Changes in negative affect and changes in heart rate variability among low-income latinos with type 2 diabetes in a randomized, controlled stress management trial

Journal of Psychosomatic Research 124 (2019) 109774

Contents lists available at ScienceDirect

Journal of Psychosomatic Research journal homepage: www.elsevier.com/locate/jpsychores

Changes in negative affect and changes in heart rate variability among lowincome latinos with type 2 diabetes in a randomized, controlled stress management trial

T

⁎

Julie A. Wagnera, , Richard Feinnb, Rachel Lampertc, Angela Bermúdez-Millánd, Rafael Pérez-Escamillae a

University of Connecticut Schools of Medicine and Dental Medicine, United States of America Quinnipiac University School of Medicine, United States of America Yale University School of Medicine, United States of America d University of Connecticut School of Medicine, United States of America e Yale University School of Public Health, United States of America b c

A R T I C LE I N FO

A B S T R A C T

Keywords: Diabetes Latinos Heart rate variability Negative affect Observational

Background: Structural equation modeling examined the relationship between change in negative affect (NA) and change in heart rate variability (HRV) among 121 Latinos with type 2 diabetes. Methods: This study leveraged data from the Community Health Workers Assisting Latinos Manage Stress and Diabetes (CALMSD) study which compared diabetes education vs diabetes education plus stress management. Participants completed surveys of NA at baseline and again 8–10 weeks later. They also wore 7‑lead, 3-channel ambulatory ECG monitors for 24 h at both time points. The latent variable NA was modeled by observed scores on symptoms of depression, anxiety, diabetes distress, and wellbeing (reversed). The latent variable HRV was modeled by observed scores in the time domain (the standard deviation of the R-R interval [SDNN] and the root mean square of the successive differences [RMSSD]) and in the frequency domain, i.e., log-transformed ultra-low frequency, very-low frequency, low frequency, and high frequency. Results: At baseline, there were strong, negative cross-sectional associations between NA and HRV. Baseline NA predicted change in HRV, whereas baseline HRV did not predict change in NA. Controlling for fasting glucose and treatment assignment did not meaningfully alter the findings. Stress management improved NA but not HRV. At followup, a greater reduction (improvement) in NA was associated with a larger increase (improvement) in HRV, with a small-to-medium negative association that approached statistical significance. Conclusions: Findings indicate a longitudinal relationship between NA and HRV, and suggest that improvement in one may be associated with improvement in the other.

1. Introduction

involved in allostasis. Low HRV is associated with diabetes disease progression; it predicts onset of the metabolic syndrome [4], can be present at diabetes diagnosis [5], is an early finding in diabetic cardiac autonomic neuropathy [6], and predicts early diabetic mortality [7]. HRV tends to decrease over the lifespan [8] and rate of decline is steeper among people with diabetes. For example, the Atherosclerosis Risk in Communities (ARIC) study followed a population-based cohort aged 45–64 years at baseline and again 9 years later. Participants with diabetes had a more rapid temporal decrease in HRV than their counterparts without diabetes [9]. Whereas hyperglycemia is the hallmark of diabetes, glucose levels account for only a small amount of the variance in changes in HRV

The ability of physiological systems to adapt to environmental demands and achieve allostasis is associated with better health [1]. The autonomic system's ability to adapt in this way can be quantified by the beat-to-beat changes in heart rate, or heart rate variability (HRV). It is hypothesized that exposure to intense, frequent, or prolonged stress causes repeated attempts at allostasis. This cycling, as well as inefficient dampening or extinguishing of it, is referred to as “allostatic load” and may engender dysregulation of the autonomic system. Decrements in HRV is thus considered an indicator of allostatic load [2,3]. HRV is an integrative indicator in that it reflects numerous ‘upstream’ processes

⁎

Corresponding author at: University of Connecticut School of Dental Medicine, 263 Farmington Ave., MC3910, Farmington CT 06030, United States of America. E-mail address: [email protected] (J.A. Wagner).

https://doi.org/10.1016/j.jpsychores.2019.109774 Received 30 January 2019; Received in revised form 5 July 2019; Accepted 5 July 2019 0022-3999/ © 2019 Elsevier Inc. All rights reserved.

Journal of Psychosomatic Research 124 (2019) 109774

J.A. Wagner, et al.

is not the most common statistical approach to HRV data, investigators have used SEM to overcome the separate but related problems of highly correlated indicators and type 1 error mentioned above. For example, investigators have used SEM to combine frequency and time domains [30], to examine the underlying factor structure of allostatic load [31], the stability of HRV measures over time [32], and to examine affective, cognitive [33] and environmental correlates of HRV [34]. Following the precedent of these authors, we propose that latent variable modeling of HRV is not only acceptable but may be preferable in some cases. Its parsimony reduces the number of statistical tests, thereby reducing chances of type I error. Importantly, it also protects against potential investigator bias to report only those HRV indices that show favorable results. Despite its parsimony, it still allows, through factor loading coefficients, the examination of direction and magnitude of the contribution of the latent factor to each individual HRV indicator. Structural equation modeling is also more forgiving of missing data which can be problematic in longitudinal studies. The study reported here adds substantially to this literature by leveraging longitudinal data from the Community Health Workers Assisting Latinos Manage Stress and Diabetes (CALMSD) study. CALMSD was a randomized controlled trial in Latinos with type 2 diabetes that compared diabetes education only to diabetes education plus a behavioral stress management intervention to improve NA and associated outcomes. Before and after treatment, HRV was measured for 24 h, capturing the full spectrum of HRV indices. Furthermore, the sample was composed of Spanish speaking Latinos with diabetes, a very high-risk and also very difficult to reach population despite being a large and fast growing demographic in the U.S. We have previously reported that CALMSD improved diabetes knowledge across all participants [35], and compared to diabetes only, those assigned to the stress management group improved symptoms of depression, anxiety, and self-reported health status [36]. With data from the CALMSD study [35,36], we used SEM to explore the temporal relationship between changes in NA and changes in 24-h HRV using the most standard, widely accepted, and interpretable HRV indices. We hypothesized that 1) at baseline, HRV and NA would show a negative association, 2) baseline HRV would predict change in NA, 3) fasting glucose and stress management treatment would not fully account for observed associations, and, 4) at followup, improvement in HRV would be associated with improvement in NA.

