Long-Term Stability of Respiratory Sinus Arrhythmia among Adults with and without a History of Depression

Abstract

Respiratory sinus arrhythmia (RSA) is an index of parasympathetic nervous system activity reflecting respiratory influences on heart rate. This influence is typically measured as high frequency heart rate variability (HF-HRV) or root mean square of successive differences (RMSSD) of adjacent inter-beat intervals. The long-term stability of the RSA measure is important it has been conceptualized as a marker of individual differences; more specifically of an individual’s autonomic regulation and affect-related processes, including emotion regulation. At present, it is now known if resting RSA levels reflect stable differences over a long-term period (i.e., >1 year). Even less is known about how RSA stability differs as a function of depression history and whether it relates to depression risk trajectories. In the present study, we examined the 1.5-year test-retest reliability of resting RSA using the intraclass correlation coefficient (ICC) in 82 adults: n = 41 with a history of depression (ever-depressed); n = 41 controls with no depression history (never-depressed). HF-HRV was fairly stable in both groups (Ever-Depressed ICC = .55, Never-Depressed ICC = .54). RMSSD was also fairly stable in ever-depressed adults (ICC = .57) and never-depressed controls (ICC = .40). ICC values for both indices did not differ between groups per overlapping 95% confidence intervals. Therefore, RSA stability as assessed by both frequency (HF-HRV) and time domain (RMSSD) measures was not attenuated by a depression history. Implications and the need for future research are discussed.

Respiratory sinus arrhythmia (RSA) is an index of parasympathetic nervous system activity that is modulated by the vagus nerve and reflects heart rate variability in synchrony with respiration (Berntson, 1997; Porges, 2007; Thayer & Lane, 2009). It is commonly assessed via frequency domain (high frequency heart rate variability, HF-HRV) or time domain (Root Mean Square of Successive Differences in R-R intervals, RMSSD) measures. Both measures have been examined as a risk factor, outcome, or correlate of the experience and regulation of affect and mood disorders (e.g., Balzarotti et al., 2017; Beauchaine, 2015; Task Force, 1996).

According to polyvagal theory (Porges, 2007), RSA levels reflect the degree to which a person is successful in orienting psychophysiological adjustment to environmental demands, such that higher levels of RSA reflect greater cardiac vagal control. At rest, cardiac vagal control (i.e., efferent vagal input to the sino-atrial node of the heart) serves to modulate heart rate, typically with vagal control attenuating sympathetic control, thereby slowing heart rate and increasing levels of RSA. In the absence of internal or environmental demands, higher RSA, as measured under resting conditions (hereafter referred to as resting RSA), has been found to reflect more control over attention, emotion, and behavior, a notion which has received empirical support in cross-sectional and longitudinal studies (e.g., Demaree et al., 2004; Gyurak & Ayduk, 2008; Ong et al., 2020). Indeed, resting RSA has been considered a trait characteristic because it reflects individual differences in tonic vagal control (see Balzarotti et al., 2017; Berntson et al., 1997; Bertsch et al., 2012). Reasonable test-retest reliability of this vagal index, however, is a prerequisite of its use as a trait measure.

There is increasing evidence for RSA as an index of emotion regulation capacity in depression (Koch et al., 2019; Koenig et al., 2016). Additionally, longitudinal research has demonstrated that resting RSA is predictive of future depression symptoms over a 1-2 year period (e.g., Kovacs et al., 2016; Yaptangco et al., 2015) and is a marker of treatment response (Chambers & Allen, 2002; Hartmann et al., 2019). While some find reduced resting RSA in depression (Koch et al., 2019), results are heterogenous (see Bylsma et al., 2021). Importantly, the extent to which measures of RSA have adequate test-retest reliability in clinical samples still remains to be characterized. Indeed, there is reason to suspect a conceptualization of RSA as trait-like may be less accurate for those with a depression history, per evidence that resting levels vary as a function of symptomatology (Chambers & Allen, 2002; Hartmann et al., 2019). Since RSA is increasingly considered to be a biomarker of risk, illness course, and treatment response, especially as related to depression, there is a need to establish evidence for its long-term stability.

