), then increased from PREP to TAPE (t(14) = 6.01, p < 0.001, d = 1.56); speed at TAPE > REST (t(14) = 3.35, p = 0.005). - Training loads: TLin and TLex both showed significant quadratic trajectories (TLin F = 340.65, p < 0.001, η² = 0.96; TLex F = 292.72, p < 0.001, η² = 0.95) with marked increase at PREP and reduction at TAPE. - Recovery–stress states: General, specific, and total stress increased from REST to PREP and decreased from PREP to TAPE (quadratic trajectories; η² = 0.56–0.68), with no REST vs TAPE differences. Recovery (general, specific, total) decreased from REST to PREP and increased from PREP to TAPE (quadratic; η² = 0.54–0.72). - Emotions: Anxiety, anger, dejection (unpleasant) increased from REST to PREP and decreased from PREP to TAPE; happiness and excitement (pleasant) decreased from REST to PREP and increased at TAPE (quadratic; η² = 0.27–0.48). - Salivary biomarkers: No significant time variations for cortisol (ACO, CAR, AUCgc) or sAA (AAA0, AAR, AUCgsAA) across periods (p = 0.06–0.85). - HR/HRV: Linear trajectories for HRex (F = 10.58, p = 0.006, η² = 0.43), MeanRR (F = 9.00, p = 0.010, η² = 0.39), lnRMSSD (F = 13.15, p = 0.003, η² = 0.48). HRex decreased from REST to TAPE; MeanRR and lnRMSSD increased from REST to PREP and from REST to TAPE. Repeated-measures correlations (within-subject): - Performance vs TL: Swimming speed negatively correlated with TLex (rrm = −0.52, p = 0.003, CI −0.74; −0.19) and TLin (rrm = −0.60, p = 0.001, CI −0.79; −0.30). - Recovery–stress vs TL: Total stress positively correlated with TLex (rrm = 0.69, p < 0.001) and TLin (rrm = 0.71, p < 0.001); total recovery negatively correlated with TLex (rrm = −0.60, p < 0.001) and TLin (rrm = −0.68, p < 0.001). General/sport-specific dimensions showed similar directions (general recovery TLex rrm = −0.74; TLin rrm = −0.77; general stress TLex rrm = 0.70; TLin rrm = 0.71). - Emotions vs TL: Unpleasant emotions correlated positively with TL (anxiety TLex rrm = 0.50, TLin rrm = 0.54; dejection TLex rrm = 0.53, TLin rrm = 0.50; anger TLex rrm = 0.56, TLin rrm = 0.50). Pleasant emotions correlated negatively (excitement TLex rrm = −0.37, p = 0.05; TLin rrm = −0.42, p = 0.02; happiness TLex rrm = −0.51, p = 0.01; TLin rrm = −0.53, p = 0.01). - HRV vs TL: lnRMSSD moderately and positively associated with TLex (rrm = 0.42, p = 0.019); associations with TLin were positive but not significant (rrm = 0.28, p = 0.12). - Salivary vs TL: AOC ratio negatively associated with TLin (rrm = −0.37, p = 0.04); associations with TLex trended negative (rrm = −0.33, p = 0.07). Individual cortisol or sAA indices showed no significant associations.

Findings support that during functional periodized training, psychological and physiological systems exhibit distinct dynamics: recovery–stress and emotions follow quadratic trajectories (impairment during intensified training, improvement during taper), whereas parasympathetic cardiac markers (MeanRR, lnRMSSD) show linear improvements indicative of positive fitness adaptation. Salivary cortisol and sAA did not vary across periods, suggesting no HPA/SAM dysregulation in this training context without competition stressors. Within-subject associations reveal that increases in training load are accompanied by higher stress and unpleasant emotions and lower recovery and pleasant emotions, consistent with periodization effects; yet parasympathetic activity increased with external load, aligning with functional overreaching and adaptation. The AOC ratio’s negative association with internal load, despite no changes in its components, bolsters its potential sensitivity to subtle stress-system coordination shifts compared to cortisol or sAA alone. Overall, the intra-individual analytic approach demonstrates meaningful links between training load and internal markers, helping differentiate functional responses from potential maladaptive states and offering practical guidance for monitoring and adjusting training to maintain an adequate stress–recovery balance.

A 12-week functional training periodization in national swimmers produced quadratic trajectories in performance, recovery–stress, and emotional states, linear improvements in parasympathetic HRV markers, and no significant changes in salivary cortisol or sAA. Within individuals, training load correlated positively with stress and unpleasant emotions and negatively with recovery, pleasant emotions, and the AOC ratio, while external load correlated positively with lnRMSSD. These results highlight the value of multidisciplinary monitoring and repeated-measures correlations for detecting functional adaptations and managing training to avoid NFOR. Future research should refine detection of the threshold at which training load precipitates psychophysiological breakdown and HPA/SAM dysregulation, integrate objective compliance tools (e.g., smartphone-linked sampling and ambient-light measures), and examine sport- and event-specific load profiles to tailor monitoring and interventions.

Generalizability is limited to high-level athletes; findings may not extrapolate to the general population. Training regimes varied by specialty (sprinters, middle distance, combined events), potentially influencing marker responses. Saliva self-collection in ambulatory conditions may introduce variability; compliance was assessed via self-report rather than objective measures. Ambient light exposure and socioeconomic status were not controlled. Small sample size favored homogeneity but constrains broader inference.