observed in individuals with diabetes [9]. Cross-sectional studies fairly consistently show an inverse relationship between HRV and negative affect (NA). The term negative affect refers broadly to negative emotions and subsumes, for example, feelings of sadness, anxiety, anger, guilt, and aversion. Lower HRV is associated with higher symptoms of depression [10,11] and anxiety [12,13] reduced wellbeing [14], and illness burden among medical patients [15]. Compared to nondiabetic controls, people with diabetes have higher rates of elevated NA, including depressive symptoms [16] and anxiety symptoms [17]. Many people with diabetes experience low subjective emotional wellbeing [18] and they may also experience diabetes distress, i.e., emotional burden from living with diabetes [16]. NA prospectively worsens diabetes outcomes including fasting glucose [19], hospitalizations [20], and early mortality [21]. Latinos show some of the highest rates of type 2 diabetes in the U.S. [22] and exhibit strong relationships among diabetes, NA, and cardiac health [23]. Yet, surprisingly little is known about NA and HRV among Latinos with diabetes. Whereas the literature is not entirely consistent and a causal relationship has not been established, longitudinal cohort studies have tended to support an HRV- > NA temporal ordering. The Whitehall II study followed approximately 2300 participants for an average of 10.5 years. Prospectively, higher HRV at baseline predicted a lower likelihood of incident depressive symptoms at follow-up; depressive symptoms at baseline were not associated with HRV at follow-up [8]. Findings from adolescents [24] also support HRV as a predictor of subsequent depression. Twins (n = 166) from the Vietnam Era Twin Registry who were discordant for depression at baseline were followed for 7 years. Findings of that study support a bidirectional relationship, but there was slightly stronger evidence for the HRV- > NA temporal ordering [25]. Little is known about the HRV- > NA relationship in persons with diabetes despite their high risk for emotional distress and autonomic disturbance. Behavioral treatments to improve mood allow examination of NA and HRV over time. In the most relevant study to date, ZimmermannSchlegel et al. [26] followed participants in a behavioral stress reduction intervention to examine relationships between HRV and NA among n = 113 White patients with type 2 diabetes. They found no crosssectional associations and no longitudinal associations. However, recordings were limited to 5 min so many HRV indices that require longer recording periods were not calculable. The lower frequency bands, which are derived only from 24-h recordings, have been shown to correlate with both mortality and psychosocial factors [27–29]. HRV is most often calculated along two domains – the time domain and the frequency domain, the latter being further decomposed into discrete frequency bands. There are many HRV indices, each reflecting a different physiological process or combination of processes aimed at allostasis such as, for example, parasympathetic, baroreflex, and temperature regulation. The underpinnings of many indices have been only partially elucidated and new indices are occasionally proposed. Whereas the clinical significance of HRV is not disputed, there is far less consensus on selection of indices and their underlying mechanisms. Published papers vary widely regarding which index or indices are reported and indeed, most report only select indices, though the justification for reporting one index over another is rarely transparent. What is well documented is that HRV indices are usually highly correlated, suggesting that HRV is an integrated indicator of allostatic load. Further, data from even small samples are often subjected to multiple tests, i.e., one for each HRV indicator under investigation, which can drastically increase type 1 error. For example, a recent report [24] tested 18 different HR and HRV indices in a sample of n = 113 without correction for type 1 error. Many other reports apply similar statistical methods. Structural equation modeling (SEM) is the preferred data analytical approach when working with ‘latent’ variables or ‘factors’, i.e., those that have multiple distinct but highly correlated indicators. Whereas it

2. Methods 2.1. Sample This is a report of secondary data analysis from the CALMSD study [35,376]. The study was approved by the institutional review boards of all institutions involved. Participants were recruited from the ‘Brownstone Clinic’, an outpatient clinic at Hartford Hospital, serving low-income patients with diabetes, approximately 80% of whom are Latinos. Participants were adult Hartford residents, self-identified Latino or Hispanic, Spanish-speaking, ambulatory, with type 2 diabetes ≥ 6 months, and most recent past year HbA1c > 7.0. Chart review excluded patients for: medical instability or intensive medical treatment that could affect diabetes control (e.g., steroids); bipolar disorder or thought disorder or suicide attempt or psychiatric hospitalization in the past 2 years. Face-to-face screening excluded recruits for alcohol problems or enrollment in another research study. Details of the sampling procedure are available elsewhere [35]. 2.2. Procedures At a morning home visit participants provided informed consent, provided a baseline fasting blood sample, answered questionnaires that were verbally administered by bilingual interviewers in the 2

Journal of Psychosomatic Research 124 (2019) 109774

J.A. Wagner, et al.

3.3. Covariates

participant's preferred language (English or Spanish) and were compensated $10. Participants were instrumented with 7‑lead, 3-channel ambulatory ECG monitors (Holters), GE Medical (Milwaukee, WI) Marquette Series 8500 direct (amplitude-modulated) recorders. Participants were de-instrumented after 24 h. Within approximately 2 weeks participants attended a 2.5 h group diabetes education session delivered by a Community Health Worker (CHW). Next, they were randomized to either diabetes education (DE) only or to diabetes education plus 8 group sessions of stress management (DE + SM). The SM intervention was a manualized, culturally based intervention comprised of 8 group sessions spread across 8–10 weeks that included psychoeducational skills training and physical relaxation training. The intervention [35] and outcomes of the study [36] have been described in detail elsewhere. Finally, after approximately 8–10 weeks, participants provided follow-up survey responses and 24-h HRV recordings.

Treatment Condition. Because treatment was specifically designed to decrease NA, follow-up SEM models controlled for treatment assignment, i.e., DE vs DE + SM. Fasting Glucose. Because glucose is related to both NA and HRV, follow-up models controlled for fasting glucose which was analyzed at the University of Connecticut (UConn Health) John Dempsey Hospital Laboratory using the LXI R system by Beckman CoulterTM. 3.4. Statistical analysis The measurement model for the latent variable “negative affect” (NA) consisted of four observed measures: depressive symptoms as measured by the Patient Health Questionnaire (PHQ), anxiety as measured by the PROMIS scale (PROMIS), diabetes distress as measured by the Problem Areas in Diabetes scale (PAID), and wellbeing as measured by the WHO-5 well-being index (WHO). The measurement model for the latent variable “heart rate variability” (HRV) consisted of observed measures from both the time and frequency domains over 24 h: standard deviation of normal-to-normal intervals (SDNN), root mean square of the successive differences (RMSSD), and lnULF, lnVLF, lnLF, and lnHF. The loadings of the observed variables for each of the two latent constructs was constrained to be equal across the two time points which was supported by model fit statistics. First, to examine baseline associations and to explain the change scores between baseline and follow-up, an SEM bivariate latent change score model was used. Paths from baseline HRV and baseline NA to the variables ΔHRV and ΔNA were estimated [42]. Model fit was assessed by the ratio of chi-square to degrees of freedom, comparative fit index (CFI), and root mean square error of approximation (RMSEA), where good fit is considered χ2/df < 2, CFI > 0.95, and RMSEA < 0.07 [43]. Models were then re-run, controlling for treatment assignment and fasting glucose. Finally, a structural equation model testing for synchronous effects was run [44] to assess the relationship between ΔHRV and ΔNA. Analyses were performed using maximum likelihood estimation in Mplus version 8 [45].