Indeed, the stability of resting RSA among emotionally healthy adults, primarily quantified using HF-HRV, has been previously demonstrated over time intervals ranging from days, weeks, and months. For example, studies including a range of N = 14-63 participants examining up to 3-week test-retest reliability reported fair-to-excellent degrees of stability: intraclass correlation coefficient (ICC) values ranged from .48 to .90; Pearson’s correlation coefficient (r) values ranged from .69 to .73 (Bertsch et al., 2012; Byrne et al., 1991; Kleiger et al., 1991; Pinna et al., 2007). Similarly, evidence supports RSA as stable over one week when quantified using RMSSD: ICC values range from .71 to .76 (Bertsch et al., 2012; Gujit et al., 2007.

Similar findings with N = 41-103 participants have been reported for test-retest intervals of 1- to 2-months, with HF-HRV and RMSSD ICC’s ranging from .50 to .80 (Borges et al., 2018; Uhlig et al., 2020), and for up to 7-month retest intervals (HF-HRV ICC’s = .32 to .54; r’s = .85 to .98; RMSSD ICC: .23; Pitzalis et al., 1996; Sinnreich et al., 1998; Uhlig et al., 2020). In a study with an 11-month follow-up among 103 healthy college students, the ICC range for both HF-HRV and RMSSD of .55 to .65 suggest a fair-to-good degree of stability (Uhlig et al., 2020). In the only study of a one-year test-retest interval in adults, which included 32 healthy women, levels of resting RSA yielded a good degree of stability (HF-HRV r = .63) (Burleson et al., 2003). While there is evidence for RSA stability (using the peak-valley method; Galles et al., 2002) over a two-year interval among young children and adolescents (r = .49; El-Sheikh, 2005), we were unable to locate studies among adults which extended beyond a year. Thus, the current state of the literature suggests that RSA is a reproducible psychophysiological index across several months among adults without psychopathology.

The scant information about longer term stability (>1 year) is constrained by small sample sizes with less than 20 participants and selective inclusion of individuals, such as healthy adults (e.g., Burleson et al., 2003; Pitzalis et al., 1996) or college students (e.g., Bertsch et al., 2012; Borges et al., 2018). Also, it is unclear whether the presence or history of psychopathology may compromise stability of resting RSA as a trait index. For example, it is unclear if findings can be generalized to individuals with psychopathology such as depression, which may compromise RSA stability given the influence of depressive status on RSA (Beauchaine, 2015; Bylsma et al., 2014; Koch et al., 2019; Rottenberg et al., 2007). Thus, when examining measurement stability, it is important to consider a history of depressive psychopathology, as participants may differ in their diagnostic status depending on the time of assessment. If levels of resting RSA were found to be a reproducible index among adults with a depression history, future researchers could more confidently utilize this index to examine the psychophysiological processes which serve as risk factors, correlates, or outcomes of its diagnosis, which could be used to inform prevention and intervention efforts.

In light of the extant literature, the goal of the present study was to examine the longitudinal stability (>1 year) of resting RSA among adults with and without depression histories. We focused on HF-HRV and RMSSD as these are the most commonly used indices of RSA in studies focused on its association with psychopathology (see Balzarotti et al., 2017; Koch et al., 2019). Based on the available evidence, we expected RSA, as indexed by HF-HRV and RMSSD, to have fair-to-excellent reliability, i.e., ICC range: .40-.80 (see Cicchetti, 1994) across a 1.5-year window. Given the disruptions to autonomic functioning associated with depression and resultant fluctuations in RSA (Bylsma et al., 2014; Koch et al., 2019; Rottenberg et al., 2007), we also expected that its levels would be more stable across time among never-depressed controls than among those with a depression history.

Method

Participants and Procedures

Participants were 82 adults (M_age = 36.98 years, SD = 11.27, Range = 20 – 56; 68% female), and recruited as part of a larger sample for a longitudinal study of psychophysiology, emotion regulation, and decision-making (e.g., Westbrook et al., 2022) across the aging process (overall N = 368). Participants identified as White/Caucasian (67.1%), 19.5% as Black/African-American, 11% as Bi/Multi-racial, and 2.4% as “Other”. There were no differences between the ever- and never-depressed groups in age, sex, nor racial composition (Table 1).

Table 1.

Demographic, clinical, and physiological characteristics by group.

