This research is a part of DYNAMIC-study which is an on-going study focusing on non-invasive recording of hemodynamics from subjects with and without cardiovascular diseases (Clinicaltrialsregister.eu 2006-002065-39; Clinicaltrials.gov NCT01742702). The study has been approved by the ethics committee of the Tampere University Hospital and all study subjects gave informed consent. Volunteers for the research were recruited by announcements delivered in public organizations including Tampere University Hospital, Tampere University, Varala sports institute and occupational health care organizations. All study subjects were interviewed and examined by a physician who documented medical history and lifestyle habits. Smoking habits were determined as never smokers, previous smokers and current smokers and smoking in pack years was registered. Alcohol consumption was documented as weekly amounts of standard drinks (i.e. ~12 g alcohol). Physical activity was determined as number of exercise sessions (lasting at least 30 minutes) per week with at least moderate level of work load (for example brisk walking or jogging).
From the present study subjects with a history or cardiovascular disease, diabetes mellitus, kidney disease, heart rhythm other than sinus, alcohol or substance abuse, concurrent malignancy, or medications with direct effects on the cardiovascular system were excluded. Also subjects with incomplete hemodynamic recordings (i.e. missing HR or HRV values) were excluded. Altogether 569 subjects (mean age 44.9, 95% confidence intervals of the mean (CI) (43.9, 45.9) years; 287 men) were included. The following stable medical conditions with adequate medications were included: asthma (medicated with inhaled corticosteroids, n=12), allergy (n=7), depression (n=29), dyslipidemia (n=13), dyspepsia (n=10), epilepsy (n=3), hypothyroidism (n=16), and rheumatoid arthritis or lupus (n=3). In total 74 females (26%) were on low dose hormone therapy, i.e. intrauterine device (44 subjects) or peroral combination therapy (30 subjects).
Blood sampling was conducted after overnight fasting. Concentrations of plasma sodium, potassium, glucose, creatinine, cystatin C, C-reactive protein, triglyceride, and total, high-density (HDL) and low-density lipoprotein (LDL) cholesterol were determined by Cobas Integra 700/800 (F. Hoffmann-LaRoche Ltd, Basel, Switzerland), or Cobas 6000, module c501 (Roche Diagnostics, Basel, Switzerland). Leukocyte count and hematocrit were determined by ADVIA 120 or 2120 analyzers (Bayer Health Care, Tarrytown, NY, USA). Estimated glomerular filtration rate was calculated using CKD-EPI creatinine-cystatin C equation .
Trained research nurse carried out hemodynamic measurements in a quiet and temperature-controlled laboratory. The study subjects were guided to avoid caffeine products, smoking or heavy meals 4 hours prior to the investigation. Before proper measurements the subjects were resting supine about 10 minutes, during which period a tonometric sensor for pulse wave analysis was placed on left radial pulsation, brachial blood pressure cuff on right upper arm for blood pressure calibration, and electrodes of whole body impedance cardiography were placed on the body surface. The following hemodynamic variables were captured in a beat-to-beat fashion during 5 min periods in supine and passive upright position: HR (1/min), pulse wave velocity (PWV, m/s), systemic vascular resistance index (SVRI, systemic vascular resistance/body surface area, dyn*s/cm5/m2), cardiac index (cardiac output/body surface area, l/min/m2) and radial blood pressure (mmHg).
For the whole body impedance cardiography measurement the CircMonR-device (JR Medical Ltd., Tallinn, Estonia) and for pulse wave analysis the SphygmoCor pulse wave monitoring system (SpygmoCor PWMx, Atcor medical, Australia) was used. The measurement protocol is described in detail in our previous publications [22, 23] and it has been shown to be repeatable and reproducible [24, 25]. In addition, PWV values measured using impedance cardiography show good correlation with the tonometric method (r=0.82, bias 0.02 m/s, 95% CI 0.21 to 0.25) , cardiac output values measured using whole body impedance cardiography are congruent with the thermodilution method (bias 0.00 l/min, 95% CI -0.26 to 0.26) and the direct oxygen Fick method (bias -0.32 l/min, 95% CI -0.69 to 0.05) , and the correlation between stroke volume recordings using impedance cardiography and 3D-echocardiography has been shown to be good (r=0.781, bias 4.1 ml, 95% CI 2.2 to 10.4) .
Heart rate variability analysis
Beat-to-beat RR-intervals were captured from electrocardiogram recordings by the impedance cardiography device at 200 Hz sampling rate. Then HRV results were analyzed separately from both RR-intervals and HR by the use of Matlab-software (Natick, Massachusetts, USA). First, normal RR-intervals were recognized and an interval was considered ectopic or an artifact if it differed by more than 20 % from the previous ones. The artifact RR-intervals were processed by the cubic spline interpolation method .
The following HRV parameters were calculated from the measurements of 5 minutes in the supine and 5 minutes in the upright position: (1) time domain HRV parameters: mean of normal to normal RR (NN)-intervals, SDNN (standard deviation of NN-intervals) and RMSSD (square root of mean squared differences of NN-intervals), and (2) frequency domain parameters by the fast Fourier transformation method [3, 29]: power in low frequency range (LF, 0.04-0.15 Hz), power in high frequency range (HF, 0.15-0.40 Hz), LF to HR ratio (LF/HF) and total power.
As described above, the magnitude of HRV depends on the prevailing level of HR. Hence, we calculated the HRV parameters also based on the HR data in addition to the RR-interval data using mean of 60/RR-interval (i.e. HR) instead of pure RR-interval values in the calculations. Furthermore, to weaken the dependence of HRV on HR, the HRV power spectra (from RR-interval data) were divided by average RR-interval (avRRI) to the fourth power and these HRV values, total power, HF power, LF power and LF/HF (TP/avRRI4, HF/avRRI4, LF/avRRI4, LFHF/avRRI4, respectively) were included in the analyses .
IBM SPSS Statistics software (version 25, Armonk, New York, USA) was used for statistical analyses and p-values <0.05 were considered significant. Continuous data was reported as means and 95% CI if normally distributed, and as medians and interquartile range if asymmetrically distributed. For statistical analyses the study subjects were divided into three tertiles according to the mean resting HR, determined from the mean of the last 3 minutes during the 5 minutes of supine measurement. Because HR and HRV are significantly dependent on sex [4, 30], the tertiles were determined separately for men and women but for statistical power men and women were analyzed together.
Differences in study population demographics, hemodynamic measurements and laboratory values between the HR tertile groups were examined using analysis of variances (ANOVA). Tukeys HSD post hoc test was performed for homogeneous and Tamhane’s T2 post hoc test for nonhomogeneous variables. Natural logarithms of HRV parameters, C-reactive protein and fasting triglyceride concentrations were used in the statistical analyses to normalize their skewed distributions. The proportions of smoking habits (never, previous or current smoker) were compared using the χ2 test.
The association of continuously measured HR with HRV during supine and upright positions was investigated using linear regression analysis (with natural logarithms of nonlinear HRV parameters). The other explanatory factors for linear regression analysis were selected as based on the comparisons between the tertile groups (ANOVA). We previously reported that blood pressure, cardiac index and SVRI are significantly related with the resting level of HR (in both supine and upright positions) . Therefore, mean blood pressure from the hemodynamic variables was selected to the regression analyses rather than cardiac index and SVRI together. Linear regression model for supine and upright HR was calculated using RR-interval -related LF/HF ratio and then the analyses were repeated using LF/HR ratio divided by average RR-interval to the fourth power (to weaken the HR dependence) .