3. Measures 3.1. Negative affect Depressive symptoms. The English and Spanish language versions of the Personal Health Questionnaire (PHQ8; the PHQ9 without the suicidality item) measure depressive symptoms over the past 2 weeks with higher scores indicating more depressive symptoms [37]. We omitted the suicidality item so that CHWs working in the field would not be in the position to have to respond to suicidality. In our sample, baseline coefficient alpha for the PHQ9 was 0.79. Anxiety symptoms. The English and Spanish language versions of the emotional distress/anxiety scale of the Patient Reported Outcomes Measurement Information System (PROMIS Anxiety Short Form 8a) assess anxiety symptoms over the past 7 days [38]. The 8-item scales have response options from 1=”never” to 5=”always” with higher scores indicating more anxiety symptoms; baseline alpha = 0.91. Diabetes Distress. The 5-item English and Spanish language versions of the Problem Areas in Diabetes (PAID) scale assess the patient's perspective of current emotional distress from living with diabetes. Each item was scored 0=“not a problem” to 4 = “serious problem” with higher scores indicating more distress [39]; baseline alpha = 0.92. Subjective Wellbeing. The English and Spanish language versions of the World Health Organization Wellbeing Index are 5-item scales designed to assess emotional well-being. The positively phrased items measure cheerfulness, energy, and life satisfaction over the past two weeks on a 4-point scale from “a little of the time” to “all of the time” [40] baseline alpha = 0.81.

4. Results 4.1. Demographics and descriptive characteristics One hundred and twenty one individuals completed baseline assessments. The sample has been described in detail elsewhere [35,36]. In brief, participants were mean = 60.3 (SD = 11.6) years old, 73% were women, modal education was 8th grade or less, 95% had health insurance (Medicare, Medicaid, or private). Most identified as Puerto Rican (71%), with mean = 34.3 (SD = 13.9) years in the mainland US, and 93% chose to respond to surveys in Spanish. Most treated their diabetes with oral agents (38%) or oral agents plus insulin (43%) and fasting glucose mg/dl was mean = 170.4 (SD = 59.9). In general, participants had high NA. At baseline, the percentage of participants with mild to severe depressive symptoms (score of 5 or higher [37]) was 48%. Baseline diabetes distress was mean = 7.97 (SD = 6.5) which is just below the clinical cutoff of ≥ 8 [46]. Mean anxiety was a T-score of 53, indicating mildly elevated anxiety symptoms. On the WHO Wellbeing Index, raw mean = 12.5 (SD = 6.0) where < 13 is indicative of poor wellbeing. When WHO Wellbeing scores were multiplied by 4, as per scoring guidelines, mean = 50 (SD = 24); substantially below the clinical target for wellbeing of ≥ 70 [47]. HRV values were substantially lower than reported for a healthy, middle-aged sample [SDNN = 96 (SD = 28) vs 141 (SD = 39)] [48]. See Tables 1 and 2 for means, standard deviations, and correlations among each HRV index and each measure of NA at baseline and at follow-up.

3.2. Heart rate variability HRV was assessed as in our previous studies [41]. Holter recordings were scanned by an experienced technician. As described in the literature [29] each tape was manually processed and edited. A list of R-R intervals with annotations denoting normal beats, types of ectopics and noise was then processed and analyzed with customized software. The R-R interval data file was edited to remove ectopics and noise, and gaps filled in by interpolated linear splines. In the time domain, we assessed the standard deviation of the R- R interval (SDNN) and root mean square of the successive differences (RMSSD). The power spectrum was computed using a fast Fourier transform with a Parzen window and corrected for attenuation due to windowing and sampling. The power spectrum was integrated over three frequency bands [29] – ultra low frequency (ULF) 0 to < 0.0033; very low frequency (VLF) 0.0033 to < 0.04 Hz; low frequency (LF) 0.04 to < 0.15 Hz; and high frequency (HF) 0.15 to 0.40 Hz. As expected, HRV values in the frequency domain were non-normally distributed, so were natural log (ln) transformed. 3

Journal of Psychosomatic Research 124 (2019) 109774

J.A. Wagner, et al.

in HRV.

Table 1 Means ± standard deviations of heart rate variability and negative affect variables by time point. Outcome

Baseline

5. Discussion

Post

Heart rate variability SD normal to normal Ultra low frequency Very low frequency Low frequency High frequency RMSSD

96.2 8.53 6.45 5.32 4.55 24.9

± ± ± ± ± ±

28.1 0.65 0.83 1.02 1.11 12.7

97.6 8.61 6.46 5.41 4.64 25.0

± ± ± ± ± ±

26.4 0.58 0.77 1.00 1.01 10.5

Negative affect Subjective well-being Anxiety Diabetes distress Depressive symptoms

12.5 15.1 7.51 6.28

± ± ± ±

6.0 7.6 6.44 5.48

13.7 14.2 6.71 5.34

± ± ± ±

5.4 7.6 6.29 5.45

Leveraging data from a behavioral intervention designed to improve NA, this study used SEM to investigate temporal relationships between NA and 24-h HRV in a sample of low-income, Spanish-speaking Latinos with type 2 diabetes. The main findings are that: 1) at baseline there were strong, negative cross-sectional associations between NA and HRV; 2) baseline NA predicted change in HRV 8–10 weeks later, whereas baseline HRV did not predict change in NA; 3) controlling for fasting glucose and treatment assignment did not alter the direction or statistical significance of these findings; and, 4) at followup, a greater reduction (improvement) in NA was associated with a larger increase (improvement) in HRV with a small-to-medium effect size that approached statistical significance. Our findings suggest that NA contributes to allostatic load. Our finding that NA and HRV were significantly associated at baseline is consistent with and strengthens the extant literature, and extends it to diabetes. Data from 967 participants in the second Midlife in the US (MIDUS II) study found inverse relationships between indices of NA and HF-HRV measured at rest for 11 min [14]. A more recent study of n = 189 patients with coronary artery disease that underwent 24-h recordings found that patients with depression showed lower HRV in the time domain [50]. Change scores in NA and HRV revealed informative temporal patterns. Contrary to our hypotheses, baseline NA predicted ΔHRV, rather than baseline HRV predicting ΔNA. These results held even after controlling for treatment assignment and baseline fasting glucose. This is in contrast to some, but not all, studies that report that HRV predicts NA [51,52]. The duration of time between assessments may be important in this respect, since our follow-up was 8–10 weeks, compared to 7–10 years in those reports. In addition, we collected 24-h HRV and reported HRV in the time domain and frequency bands. Longer intervals between assessments, shorter HRV recording periods, and select HRV indices may yield differing results. At follow-up, ΔHRV and ΔNA were marginally associated. Specifically, decreases in NA were marginally associated with increases in HRV. Our findings may seem to contradict a meta-analysis that concludes that resolution of depression following treatment with antidepressant medication does not improve HRV [53]. However, although taking antidepressants and resolving depression may co-occur, they may produce differing - and potentially even countervailing – effects on HRV [54]. Not all subjects in treatment groups enjoy improvement in depression, some in the control arm may improve through other means, and medication almost certainly exert physiological effects in addition to any on mood. Non-pharmacological interventions designed to improve NA provide a unique opportunity to examine changes in NA and HRV that are independent of the effects of medication per se. Our stress management intervention did not involve medication so the observed reduction in depressive symptoms was not pharmacologically driven. Behavioral interventions may not directly improve HRV as found in the present study and others [55–58]. But any improvements in NA caused by these interventions may in turn improve HRV. Our findings were only marginal for associations between changes in NA and HRV. The effect of NA reduction in this medical, but not psychiatric, sample may have been of insufficient magnitude to alter HRV. Rigorous studies of the effect of behavioral treatment for major depressive disorder on HRV are warranted. Other behavioral interventions and those with larger samples may yield significant results. Exercise benefits both NA and HRV among persons with diabetes [59–61]. The efficacy of such low-risk, non-pharmacological interventions should be tested for combined psychological and autonomic benefits. This study could not establish that improvement in HRV or NA causes improvement in the other because we measured NA and HRV at two timepoints. At least one additional time point would be needed at