Variables	Ever-Depressed (n = 41)	Never-Depressed (n = 41)	χ2 or t statistic
Age at Time 1 (Mean years, SD)	38.38 (9.90)	35.58 (12.46)	−1.13
Female (n, %)	26 (63.40)	30 (73.17)	0.90
Race (White, n, %)	31 (75.61%)	24 (58.54%)	2.71
Race (Black, n, %)	10 (24.39%)	6 (14.63%)	1.24
Respiration Rate (Breaths per minute), Time 1	13.70 (4.72)	13.96 (4.26)	0.25
Respiration Rate (Breaths per minute), Time 2	14.43 (5.46)	13.67 (4.88)	−0.67
BDI-II (Mean, SD), T1	18.00 (11.42)	5.61 (6.22)	−6.11***
BDI-II (Mean, SD), T2	15.78 (10.97)	4.05 (3.52)	−6.45***
Any cardiovascular problem/cardioactive medication, T1 (n, %)	18 (43.90%)	5 (12.19%)	10.21**
Cardiovascular problem	11 (26.83%)	3 (7.32%)	5.51*
Cardioactive medication	12 (29.27%)	2 (4.87%)	8.61**
Any cardiovascular problem or cardioactive medication, T1 (n, %)	18 (43.90%)	4 (9.75%)	12.18***
Cardiovascular problem	10 (24.39%)	2 (4.87%)	6.03*
Cardioactive medication	14 (34.14%)	2 (4.87%)	11.18***
Time of Day for Physiological Assessment, T1	11:58am (124 min)	12:29pm (142 min)	−1.02
Time of Day for Physiological Assessment, T2	1:16pm (144 min)	1:07pm (184 min)	−0.26
Regular Smoker, T1 only (n, %)	11 (26.83%)	6 (14.63%)	1.19
HF-HRV, T1	5.69 (1.92)	6.21 (1.33)	1.51
HF-HRV, T2	5.14 (1.74)	6.27 (1.48)	3.15***
RSA ICC	.55	.54	N/A
RMSSD, T1	38.54 (21.54)	39.10 (31.98)	−.09
RMSSD, T2	43.07 (27.33)	30.53 (37.18)	1.74
RMSSD ICC	.57	.40	N/A

Note.

^*Bolded values are statistically significant. Regular smoker defined as smoking daily.

Participants were originally recruited from a variety of outpatient clinical settings, as described in detail previously (see Kovacs et al., 1984), for a longitudinal Program Project on risk factors for and outcomes of pediatric onset depression, during which they completed a comprehensive clinical assessment including a semi-structured psychiatric diagnostic interview (see Miller et al., 2002 for details). Studies following these and other participants, including siblings, controls, and their offspring have been published elsewhere (e.g., Kovacs et al., 2009; Miller et al., 2002; Yaroslavsky et al., 2014). Final psychiatric diagnoses were reached upon consensus among clinicians (Maziade et al., 1992). Exclusionary criteria included the presence of major systemic medical illness or intellectual disability. The initial project and present study were approved for ethical procedures by the University of Pittsburgh Institutional Review Board (IRB).

In the present study, a target sample size of N = 100 was recruited for follow-up (Time 2) to examine the long-term stability of a behavioral task of cognitive effort (see Westbrook et al., 2022). Participants were recruited from the parent study sample of N = 368. The sample size of N = 100 was determined by a multilevel power analysis (Bujang & Baharum, 2017), which ensured the present study was sufficiently powered to detect a poor-to-fair degree of test-retest reliability (ICC = .42 vs. null hypothesis ICC = .20) at 80% power and alpha at .05. Data collection for the parent study began in October 2017, and participants were only eligible for Time 2 if they completed their baseline assessment during or after December 2018. There were no further exclusion criteria. We began recruitment at this point to initially focus on securing participation and having available clinical staff for the baseline assessment in the parent study, which also included a comprehensive clinical interview (see Westbrook et al., 2022). The COVID-19 pandemic began in the same month (March 2020) of follow-up, and participation ended in September 2021. N = 95 participants were successfully recruited, of which 82 had usable psychophysiological data at both timepoints: n = 41 had a history of depression (ever-depressed) and n = 41 were never-depressed controls. There was a similar degree of data loss between both groups, χ² = .07, p = .80. The mean interval between Time 1 and 2 was 1.51 years (Median = 1.43, SD = 0.39, Range = 0.84 – 2.48 years). As a result of COVID-19 pandemic-related disruptions in data collection, there was variability in follow-up interval span. However, interval length did not differ between groups (p = .90).

