Cardiorespiratory fitness modulates parasympathetic withdrawal and recovery: Wavelet-estimated HRV across graded exercise phases

This study aimed to determine which wavelet-derived heart rate variability (HRV) indices reliably track autonomic nervous system changes across the key transitions of a graded exercise test (rest, onset, peak, and early recovery) and to assess whether cardiorespiratory fitness (VO 2 max) moderates these transition-specific responses. Although wavelet analysis of HRV accommodates non-stationarity better than traditional spectral methods, frequency-domain power may remain sensitive to rapid, breathing-driven shifts in respiratory frequency. We hypothesized that wavelet time-domain HRV would track autonomic transitions of graded exercise and be moderated by VO 2 max, whereas spectral indices would destabilize, especially from peak to recovery, due to breathing-driven frequency shifts. Methods: This secondary analysis used data from the PhysioNet ACTES database. Eighteen adolescent athletes (8 females; 15.2 ± 2 years) completed a maximal graded cycling test (15 W/min to exhaustion) with continuous R to R intervals, respiratory gas exchange, and power output recorded. R to R intervals were processed using a continuous Morlet wavelet transform (Wavelet Comp package in R) to obtain time–frequency HRV estimates across standard low-frequency (LF) and high-frequency (HF) bands. Four 3-minute segments were extracted: rest, exercise onset, peak exercise, and early recovery. HR, SDNN, RMSSD, and wavelet-derived LF, HF, and total power were computed. Phase-specific autonomic changes were modeled using linear mixed-effects models with VO 2 max as a moderator. Potential moderating effects of age, BMI, and sport type were also explored. Results: HR increased from 78 bpm (95% CI: 73.9–81.3) at rest to 180 bpm (95% CI: 177–183) at peak, while VO 2 max rose from 0.37 to 2.25 L·min - ¹. RMSSD sharply declined from rest to onset (β = −140 ms, p = 0.035). From rest to peak, RMSSD decreased by 94.1%, HF by 65.3%, and LF by 76.4%, whereas HR increased by 134%. During early recovery, HR decreased (β = −39.2 bpm, p = 0.018), and RMSSD and SDNN showed partial rebound (+4.97% and +57.2%, respectively). VO 2 max moderated autonomic adjustments from onset to peak, influencing SDNN (β = 1.41, p = 0.030), LF (β = −35.2, p = 0.001), and total power (β = −18.7, p = 0.010). VO 2 max was strongly associated with RMSSD recovery slopes in unadjusted (β = 0.144, p = 0.019; R 2 = 0.299) and adjusted models (β = 0.168, p = 0.036). Spectral indices showed marked instability, despite recovery in HR and RMSSD. HF decreased an additional 172% and LF decreased an additional 18% from peak to recovery. Conclusion: Wavelet-derived time-domain indices showed coherent autonomic transitions, with strong vagal withdrawal at onset and partial reactivation during early recovery. Higher VO 2 max was linked to reduced vagal withdrawal at peak and stronger RMSSD recovery, indicating greater autonomic resilience. In contrast, wavelet spectral indices were phase-dependent and failed to track recovery, likely due to rapid, irregular shifts in breathing rate and cardiac dynamics. These results indicate that, without respiration-aligned HF and LF bands, frequency-domain HRV indices may not reliably characterize autonomic changes during peak exercise and early recovery. Funding: No external funding was received. This abstract was presented at the American Physiology Summit 2026 and is only available in HTML format. There is no downloadable file or PDF version. The Physiology editorial board was not involved in the peer review process.