4.2. Baseline HRV and NA Fig. 1 shows the path diagram with unstandardized coefficients for the bivariate latent variable change score model where the paths with a “1.0” were fixed, i.e., not estimated. The model fit the data well (χ2(174) = 304 or χ2/df = 1.75, CFI = 0.92, and RMSEA = 0.08). All observed measures loaded significantly on their respective factors (all pvalues < .001). As hypothesized, there was a strong negative relationship between HRV and NA at baseline (cov = −3.81, p = .003) such that subjects reporting higher NA had lower HRV. 4.3. Baseline HRV and NA predicting change in HRV and NA Baseline NA significantly predicted ΔNA (b = −0.21, p = .02); participants with higher (worse) baseline NA showed greater decrease (more improvement) in NA. Baseline NA also predicted ΔHRV (b = 0.06, p = .047); participants with higher (worse) baseline NA showed greater positive change (more improvement) in HRV. Baseline HRV significantly predicted ΔHRV (b = −0.33, p < .001); participants with lower (worse) baseline HRV showed greater increase (more improvement) in HRV. Contrary to our hypothesis, baseline HRV did not predict ΔNA (b = 0.28, p = .38). 4.4. Controlling for treatment assignment and fasting glucose The model was rerun with treatment assignment and baseline fasting glucose as predictors of ΔHRV and ΔNA. The addition of these variables retained a good model fit (χ2/df = 1.75, CFI = 0.91, and RMSEA = 0.08). Treatment assignment did not predict ΔHRV (b = 0.13, p = .71) but did predict ΔNA (b = −1.77, p = .05) such that participants assigned to DE + SM showed a greater decrease (improvement) in NA than those assigned to DE only. Baseline glucose did not predict ΔHRV (b = −0.01, p = .32) but did predict ΔNA (b = −0.03, p = .003) such that participants with higher baseline glucose showed a greater decrease (improvement) in NA. Further, with treatment assignment and baseline glucose in the model, baseline NA still predicted ΔHRV (b = 0.06, p = .040) and baseline HRV still did not predict ΔNA (b = 0.09, p = .76). 4.5. Relationship between change in HRV and change in NA The SEM synchronous model examining the association between ΔHRV and ΔNA had good model fit (χ2(160) = 282 or χ2/df = 1.76, CFI = 0.927, and RMSEA = 0.080). As hypothesized, a larger reduction (improvement) in NA was associated with a larger increase (improvement) in HRV. The coefficient for the change in HRV with the change in NA was negative and approached statistical significance (b = −1.04, p = .06). The standardized coefficient equaled −0.286 indicating a small-to-medium correlation [49] between change in NA with change 4

5

SDNN ULF VLF LF HF RMSDD WHO Anx PAID PHQ SDNN ULF VLF LF HF RMSDD WHO Anx PAID PHQ Mean SD

0.87 0.69 0.71 0.57 0.48 0.15 −0.08 −0.06 −0.09 0.57 0.50 0.58 0.50 0.32 0.35 0.06 −0.09 −0.02 −0.08 96.2 28.1

0.70 0.67 0.54 0.45 0.18 −0.12 −0.14 −0.11 0.65 0.65 0.53 0.44 0.38 0.47 0.16 −0.12 0.1 −0.12 8.5 0.65

ULF

0.90 0.72 0.55 0.30 −0.29 −0.17 −0.31 0.63 0.48 0.72 0.60 0.45 0.44 0.21 −0.07 −0.14 −0.21 6.4 0.83

VLF

0.81 0.65 0.30 −0.26 −0.13 −0.28 0.52 0.38 0.70 0.69 0.53 0.54 0.21 −0.12 −0.18 −0.21 5.3 1.02

LF

0.91 0.23 −0.19 −0.14 −0.21 0.57 0.47 0.62 0.61 0.71 0.70 0.15 −0.07 −0.22 −0.23 4.5 1.11

HF

0.11 −0.06 −0.12 −0.08 0.51 0.45 0.54 0.52 0.67 0.74 0.03 −0.04 −0.19 −0.05 24.9 12.7

RMSDD

−0.71 −0.38 −0.67 0.14 0.14 −0.02 0.01 −0.01 0.10 0.65 −0.59 −0.42 −0.60 12.5 6.0

WHO

0.41 0.76 −0.10 −0.10 0.05 0.03 0.08 −0.03 −0.60 0.69 0.50 0.60 15.1 7.6

Anx

0.38 −0.09 −0.18 0.14 0.13 0.17 0.11 −0.28 0.37 0.53 0.35 7.5 6.4

PAID

−0.16 −0.09 −0.08 −0.05 −0.04 0.02 −0.61 0.47 0.42 0.66 6.3 5.5

PHQ

0.89 0.61 0.47 0.35 0.33 0.07 −0.11 0.07 −0.17 97.6 26.4

SDNN

0.47 0.32 0.21 0.24 0.06 −0.18 0.03 −0.21 8.6 0.58

ULF

0.89 0.62 0.50 0.06 −0.01 0.02 −0.02 6.5 0.77

VLF

0.77 0.65 0.05 −0.04 −0.02 −0.04 5.4 1.00

LF

Follow-up measurements

0.86 0.03 0.06 0.01 −0.05 4.6 1.01

HF

0.06 −0.07 −0.01 −0.07 25.0 10.5

RMSDD

−0.68 −0.50 −0.69 13.7 5.4

WHO

0.61 0.74 14.2 7.6

Anx

0.53 6.7 6.3

PAID

5.3 5.5

PHQ

Legend: SDNN = standard deviation of the normal-to-normal R-R interval; ULF = ultra-low frequency; VLF = very low frequency; LF = low frequency; HF = high frequency; RMSSD = root mean square of the successive differences; Anx = PROMIS Emotional distress/anxiety scale (Short Form 8a); WHO = World Health Organization Wellbeing Index; PAID = Problem Areas in Diabetes; PHQ = Personal Health Questionnaire, SD = Standard deviation.

