Optimizing Intersession Reliability of Heart Rate Variability-The Effects of Artifact Correction and Breathing Type.

Gisselman, AS, D'Amico, M, and Smoliga, JM. Optimizing intersession reliability of heart rate variability-the effects of artifact correction and breathing type. J Strength Cond Res 34(11): 3199-3207, 2020-This study aimed to comprehensively investigate the reliability of multiple heart rate variability (HRV) parameters and to explore the influence of artifact removal and breathing condition on HRV reliability. Resting HRV was collected using Polar Team monitors on 41 participants (age: 19.9 ± 1.2 years; 28 females and 13 males) during 2 separate days. Within each session, participants performed 10 minutes each of spontaneous and controlled breathing (randomized order). Kubios HRV analysis software was used to analyze 180-second data epochs using "low" or "strong" artifact removal. Relative reliability was assessed using intraclass correlation coefficient (ICC2,1), and absolute reliability was quantified using mean-normalized SEM%. Time-domain and nonlinear parameters produced poor to good intersession agreement (ICC = 0.34-0.68; SEM% = 11.0-39.0) with "low" artifact removal, regardless of breathing condition. Frequency-domain parameters demonstrated fair intersession agreement during controlled breathing (ICC = 0.40-0.45; SEM% = 26.0-70.0) but poor agreement during spontaneous breathing (ICC = 0.07-0.13; SEM% = 32.0-81.0). Minimal differences in ICCs were observed between "low" and "strong" artifact removal. Thus, this study provides 3 practical applications: (a) HRV monitoring is most reliable when using time-domain and nonlinear parameters, regardless of breathing or filtering condition, but no single parameter is especially reliable. The large variation and poor intersession reliability of frequency indices during spontaneous breathing are improved by controlling breathing rate; (b) "Low" artifact removal seems superior to more aggressive artifact removal; and (c) intersession differences in HRV measurements <30% may be indicative of normal daily variation rather than true physiologic changes.