Among ever-depressed cases, the mean age of onset of the first episode of major depression was 17.48 years (SD = 8.91). On average, ever-depressed participants had experienced four previous major depressive episodes (M = 3.90, SD = 3.64). At baseline (Time 1) in the present study, 44% were diagnosed as being in a current episode of major depression using the Structured Clinical Interview for DSM-IV Disorders (SCID; First et al., 1995). At follow-up (Time 2), 39% scored ≥20 on the Beck Depression Inventory-II (BDI-II; Beck et al., 1996), indicating the likely presence of a major depressive episode. Using the BDI cut-off score as a proxy of the likelihood of being in a depressive episode at Time 2, 22% were currently depressed at Time 1 and 2, 17% transitioned from no episode to current episode, and 22% transitioned from a current episode to no episode.

Psychophysiological Procedures

Cardiac physiology was continuously monitored throughout an experimental protocol involving emotion regulation and decision-making tasks (see Westbrook et al., 2022). At Time 1, participants completed a clinical assessment, including an interview and self-rated questionnaires assessing levels of depression symptoms over the past two weeks using the BDI-II. They also reported on the diagnosis of a cardiovascular health problem (e.g., blood pressure), current use of cardioactive medication, including psychotropic (e.g., selective serotonin reuptake inhibitors) and non-psychotropic medication (e.g., blood pressure or thyroid medication), and smoking regularity (i.e., daily smoker or not). Then, participants were brought to the testing room and connected to the physiological recording devices. Subjects sat upright in a chair facing a computer screen for the entire protocol. After habituating to the room, participants were asked to breathe normally for three minutes to assess resting RSA.

All physiological signals were continuously sampled at 1,000 Hz using a MindWare BioNex system 3711-02 and BioLab software Version 3.3.0 (MindWare Technologies Ltd., Gahanna, OH). Electrocardiography (ECG) was measured with disposable, pregelled stress-testing Ag/AgCl electrodes (ConMed Andover Medical, Haverhill, MA). Following guidelines (Berntson et al., 1997), a modified Lead II configuration was used for ECG; electrodes were placed on the torso. Respiration rate was assessed via the Ambu Sleepmate Piezo respiration transducer placed around the subject’s abdomen. At Time 2, the above noted psychophysiological procedures were repeated, along with an abbreviated clinical assessment.

ECG data were inspected, and artifacts were corrected manually by trained raters (n = 4) using MindWare HRV 3.0.21 software (MindWare Technologies Ltd.). Inter-rater reliability was strong, r = .95 [95% CI: .77, .99]. If 5% or more of ECG signals in an epoch (3 minutes) were influenced by artifacts, that epoch was excluded from analyses. Data were processed as a time series of interbeat intervals (IBI) of adjacent R peaks extracted from the ECG for the 3-minute baseline period. R-wave markers in the ECG signal were processed with the MAD/MED artifact detection algorithm (Berntson et al., 2007). The IBI series data were linearly detrended, mean centered and tapered using a Hanning Window. We computed both HF-HRV and RMSSD as measures of RSA based on published guidelines (Berntson et al., 1997; Jennings, et al., 1981). Spectral-power values were derived using Fast Fourier transformations (quantified in ms²/Hz). HF-HRV was defined as the natural logged (ln) spectral-power value in the high frequency band (0.12–0.40 Hz). RMSSD was computed by Mindware as the root-mean square of successive R-R interval deviations and log-transformed, consistent with prior research (Bertsch et al., 2012). The Mindware HRV software was also used to calculate respiration rate from the respiratory transducer signal.