Follow-Up

Baseline

SDNN

Baseline measurements

Table 2 Correlations, means, and standard deviations.

J.A. Wagner, et al.

Journal of Psychosomatic Research 124 (2019) 109774

Journal of Psychosomatic Research 124 (2019) 109774

J.A. Wagner, et al.

Fig. 1. Path diagram for the bivariate latent variable change score model. Legend: SDNN = standard deviation of the normal-to-normal R-R interval; ULF = ultra-low frequency; VLF = very low frequency; LF = low frequency; HF = high frequency; RMSSD = root mean square of the successive differences; PROMIS = PROMIS Emotional distress/anxiety scale (Short Form 8a); WHO = World Health Organization Wellbeing Index; PAID = Problem Areas in Diabetes; PHQ = Personal Health Questionnaire; HRV = heart rate variability; Neg Aff = negative affect; chg HRV = change in heart rate variability; chg Neg Aff = change in negative affect. Variables labeled “1” were measured at baseline, variables labeled “2” were measured at follow-up.

some studies it is the most closely associated with health outcomes [29] and in other, although not all studies, it is most closely associated with psychological stressors [28]. One study has shown associations with temperature regulation [76] and others with physical activity [77]. Provisional recommendations for the handling and reporting of HRV indices have recently been made [78]. Those recommendations include being explicit about the theoretical research model and the use of an analytic strategy that is in line with that model. Those authors specifically offer a “plea” for approaches such as structural equation models that employ latent factors to expand the HRV indices under investigation. Similarly, we propose that how HRV is statistically modeled turns entirely on the research question at hand. If the goal is prediction of a clinical outcome such as depressive symptoms as in this paper, a latent variable approach may be preferable. A corollary example from cardiovascular disease may be illustrative. Whereas a single variable such as age or smoking status can predict cardiovascular outcomes, it is well-recognized that composite scores that combine risk factors provide better prediction, and risk-assessment guidelines continue to use composites [79]. However, when the variables in a given risk equation are highly correlated and are thought to reflect different aspects of a more general factor, common statistical tests such as linear regression can be problematic for several reasons including most specifically multicollinearity. SEM-derived latent factors are more suitable under these conditions which apply to HRV. Our own study suggests that a latent HRV variable is a suitable and efficient strategy as evidenced by the high factor loadings for each HRV index on the latent HRV variable and the very good model fit. Modeling a latent HRV factor provides an integrated, easily interpretable measure of autonomic function and allostatic load that can be related to clinical variables such

which change between times 1 and 2 could predict change between times 2 and 3. It remains possible and even likely that a bidirectional relationship may exist, with the observed temporal ordering depending on when in the loop the observation occurs. Alternatively, both HRV and depressive symptoms may be partially modulated by a common third factor that was not measured such as, for example, inflammation [62]. HRV is a powerful, aggregated risk factor that reflects numerous ‘upstream’ mechanisms. Our control of fasting glucose suggests that glucose control does not account for our findings. Finally, our study supports the use of SEM for modeling HRV. The clinical significance of HRV measured in the time and frequency domains is well accepted [63–66]. For example, in the Framingham Heart Study, a one standard deviation difference in the lnLF power nearly doubled the risk of all-cause mortality [67]. Yet, what each specific HRV index represents is less clear and how psychosocial experiences and exposures map onto each index is not well elucidated [68]. There are varying opinions regarding the underlying physiology of time domains and frequency bands and the relative contribution of sympathetic and vagal control to each [69]. Notwithstanding, all HRV indices have been linked to adaptability of regulatory functions. Both HF power and RMSSD are recognized to reflect parasympathetic activity [70–72]. While many studies suggest LF contains components of both vagal and sympathetic activity [70] others suggest that LF reflects the ability to modulate autonomic outflow via the baroreflex, rather than serving as a direct measure of autonomic activity per se [73]. VLF activity may reflect the renin-angiotensin system as well as parasympathetic activity [74]. 24-h SDNN captures diurnal variation, known to predict health [75] as 30–40% of SDNN is related to day-night differences in heart rate [70]. The physiological underpinnings of ULF are least understood – in

6

Journal of Psychosomatic Research 124 (2019) 109774

J.A. Wagner, et al.

as NA. Alternatively, if the research goal is to elucidate underlying physiological mechanisms, then examining specific indices individually may be most appropriate. For example, which frequency band best reflects temperature regulation? When theory requires that such an index-by-index approach should be taken, investigators are cautioned to ensure that the number of statistical tests does not exceed the limits of the sample size and that there is reasonable protection against type I error.

