Heart Rate Variability Computational Benchmarks And Associations With Hot Flash Events Under Free-living Conditions

Abstract

PURPOSE: The primary purpose was to compare benchmark times for deriving real-time heart rate variability (HRV) parameters from raw electrocardiogram (ECG) data using two methods (BioSPPy+pyHRV vs custom algorithm). The secondary purpose was to investigate the association between HRV and hot flash events (HF) among menopausal women during free-living conditions. METHODS: Fifteen women with daily menopausal HF (age (mean ± SD): 54.4 ± 4.3 years, BMI: 28.7 ± 4.4 kg/m2) completed up to three 48-hour free-living ambulatory monitoring sessions while wearing a montage of devices to measure the electrophysiological data. Participants also used an event marking button to log the time of subjective HFs and were asked to maintain their habitual daily routine. Physiological data streams were expertly scored to identify objective HF. Raw 1024 Hz ECG data were processed using existing python libraries (BioSPPY+pyHRV) and a custom algorithm adapted from Hamilton et al 2002 to derive N-N intervals and HRV time and frequency domain parameters. Differences in log-transformed HRV parameters between physionormal and HF (objective) were investigated using mixed effects linear regression. RESULTS: On average, women experienced 9 ± 6 HF/day, resulting in 16716 data windows for analysis (8319 physionormal and 8397 HF). Time to compute HRV parameters from raw ECG data was 51.0% ± 8.9% faster using the custom algorithm compared to BioSPPY+pyHRV. When the custom algorithm was used to extract N-N intervals from raw ECG data and pyHRV for HRV parameters, the full custom algorithm remained 2.3% ± 1.5% faster. HF were associated with significant reductions in HRV parameters related to N-N interval variability and timing between successive N-N intervals (p < 0.05). In addition, there were significant reductions in the high-frequency power and ratio of low-frequency to high-frequency power during HF (p < 0.05). However, there were no significant differences in the low-frequency power component (p > 0.05). CONCLUSION: Disruptions in autonomic cardiac control were observed in the presence of HF via reductions in parasympathetic control. The custom algorithm may provide a computationally efficient method for providing HRV parameters in real-time. Supported by Massachusetts Life Sciences Center and Embr Labs