Assessment of cardiac vagal control in daily life: approaches to control for respiratory effects on respiratory sinus arrhythmia

Respiratory sinus arrhythmia (RSA) is the natural lengthening of heart period during inhalation and shortening during exhalation. A number of pharmacological blockade studies showed that this heart period variability linked to respiration was largely eliminated by cardiac vagal blockade but not by sympathetic blockade. Consequently, the prospect of non-invasively monitoring cardiac parasympathetic (vagal) control via the calculation of parameters such as peak-valley RSA and high-frequency heart rate variability (HF-HRV) has become attractive in psychophysiology research (Quigley et al., 2024). This way, changes in parasympathetic control on the heart in response to changing behavioral and cognitive demands, as well as psychological states, could in principle be studied. However, changes in RSA parameters are substantially affected by changes in respiratory pattern even in the absence of cardiac vagal control changes. Previous literature, some of which utilized autonomic blockade, showed up to 60% of variation in RSA could be attributed to changes in tidal volume and respiration rate. Despite the evidence, studies using peak-valley RSA, HF-HRV, or RMSSD to try and index cardiac vagal control have continued to disregard the effects of respiration on heart rate variability. When conducting laboratory research, this problem may be attenuated by a relatively mild variation in respiratory behavior (although not always). Increasingly, however, research is moved outside the lab to study physiological reactivity to events and thoughts that are more meaningful to the individual in settings with more ecological validity, using daily life (ambulatory) monitoring through wearables. Due to physical activity, the within-person variations in respiration and RSA tend to be larger in daily life as compared to that in a controlled laboratory setting (Grossman et al., 2004). Even during static moments, with no or light physical activity only, variation in respiratory behavior is likely to be more pronounced in ambulatory recordings. Accordingly, ambulatory research using RSA parameters (e.g., peak-valley RSA, RMSSD, HF-HRV), even when correcting for movement, may run the risk of detecting relationships of psychological states with respiration rather than with cardiac vagal control. So far, research on the relationship between RSA reactivity and psychological states in daily life has been scarce (Schmid et al., 2024), and rarely the effects of respiration on RSA are controlled for. Good approaches to control for respiration might help improve the association of RSA parameters with momentary psychological state, improving our understanding of biopsychological dynamics. As proposed early on (Grossman et al., 2004) and recently adopted in the publication guidelines for human heart rate and heart rate variability studies in psychophysiology (Quigley et al., 2024) the assessment of cardiac vagal control using ambulatory RSA recording would be greatly helped by performing some form of respiratory control. However, this advice is hard to put in practice as long as no clear consensus exist on the best way to do this. To address this knowledge gap, we compare between different approaches (Approaches 0-4 explained below) that aim to control the effects of respiration on RSA in order to obtain a parameter that best captures changes in cardiac vagal control in daily life settings. Because our focus of interest is in changes in cardiac vagal control that are related to changes in psychological state, we restrict our comparison to daily recordings during the awake time where there is no or light physical activity only. As the criterion for changes in cardiac vagal control, we use the within-person minute-to-minute changes in heart period (as recorded by electrocardiography). A number of studies to date suggested that in the absence of moderate-to-vigorous physical activity, the variations in heart period are largely linked to changes in cardiac vagal control (see Grossman et al., 2004; Maciel et al., 1986). Although the relation is imperfect due to an admixture of changes in cardiac sympathetic control, it would still be expected that the approach best controlling for the effects of respiration on RSA would also better predict heart period variations, reflecting more variation captured in cardiac vagal control. We will therefore test which of the different approaches best predicts minute-to-minute heart period variations. Likewise, in view of the known vagal effects induced by psychological states, it is expected that approaches that control for the effects of respiration on RSA better predict variation in psychological states. To test this, we will compare which of the different approaches best explains the variance in self-reported momentary psychological states - namely, perceived stress, positive affect, negative affect, perceived safety as assessed with