[2] I. Kawachi, D. John, T. Catherine, MacArthur Foundation Research Network on Socioeconomic Status and Health: Heart Rate variability, (1997). [3] B.S. McEwen, E. Stellar, Stress and the individual. Mechanisms leading to disease, Arch. Intern. Med. 153 (1993) 2093–2101. [4] L.R. Wulsin, P.S. Horn, J.L. Perry, J.M. Massaro, R.B. DʼAgostino Sr., The contribution of autonomic imbalance to the development of metabolic syndrome, Psychosom. Med. 78 (4) (2016 May) 474–480. [5] K.P. Ratzmann, M. Raschke, I. Gander, E. Schimke, Prevalence of peripheral and autonomic neuropathy in newly diagnosed type II (noninsulin-dependent) diabetes, J. Diabet. Complicat. 5 (1) (1991) 1–5. [6] A.I. Vinik, R.E. Maser, B.D. Mitchell, R. Freeman, Diabetic autonomic neuropathy, Diabetes Care 26 (5) (2003) 1553–1579. [7] J. Gerritsen, J.M. Dekker, B.J. TenVoorde, et al., Impaired autonomic function is associated with increased mortality, especially in subjects with diabetes, hypertension, or a history of cardiovascular disease: the Hoorn Study, Diabetes Care 24 (10) (2001) 1793–1798. [8] V.K. Jandackova, S. Scholes, A. Britton, A. Steptoe, Are changes in heart rate variability in middle-aged and older people normative or caused by pathological conditions? Findings from a large population-based longitudinal cohort study, J. Am. Heart Assoc. 5 (2) (2016 Feb 12). [9] E.B. Schroeder, L.E. Chambless, D. Liao, et al., Diabetes, glucose, insulin, and heart rate variability: the atherosclerosis risk in communities (ARIC) study, Diabetes Care 28 (3) (2005) 668–674. [10] M.E. Bleil, P.J. Gianaros, J.R. Jennings, J.D. Flory, S.B. Manuck, Trait negative affect: toward an integrated model of understanding psychological risk for impairment in cardiac autonomic function, Psychosom. Med. 70 (3) (2008) 328–337. [11] C. Koch, M. Wilhelm, S. Salzmann, W. Rief, F. Euteneuer, A meta-analysis of heart rate variability in major depression, Psychol. Med. (2019 Jun 26) 1–10, https://doi. org/10.1017/S0033291719001351 (Epub ahead of print). [12] J.L. Francis, A.A. Weinstein, D.S. Krantz, M.C. Haigney, P.K. Stein, P.H. Stone, J.S. Gottdiener, W.J. Kop, Association between symptoms of depression and anxiety with heart rate variability in patients with implantable cardioverter defibrillators, Psychosom. Med. 71 (8) (2009 Oct) 821–827. [13] J.A. Chalmers, D.S. Quintana, M.J. Abbott, A.H. Kemp, Anxiety disorders are associated with reduced heart rate variability: a meta-analysis, Front Psychiatry. 5 (2014 Jul 11) 80. [14] R.P. Sloan, E. Schwarz, P.S. McKinley, et al., Vagally-mediated heart rate variability and indices of well-being: results of a nationally representative study, Health Psychol. 36 (1) (2017) 73. [15] M. Princip, M. Scholz, R.E. Meister-Langraf, J. Barth, U. Schnyder, H. Znoj, J.P. Schmid, J.F. Thayer, R. von Känel, Can illness perceptions predict lower heart rate variability following acute myocardial infarction? Front. Psychol. 7 (2016 Nov 18) 1801. [16] F. Rotella, E. Mannucci, Diabetes mellitus as a risk factor for depression. A metaanalysis of longitudinal studies, Diabetes Res. Clin. Pract. 99 (2) (2013) 98–104. [17] K.J. Smith, M. Béland, M. Clyde, et al., Association of diabetes with anxiety: a systematic review and meta-analysis, J. Psychosom. Res. 74 (2) (2013) 89–99. [18] A. Nicolucci, K. Kovacs Burns, R.I. Holt, et al., Diabetes attitudes, wishes and needs second study (DAWN2): cross-national benchmarking of diabetes-related psychosocial outcomes for people with diabetes, Diabet. Med. 30 (7) (2013) 767–777. [19] M.M. Skaff, J.T. Mullan, D.M. Almeida, L. Hoffman, U. Masharani, D. Mohr, L. Fisher, Daily negative mood affects fasting glucose in type 2 diabetes, Health Psychol. 28 (3) (2009 May) 265–272, https://doi.org/10.1037/a0014429. [20] D.S. Davydow, W.J. Katon, E.H. Lin, et al., Depression and risk of hospitalizations for ambulatory care-sensitive conditions in patients with diabetes, J. Gen. Intern. Med. 28 (7) (2013) 921–929. [21] K. Naicker, J.A. Johnson, J.C. Skogen, et al., Type 2 diabetes and comorbid symptoms of depression and anxiety: longitudinal associations with mortality risk, Diabetes Care 40 (3) (2017) 352–358. [22] A. Menke, S. Casagrande, L. Geiss, C.C. Cowie, Prevalence of and trends in diabetes among adults in the United States, 1988-2012, JAMA. 314 (10) (2015 Sep 8) 1021–1029, https://doi.org/10.1001/jama.2015.10029. [23] J.A. Zimmerman, B.T. Mast, T. Miles, K.S. Markides, Vascular risk and depression in the Hispanic Established Population for the Epidemiologic Study of the Elderly (EPESE), Int J Geriatr Psychiatr. 24 (4) (2009 Apr) 409–416. [24] L. Vazquez, J.D. Blood, J. Wu, T.M. Chaplin, R.E. Hommer, H.J. Rutherford, M.N. Potenza, L.C. Mayes, M.J. Crowley, High frequency heart rate variability predicts adolescent depressive symptoms, particularly anhedonia, across one year, J. Affect. Disord. 196 (2016 May 15) 243–247. [25] M. Huang, A. Shah, S. Su, J. Goldberg, R.J. Lampert, O.M. Levantsevych, L. Shallenberger, P. Pimple, J.D. Bremner, V. Vaccarino, Association of sepressive symptoms and heart rate variability in Vietnam war-era twins: a longitudinal twin difference study, JAMA Psychiatry. 75 (7) (2018 Jul 1) 705–712. [26] V. Zimmermann-Schlegel, B. Wild, P. Nawroth, S. Kopf, W. Herzog, M. Hartmann, Impact of depression and psychosocial treatment on heart rate variability in patients with type 2 diabetes mellitus: an exploratory analysis based on the HEIDIS trial, Exp. Clin. Endocrinol. Diabetes 127 (6) (2019 Jun) 367–376, https://doi.org/ 10.1055/s-0043-125445 Epub 2018 Feb 8. [27] A.J. Shah, R. Lampert, J. Goldberg, E. Veledar, J.D. Bremner, V. Vaccarino, Posttraumatic stress disorder and impaired autonomic modulation in male twins, Biol. Psychiatry 73 (2013) 1103–1110. [28] V. Vaccarino, R. Lampert, J.D. Bremner, F. Lee, S. Su, C. Maisano, N.V. Murrah, L. Jones, F. Jawed, N. Afzal, A. Ashraf, J. Goldberg, Depressive symptoms and heart rate variability: evidence for a shared genetic substrate in a study of twins, Psychosom. Med. 70 (2008) 628–636. [29] J.T. Bigger Jr., J.L. Fleiss, R.C. Steinman, L.M. Rolnitzky, R.E. Kleiger,

6. Limitations and future directions These findings should be interpreted in light of several limitations. The sample is small and relatively homogenous, though comparable to or larger than other behavioral interventions examining HRV. As discussed above, the design does not allow conclusions about causality. Only a single measure of wellbeing was measured so the unique contribution of a variety of positive affective states cannot be partialed out. We were not able to control for antidepressant use which may be important [80]. We did not measure sleep onset and wake time during the 24-h HRV assessments, so we cannot address whether diurnal and nocturnal values of certain HRV indices show different associations with NA. This shortcoming is mitigated to some extent by our collection of SDNN which reflects diurnal differences. These limitations are generally outweighed by strengths, including multiple measures of negative affect and the sample's unique composition, i.e., hard-to-reach Latinos with type 2 diabetes. Such participants are often not included in intensive research protocols such as 24-h HRV recording because most studies are not designed to overcome socio-cultural barriers to their inclusion such as language, literacy, numeracy, distrust, nescience of research, and logistical challenges such as transportation. Future studies should employ larger samples, incorporate interventions such as exercise that may extend benefits commonly to both HRV and depressive symptoms, and explore potential shared, ‘upstream’ mechanisms. Funding This work was supported by grants from the National Institute of Minority Health and Health Disparities, grant number R01MD005879, and the American Diabetes Association, grant number 7-13-TS-31. Declaration of Competting Interest The authors declare that they have no conflict of interest. The authors have no competing interests to report. Ethical standards All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/ or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all individual participants included in the study. Acknowledgements The authors would like to thank Hartford Hospital, the staff of the Hispanic Health Council, and the study participants for their contributions to this project. References [1] T. Seeman, B. Singer, J. Rowe, R. Horwitz, B. McEwen, Price of adaptation–allostatic load and its health consequences, Arch. Intern. Med. 157 (1997) 2259–2268.

