Background: Cardiac autonomic dysfunction (cAD), representing an imbalance between increased sympathetic and diminished parasympathetic activity, is frequently found in individuals with heart failure (HF), associated with exercise intolerance and a worse prognosis. Heart rate variability (HRV) is one established tool to assess cAD. Previous smaller studies in athletes and smaller studies in patients with coronary artery disease suggest an association between HRV and measurements of cardiopulmonary exercise testing (CPET). However, the relationship of cAD and cardiopulmonary capacity in patients with HF is not well understood.

Methods: Data from the MyoVasc study (N=3,289), a prospective cohort study of chronic HF (NCT04064450), were analyzed. Participants received a five-hour, highly standardized examination at a dedicated study center including CPET on a cycle ergometer according to an adapted WHO protocol. Data on HRV was derived from 24 hour Holter ECG recordings by extraction of RR intervals and subsequent processing via an established analysis pipeline. Based upon prior analyses identifying top predictors of HF outcome and survival out of a set of more than 70 HRV markers using regularized regression, the following HRV markers were selected for investigation in the present study: rMSSD, LF/HF IQR (wavelet), and the non-linear measures SD1/SD2 (Poincaré plot index), acceleratory capacity (AC) and deceleration capacity (DC). Multivariable regression analyses were calculated to investigate the impact of HRV markers on CPET measures and to evaluate potential differences in HF phenotypes.

Results: The analysis sample included N=626 subjects (mean age 64.3±10.4 years; female sex: 31%) with available data on both HRV and CPET. In random forest analyses, percent predicted peak Vo2 was identified as strongest predictor of worsening of heart failure and VE/Vco2 slope as strongest predictor of cardiac death, respectively. In linear regression analysis with adjustment for age and sex, LF/HF IQR (wavelet) (ß-estimate: 0.23 (95% confidence interval (95%CI) -0.15; 0.32), P<0.0001), log(SD1/SD2) (ß-estimate: -0.21 (95%CI -0.28; -0.13), P<0.0001), AC (ß-estimate: -0.22 (95%CI -0.30; -0.14), P<0.0001) and DC (ß-estimate: 0.24(95%CI 0.16; 0.33), P<0.0001) were associated with percent predicted peak VO2. In addition, AC (ß-estimate: 0.11 [95%CI 0.04; -0.19]; P=0.004) and DC (ß-estimate: -0.14 [95%CI - 0.22; -0.06]; P=0.00087) were associated with VE/Vco2 slope. Subsequent analysis in HF groups revealed phenotype-specific differences: in subjects with HF with (mildly) reduced ejection fraction, only log(SD1/SD2) was associated with percent predicted peak VO2 (ß-estimate: -0.13 [95%CI -0.26;0.00]; P=0.046). In contrast, AC (ß-estimate: -0.22 [95%CI -0. 36; -0.08]; P=0.002) and DC (ß-estimate: 0.30 [95%CI 0.15; 0.46]; P=0.0002) were associated with percent predicted peak VO2 in subjects with HF and preserved ejection fraction (HFpEF). In addition, DC reflecting diminished parasympathetic activity was also inversely related to VE/Vco2 slope (ß-estimate: -0.15 [95%CI - 0.30; - 0.01]; P=0.043) in HFpEF subjects.