ecological momentary assessment in daily life. Different approaches to control RSA for respiratory behavior (see Statistical Analyses for the model equations) In Approach 0, outcome variables (e.g., minute-to-minute heart period and hourly smartphone-based self-reports of perceived stress) are predicted from minute-to-minute or the 5-minute pre prompt RSA values without controlling for respiratory behavior. This approach represents what has been the common practice in the field of psychophysiology to date. To implement Approach 1, prior to the start of an experimental recording (i.e., ambulatory assessment for the current study) participants need to undergo paced breathing trials covering a pace range from approximately 8 beats per min to 17 bpm. Using the impedance-based respiration and electrocardiography heart period data collected during the paced breathing trials across a range of respiratory frequencies, a regression line, RSA/Vt versus Ttot, is calculated per participant where Vt is tidal volume of respiration and Ttot is breathing period. Each data point on the regression represents a breath from the paced breathing calibration. This approach is based on the observation that when levels of cardiac vagal control change, this regression line shifts across the y-axis but its slope remains mostly stable (for more details on this approach see Ritz & Dahme, 2006). After getting the regression line for a participant, the respiration-controlled RSA in the epochs of interest are calculated as the deviation of the present RSA/Vt from what would be suggested by the baseline regression line for the present Ttot (X-axis location). Thus, the respiratory controlled RSA for a given epoch is expressed as ΔRSA/Vt (the shift on the Y-axis) with which we would predict the outcome variables of interest (e.g., minute heart period, hourly perceived momentary stress). As prior observations showed that the slope and the y-intercept of RSA/Vt versus Ttot regression vary substantially between-individuals, it is assumed it is crucial to have all participants undergo the paced breathing calibration rather than doing corrections at the group level. In Approach 2, we will emulate Approach 1 but calculate a regression line of RSA/VT vs Ttot per individual using all observed ambulatory minute RSA, Vt, and Ttot values but not using the experimentally induced paced breathing data. We will use the minute data from daytime and evening time, and exclude periods of sleep and physical activity. We will thus also see the feasibility of making this regression line without using a paced breathing procedure but rather sticking with each individual’s recording during spontaneous breathing. Again, the respiratory controlled ΔRSA/Vt for each minute specifically can then be calculated. In Approach 3, tidal volume and breathing period (inverse of respiratory rate) are simply added alongside RSA in the regression model as within-individual variables with random effects. As Approach 3 does not require a separate calibration stage, this approach is time-efficient and relatively more straightforward. We also include an Approach 4 that has been used in some previous research to control for the effects of respiration on RSA in ambulatory research (Grossman et al., 2004). In this approach, the mean RSA within a given epoch is divided by the mean tidal volume during the given epoch. Thus, the respiratory controlled RSA is expressed as RSA/Vt. No additional correction for respiratory rate is conducted as it is strongly (inversely) corelated to Vt. This approach is also time-efficient and does not rely on calibration (i.e., paced breathing). In the current study, participants wear a device recording the electrocardiograms (ECG), thorax impedance (dZ), and movement via triaxial upper-sternum positioned accelerometer and gyroscopes) for 2 days. Heart period is extracted from the ECG, respiration rate and depth from the dZ and peak-valley RSA from the combined ECG and dZ signals. Participants answer hourly ecological momentary assessment prompts on their momentarily perceived stress, negative affect, positive affect, and perceived safety levels. We extract the data during the awake periods of the two days, when the participants are not engaged in physical activity. In our within-subject design, level 2 constitutes an individual (and covariates, e.g., age, sex, physical activity) and level 1 an individual’s minute-by-minute values (RSA, RR, VT, HP) or 5-minute pre prompt RSA values and hourly prompted self-report values (perceived stress, negative affect, positive affect, and perceived safety levels at the moment). The remainder of the preregistration further explains the study including the variable operationalizations and the planned analytical strategy. Grossman, P., Wilhelm, F. H., & Spoerle, M. (2004). Respiratory sinus arrhythmia, cardiac vagal control, and daily activity. American journal of physiology. Heart and circulatory physiology, 287(2), H728–H734. https://doi.org/10.1152/ajpheart.00825.2003 Maciel, B.