7

Journal of Psychosomatic Research 124 (2019) 109774

J.A. Wagner, et al.

[30]

[31] [32]

[33]

[34]

[35]

[36]

[37]

[38]

[39]

[40]

[41]

[42] [43] [44] [45] [46]

[47]

[48]

[49] [50]

[51]

[52]

[53]

[54]

[55]

J.N. Rottman, Frequency domain measures of heart period variability and mortality after myocardial infarction, Circulation 85 (1) (1992) 164–171. M.X. Hu, B.W.J.H. Penninx, E.J.C. de Geus, F. Lamers, D.C. Kuan, A.G.C. Wright, A.L. Marsland, M.F. Muldoon, S.B. Manuck, P.J. Gianaros, Associations of immunometabolic risk factors with symptoms of depression and anxiety: the role of cardiac vagal activity, Brain Behav. Immun. 73 (2018 Oct) 493–503. J.M. McCaffery, A.L. Marsland, K. Strohacker, M.F. Muldoon, S.B. Manuck, Factor structure underlying components of allostatic load, PLoS One 7 (10) (2012) e47246. K. Bertsch, D. Hagemann, E. Naumann, H. Schächinger, A. Schulz, Stability of heart rate variability indices reflecting parasympathetic activity, Psychophysiology. 49 (5) (2012 May) 672–682. S.L. Mann, E.A. Selby, M.E. Bates, Contrada RJ4 integrating affective and cognitive correlates of heart rate variability: a structural equation modeling approach, Int. J. Psychophysiol. 98 (1) (2015 Oct) 76–86. E.S. Baja, J.D. Schwartz, B.A. Coull, G.A. Wellenius, P.S. Vokonas, H.H. Suh, Structural equation modeling of parasympathetic and sympathetic response to traffic air pollution in a repeated measures study, Environ. Health 12 (1) (2013 Sep 23) 81. J. Wagner, A. Bermudez-Millan, G. Damio, et al., Community health workers assisting Latinos manage stress and diabetes (CALMS-D): rationale, intervention design, implementation, and process outcomes, Transl. Behav. Med. 5 (4) (2015) 415–424. J.A. Wagner, A. Bermudez-Millan, G. Damio, et al., A randomized, controlled trial of a stress management intervention for Latinos with type 2 diabetes delivered by community health workers: outcomes for psychological wellbeing, glycemic control, and cortisol, Diabetes Res. Clin. Pract. 120 (2016) 162–170. L. Wulsin, E. Somoza, J. Heck, The feasibility of using the Spanish PHQ-9 to screen for depression in primary care in Honduras, Primary Care Companion to the J Clin Psychiatry. 4 (5) (2002) 191. D. Cella, S. Yount, N. Rothrock, et al., The patient-reported outcomes measurement information system (PROMIS): progress of an NIH roadmap cooperative group during its first two years, Med. Care 45 (5Suppl 1) (2007) S3–S11. G. Welch, C.E. Schwartz, P. Santiago-Kelly, J. Garb, R. Shayne, R. Bode, Diseaserelated emotional distress of Hispanic and non-Hispanic type 2 diabetes patients, Ethn. Dis. 17 (3) (2007 Summer) 541–547. R. Lucas-Carrasco, Reliability and validity of the Spanish version of the World Health Organization-five well-being index in elderly, Psychiatry Clin. Neurosci. 66 (6) (2012) 508–513. R. Lampert, J. Ickovics, R. Horwitz, F. Lee, Depressed autonomic nervous system function in African Americans and individuals of lower social class: a potential mechanism of race- and class-related disparities in health outcomes, Am. Heart J. 150 (1) (2005) 153–160. K.J. Grimm, G.L. Mazza, M.M. Mazzocco, Advances in methods for assessing longitudinal change, Educ. Psychol. 51 (3–4) (2016) 342–353. D. Hooper, J. Coughlan, M. Mullen, Structural Equation Modelling: Guidelines for Determining Model Fit. Articles, (2008), p. 2. J. Newsom, Longitudinal Structural Equation Modeling: A Comprehensive Introduction, Routledge, New York, NY, 2015. L. Muthén, B. Muthén, Mplus User's Guide, Eight ed., Muthén & Muthén, Los Angeles, 1998. B.E. McGuire, T.G. Morrison, N. Hermanns, et al., Short-form measures of diabetesrelated emotional distress: the problem areas in diabetes scale (PAID)-5 and PAID-1, Diabetologia. 53 (1) (2010) 66–69. C.W. Topp, S.D. Østergaard, S. Søndergaard, P. Bech, The WHO-5 well-being index: a systematic review of the literature, Psychother. Psychosom. 84 (3) (2015) 167–176. J.T. Bigger Jr., J.L. Fleiss, R.C. Steinman, L.M. Rolnitzky, W.J. Schneider, P.K. Stein, RR variability in healthy, middle-aged persons compared with patients with chronic coronary heart disease or recent acute myocardial infarction, Circulation. 91 (7) (1995) 1936–1943. J. Cohen, Statistical Power Analysis for the Behavioral Sciences, 2nd ed., Lawrence Earlbaum Associates, Hillsdale, NJ, 1988. Y. Luo, S. Zhang, R. Zheng, L. Xu, J. Wu, Effects of depression on heart rate variability in elderly patients with stable coronary artery disease, J Evid Based Med. 11 (4) (2018 Nov) 242–245, https://doi.org/10.1111/jebm.12310 (Epub 2018 Aug 9). C.W. Sung, H.C. Lee, Y.H. Chiang, W.T. Chiu, S.F. Chu, J.C. Ou, S.H. Tsai, K.H. Liao, C.M. Lin, J.W. Lin, G.S. Chen, W.J. Li, J.Y. Wang, Early dysautonomia detected by heart rate variability predicts late depression in female patients following mild traumatic brain injury, Psychophysiology. 53 (4) (2016 Apr) 455–464. A. Tessier, I. Sibon, M. Poli, M. Audiffren, M. Allard, M. Pfeuty, Resting heart rate predicts depression and cognition early after ischemic stroke: a pilot study, J. Stroke Cerebrovasc. Dis. 26 (10) (2017 Oct) 2435–2441. A.H. Kemp, D.S. Quintana, M.A. Gray, K.L. Felmingham, K. Brown, J.M. Gatt, Impact of depression and antidepressant treatment on heart rate variability: a review and meta-analysis, Biol. Psychiatry 67 (11) (2010 Jun 1) 1067–1074. C.M. Licht, E.J. de Geus, R. van Dyck, B.W. Penninx, Longitudinal evidence for unfavorable effects of antidepressants on heart rate variability, Biol. Psychiatry 68 (9) (2010 Nov 1) 861–868. A.R. Brunoni, A.H. Kemp, E.M. Dantas, A.C. Goulart, M.A. Nunes, P.S. Boggio, J.G. Mill, P.A. Lotufo, F. Fregni, I.M. Benseñor, Heart rate variability is a trait marker of major depressive disorder: evidence from the sertraline vs. electric