Data Analysis Plan

First, we computed descriptive statistics (Table 1). Next, we computed correlations between HF-HRV and RMSSD, separately, for each group within each timepoint (e.g., HF-HRV and RMSSD at Time 1 among the ever-depressed). Next, to test for group differences in HF-HRV, RMSSD, and respiratory rate, we conducted multivariate Repeated Measures Analysis of Covariance (RM-ANCOVA). To control for the effect of variables known to affect these parameters, covariates included respiration rate, age at visit, biological sex, self-reported frequency of depression symptoms over the past two weeks, time-of-day, length of interval between test-retest, and smoking regularity. Of central interest to the present study, we evaluated the reliability of RSA using ICC, defined as HF-HRV (Table 2) and RMSSD (Table 3). The ICC value quantifies the stability of RSA by partitioning the between-person variance, or the proportion of total variance that can be attributed to individual differences, from the within-person variance, or remaining variability which can be attributed to random error (Weir, 2005). Higher scores (range: 0-1) reflect increased reproducibility among participants. Guidelines for interpretation of ICC values were determined as follows: poor (< .40), fair (.40 to .59), good (.60 to .74), and excellent (.75 to 1.00) (Cicchetti, 1994).

Table 2.

Regression models examining RSA stability (HF-HRV) via the ICC in both groups.

Variables	Ever-Depressed (n = 41)		Never-Depressed (n = 41)
	B	SE	B	SE
Intercept	10.52***	1.29	9.38***	0.83
Age at Visit	−0.09***	0.02	−0.05**	0.01
Sex	−0.47	0.48	0.44	0.34
Respiration Rate	−0.03	0.03	−0.08**	0.03
Depression symptoms (BDI-II)	−0.01	0.02	−0.00	0.03
Any cardiovascular problem/cardioactive medication	0.29	0.40	−1.11*	0.49
Regular Smoker (Time 1 Only)	−0.34	0.52	−0.25	0.43
Time of Day for Physiological Assessment	−0.00	0.00	−0.00	0.00
Days between Time 1 and Time 2	−0.29	0.15	0.06	0.11

Note.

^*p < .05,

^**p < .01,

^***p < .001.

All covariates were time-varying except sex, follow-up interval, and being a regular smoker (defined as smoking daily). Sex coded as 0 = female, 1 = male. Any cardiovascular problem/cardioactive medication coded as 0 = no, 1 = yes. Regular smoker coded as 0 = no, 1 = yes.

Table 3.

Regression models examining RSA stability (RMSSD) via the ICC in both groups.

Variables	Ever-Depressed (n = 41)		Never-Depressed (n = 41)
	B	SE	B	SE
Intercept	5.23***	0.65	4.67***	0.41
Age at Visit	−0.04**	0.01	−0.02**	0.00
Sex	−0.29	0.24	0.14	0.16
Respiration Rate	−0.01	0.02	−0.02	0.01
Depression symptoms (BDI-II)	−0.00	0.00	0.01	0.01
Any cardiovascular problem/cardioactive medication	0.22	0.20	−0.57**	0.23
Regular Smoker (Time 1 Only)	−0.20	0.26	−0.23	0.20
Time of Day for Physiological Assessment	−0.00	0.00	−0.00	0.00
Days between Time 1 and Time 2	−0.14	0.07	0.08	0.06

Note.

^*p < .05,

^**p < .01,

^***p < .001.

All covariates were time-varying except sex, follow-up interval, and being a regular smoker (defined as smoking daily). Sex coded as 0 = female, 1 = male. Any cardiovascular problem/cardioactive medication coded as 0 = no, 1 = yes. Regular smoker coded as 0 = no, 1 = yes.

To derive separate ICC’s for diagnostic groups (ever-depressed, never-depressed), we tested two linear mixed effect models using the macro ICC9 in SAS Version 9.4 (Hertzmark & Spiegelman, 2010; SAS Institute, 2013). The same covariates from the RM-ANCOVA were used in the mixed effects models. To account for potential variance in RSA attributable to cardiac health problem and/or use of cardioactive medication, we created a binary variable to indicate if at least one of these issues were present. Doing so enabled us to reduce model multicollinearity and power per the low and restricted range (4.87-7.32%) of participants in the never-depressed group with potential cardiac effects attributable to these issues at either timepoint.

Results

We present descriptive statistics, ICCs, means and standard deviations for RSA per group in Table 1. Correlational analysis indicated strong relationships between each RSA measure within each timepoint (Ever-Depressed r’s = .93, Never-Depressed r’s = .92 to .93).

