A broken power-law model of heart rate variability spectra in sleep

AimsThe aim of the study was to introduce a parametric description of RR-interval spectra using a broken power-law model, in addition to the parametrization of oscillatory peaks. Furthermore, to use this model to evaluate effects of age, sex and sleep architecture on overnight heart rate variability (HRV) in healthy subjects. Methods & ResultsFrom a polysomnography database, 215 whole-night, high quality electrocardiograms (ECGs) were extracted. The fractal and oscillatory power-spectral densities (PSDs) were calculated from evenly resampled RR-interval time-series, then a broken power-law model was fitted using piecewise linear regression to the double-logarithmic PSD, determining a custom breaking point in the fractal component, and allowing for two independent spectral slopes in the lower and higher frequency domains. The two-slope model provided a more optimal description compared to linear regression in all cases, even when penalizing increased model complexity. Peak detection was applied to the oscillatory component in the LF (0.04-0.15 Hz) and HF (0.15-0.4 Hz) bands, extracting the frequency and prominence of the dominant peak from each. The high frequency domain intercept, the breaking point frequency and the LF peak frequency decreased significantly with age. Both slopes were flatter in females, while the high domain intercept and the HF peak prominence was significantly increased. Waking after sleep onset and lightest sleep (N1) were associated with lower intercept values, while REM sleep had an opposite effect. ConclusionsThe broken power-law model proved to be more appropriate for the description of RR-interval spectra than the single-slope model, and also captured effects of age, sex and sleep structure that were corroborated by the literature. We would like to highlight that while HRV changes are often assumed to be of oscillatory origin, the fractal component has a major contribution to the total PSD, and thus to all measures derived from it.