A multivariate Repeated Measures Analysis of Covariance (RM-ANCOVA) indicated between-group differences, F(1, 77) = 7.75, p = .007, such that never-depressed adults had higher levels of resting HF-HRV, F(1, 80) = 7.96, p = .006, and RMSSD, F(1, 80) = 6.47, p = .01. There was a non-significant effect for time (p = .13) but a significant group X time interaction, F(1, 79) = 4.35, p = .04).¹ In particular, groups did not differ on either index at Time 1 (p’s > .11). At Time 2, HF-HRV and RMSSD were lower in the ever-depressed group (p’s < .002). Respiratory values were routinely within the HF band for RSA. Further, an RM-ANCOVA found no between-group differences, effects of time, nor their interaction (p’s > .13)

The results from the separate linear mixed effect models for the ever- and never-depressed groups, including relevant covariates, are shown in Table 2 for HF-HRV and Table 3 for RMSSD. In the ever-depressed adults, ICC for RSA using HF-HRV at 1.5-year follow-up was .55, 95% CI = [.28, .73] and .54 among the never-depressed controls, 95% CI = [.31, .77]. ICC for RSA using RMSSD in the ever-depressed adults was .57, 95% CI [.32, .79], and .40 in never-depressed controls, 95% CI [.17, .68]. The overlapping 95% CI’s between groups indicate these values are not significantly different. In other words, after controlling for a host of related variables, having a history of depression was not found to attenuate the stability of resting RSA.

Discussion

RSA, measured using HF-HRV and RMSSD, is a commonly used index of cardiac psychophysiology, but there is scant evidence for its long-term stability (>1 year), particularly in clinical samples. Thus, the findings of the present study extend the literature on the test-retest reliability of resting HF-HRV and RMSSD, two indices of RSA: over an average of 1.5-year retest interval, we found a fair degree of stability of HF-HRV and RMSSD levels in two samples of participants with a depression history (HF-HRV ICC = .55, RMSSD ICC = .57) and without (HF-HRV ICC = .54, RMSSD ICC = .40). However, overlapping confidence intervals between groups for both HF-HRV and RMSSD indicate no significant differences. Consistent with prior research, there was a higher proportion of participants in the ever-depressed group who endorsed a cardiovascular problem and use of cardioactive medication (e.g., Joynt et al., 2003). Notably, equivalent levels of stability were found despite these differences, indicating that the reproducibility of RSA (as indexed by HF-HRV and RMSSD) is robust to group differences in cardiovascular health. Therefore, our findings indicate that a depression history does not attenuate the stability of RSA.

The coefficients found in the present study are in the range found among samples without psychopathology for shorter re-test intervals including up to 3 weeks (HF-HRV ICC range: .48 to .90; Bertsch et al., 2012; Byrne et al., 1991; Kleiger et al., 1991; Pinna et al., 2007) and two months (ICC range: .65 to .80; Borges et al., 2018; Sinnreich et al., 1998). Our findings are also consistent with those documented in studies up to seven months (HF-HRV ICC’s = .40 to .54; r’s = .85 to .98; RMSSD ICC’s: .32 to .49; Pitzalis et al., 1996; Sinnreich et al., 1998; Uhlig et al., 2020), eleven months in healthy students (ICC range for both HF-HRV and RMSSD of .55 to .65; Uhlig et al., 2020), and a year follow-up among a sample of healthy women (HF-HRV r = .63; Burleson et al., 2003). Thus, the present study provides support for interpreting resting RSA as a fairly stable index of cardiac psychophysiology when measured using HF-HRV and RMSSD, by examining its reproducibility over a longer period among a larger sample of adults with and without depression histories.

In spite of evidence for the detrimental associations found between depression and RSA (Koch et al., 2019), we did not find that a depression history attenuated RSA stability, as assessed by HF-HRV. Our findings of fair degrees of stability in the presence of psychopathology (ICC HF-HRV = .55; RMSSD = .57) are consistent with values reported among participants with personality and anxiety disorders: HF-HRV ICC’s = .51 to .85, RMSSD ICC’s = .50 to .80 (Eikeseth et al., 2020; Schmidt et al., 2012). However, in these studies, the follow-up interval was as short as one day and limited to up to one week. As these stability coefficients are similar to those reported in non-clinical samples, the extant literature suggests that psychopathology might not constrain the reproducibility of resting RSA, but longer-term research is needed to examine its stability across various types of psychopathologies.

Overall, our findings provide support for the stability of resting RSA when indexed as HF-HRV and RMSSD among adults with and without a history of depression. Therefore, the present study helps corroborate the utility of its interpretation as a physiological index of individual differences in autonomic nervous system functioning and marker of emotion (dys)regulation (e.g., Beauchaine, 2015; Porges, 2007). While the present study cannot account for the potential psychosocial or physical health influence of COVID-19 on RSA stability, we remind the readers that follow-up began the same month of the pandemic. Recent research has found RSA levels, when indexed both as HF-HRV and RMSSD, to differ between COVID-19 infected patients and controls; more, these differences appear to remain after recovery (e.g., Kaliyaperumal et al., 2021; Kurtoğlu et al., 2022). Therefore, it is possible that some variation in indices of stability indices may be attributable to infection prior to participation. Although the present study suggests that resting RSA is generally reproducible when measured in a laboratory setting, an important future area of research includes examining its reliability in daily life given the extensive use of ambulatory monitoring of cardiovascular psychophysiology as a risk factor and outcome of affect-related processes in healthy functioning and psychopathology (see Raugh et al., 2019).

Limitations

The present study is the first to examine the test-retest reliability of resting RSA among adults with and without a depression history (obtained in present and past studies via use of a structured clinical interview for diagnosis), does so using two popular indices (HF-HRV and RMSSD), and is also strengthened by a long-term test-retest window of 1.5 years. Still, there are a few limitations to note. First, while we did control for respiratory rate, we were not able to assess respiratory depth due to width and placement of the transducer belt (Egizio et al., 2011), which may be another important influence on test-retest reliability. Second, while the sample size in both groups (n = 41) was large relative to prior similar studies, it is still modest. For this reason, we did not derive separate indices of RSA stability for subgroups of ever-depressed adults (e.g., currently depressed vs. remitted at T1 or remitted at T2). Relatedly, the assessment of depression at T2 was limited to self-report (i.e., BDI-II). While our analysis controls for eight related demographic and psychophysiological variables (e.g., age, respiration rate, smoking regularity), we did not assess other potential influences (e.g., current stress levels, hours of sleep in prior night, diet, menstrual cycle, or use of hormonal contraceptives) which can potentially influence the reproducibility of RSA, especially when these variables change over time. Future work examining the impact of such factors on long-term stability is advisable. Also, we could not reliably account for all possible influences to retain statistical power. Finally, our sample was mainly female (68%), and we did not assess gender identity. Since research has found interactions between resting levels of RSA and sex in predicting several future distinct trajectories of psychopathology symptoms (Hinnant & El-Sheikh, 2013), and adults with minoritized gender identities experience disparities in cardiovascular health (Caceres et al., 2020), future studies with more balanced samples which consider sex and gender are needed.

Conclusion

RSA, as indexed by HF-HRV and RMSSD, is commonly examined as a psychophysiological index of emotion regulation associated with psychopathology risk, correlates, and outcomes. To address the limited research on the longitudinal stability of RSA over time frames longer than a year, and lack of known studies doing so with regard to depression, we examined levels of resting HF-HRV and RMSSD among adults with a history of depression and those with no history of depression over approximately 18 months. Overall, we found fair levels of stability in both groups for HF-HRV and RMSSD. Importantly, the presence of a depression history did not attenuate the long-term reproducibility of HF-HRV nor RMSSD. Findings suggest that researchers can consider HF-HRV and RMMSD as fairly stable indices of RSA that is appropriate for clinical longitudinal research. Given the interest in RSA as a transdiagnostic index of emotion regulation across types of psychopathology (e.g., anxiety, substance use disorders, personality pathology; Chalmers et al., 2014; Clamor et al., 2016), future longitudinal research should examine RSA stability across other disorders where emotion dysregulation is a core feature.

Impact Statement:

Researchers commonly utilize resting levels of parasympathetically-mediated heart rate variability (e.g., HF-HRV, RMSSD) to assess emotion regulation and other affective processes in healthy and clinical samples, particularly depression (Koch et al., 2019). However, little is known about its long-term stability, and even less so in the presence of psychopathology. Our findings indicate that resting HF-HRV and RMMSD are fairly stable indices of RSA over a period of 1.5 years, on average, among adults with and without a history of depression.