[56]

[57]

[58]

[59]

[60]

[61]

[62]

[63]

[64]

[65]

[66]

[67]

[68] [69] [70]

[71]

[72]

[73]

[74] [75]

[76] [77]

[78]

[79] [80]

8

current therapy to treat depression clinical study, Int. J. Neuropsychopharmacol. 16 (9) (2013 Oct) 1937–1949. T.L. Zucker, K.W. Samuelson, F. Muench, M.A. Greenberg, R.N. Gevirtz, The effects of respiratory sinus arrhythmia biofeedback on heart rate variability and posttraumatic stress disorder symptoms: a pilot study, Appl Psychophysiol Biofeedback. 34 (2) (2009 Jun) 135–143. X. Chen, R. Yang, L. Ge, J. Luo, R. Lu, Hypnosis in the treatment of major depression: an analysis of heart rate variability, Int. J. Clin. Exp. Hypn. 65 (1) (2017 Jan-Mar) 52–63. E. Ionson, J. Limbachia, S. Rej, K. Puka, R.I. Newman, S. Wetmore, A.M. Burhan, A. Vasudev, Effects of Sahaj Samadhi meditation on heart rate variability and depressive symptoms in patients with late-life depression, Br. J. Psychiatry 28 (2018 Nov) 1–7. S. Villafaina, D. Collado-Mateo, J.P. Fuentes, E. Merellano-Navarro, N. Gusi, Physical exercise improves heart rate variability in patients with type 2 diabetes: a systematic review, Curr Diab Rep. 17 (11) (2017) 110. M. de Groot, T. Doyle, M. Kushnick, J. Shubrook, J. Merrill, E. Rabideau, F. Schwartz, Can lifestyle interventions do more than reduce diabetes risk? Treating depression in adults with type 2 diabetes with exercise and cognitive behavioral therapy, Curr Diab Rep. 12 (2) (2012 Apr) 157–166. G. Toni, M. Belvederi Murri, M. Piepoli, S. Zanetidou, A. Cabassi, S. Squatrito, L. Bagnoli, A. Piras, et al., Physical exercise for late-life depression: effects on heart rate variability, Am. J. Geriatr. Psychiatry 24 (11) (2016 Nov) 989–997. M.X. Hu, B.W.J.H. Penninx, E.J.C. de Geus, F. Lamers, D.C. Kuan, A.G.C. Wright, A.L. Marsland, M.F. Muldoon, S.B. Manuck, P.J. Gianaros, Associations of immunometabolic risk factors with symptoms of depression and anxiety: the role of cardiac vagal activity, Brain Behav. Immun. 73 (2018 Oct) 493–503. R.E. Kleiger, J.P. Miller, J.T. Bigger Jr., A.J. Moss, Decreased heart rate variability and its association with increased mortality after acute myocardial infarction, Am. J. Cardiol. 59 (4) (1987) 256–262. D. Liao, J. Cai, W.D. Rosamond, et al., Cardiac autonomic function and incident coronary heart disease: a population-based case-cohort study. The ARIC study Atherosclerosis Risk in Communities Study, Am J Epidemiol. 145 (8) (1997) 696–706. H.V. Huikuri, V. Jokinen, M. Syvänne, et al., Heart rate variability and progression of coronary atherosclerosis, Arterioscler. Thromb. Vasc. Biol. 19 (8) (1999) 1979–1985. M.T. La Rovere, G.D. Pinna, R. Maestri, et al., Short-term heart rate variability strongly predicts sudden cardiac death in chronic heart failure patients, Circulation. 107 (4) (2003) 565–570. H. Tsuji, F.J. Venditti, E.S. Manders, et al., Reduced heart rate variability and mortality risk in an elderly cohort. The Framingham heart study, Circulation. 90 (2) (1994) 878–883. A.E. Draghici, J.A. Taylor, The physiological basis and measurement of heart rate variability in humans, J. Physiol. Anthropol. 35 (2016) 22. D.L. Eckberg, Sympathovagal balance: a critical appraisal, Circulation 96 (1997) 3224–3232. Task Force of the European Society of Cardiology and the North American Society of Paicng and Electrophysiology, Heart rate variability: Standards of measurement, physiological interpretation, clinical use, Eur. Heart J. 17 (1996) 354–381. P. Grossman, J. van Beek, C. Wientjes, A comparison of three quantification methods for estimation of respiratory sinus arrhythmia, Psychophysiology 27 (1990) 702–714. S. Akselrod, D. Gordon, F.A. Ubel, D.C. Shannon, A.C. Barger, R.J. Cohen, Power spectrum analysis of heart rate fluctuation: a quantitative probe of beat-to-beat cardiovascular control, Science 213 (1981) 220–222. D.S. Goldstein, O. Bentho, M.-Y. Park, Y. Sharabi, Low-frequency power of heart rate variability is not a measure of cardiac sympathetic tone but may be a measure of modulation of cardiac autonomic outflows by baroreflexes, Exp. Physiol. 96 (2011) 1255–1261. J. Taylor, D. Carr, C. Myers, D. Eckberg, Mechanisms underlying very-low-frequency RR-interval oscillations in humans, Circulation 98 (1998) 547–555. L. Boer-Martins, V.N. Figueiredo, C. Demacq, L.C. Martins, F. Consolin-Colombo, M.J. Figueiredo, F.P.S. Cannavan, H.R. Moreno, Relationship of autonomic imbalance and circadian disruption with obesity and type 2 diabetes in resistant hypertensive patients, Cardiovasc. Diabetol. 10 (2011) 24. L.A. Fleisher, S.M. Frank, D.I. Sessler, C. Cheng, T. Matsukawa, C.A. Vannier, Thermoregulation and heart rate variability, Clin. Sci. 90 (1996) 97–103. D. Roach, W. Wilson, D. Ritchie, R. Sheldon, Dissection of long-range heart rate variability: controlled induction of prognostic measures by activity in the laboratory, J. Am. Coll. Cardiol. 43 (2004) 2271–2277. E.J.C. de Geus, P.J. Gianaros, R.C. Brindle, J.R. Jennings, G.G. Berntson, Should heart rate variability be “corrected” for heart rate? Biological, quantitative, and interpretive considerations, Psychophysiology. 56 (2) (2019 Feb) e13287 Published online 2018 Oct 25 https://doi.org/10.1111/psyp.13287. D.C. Goff, D.M. Lloyd-Jones, G. Bennett, et al., ACC/AHA guideline on the assessment of cardiovascular risk, Circulation 2014 (129) (2013) S49–S73. A.H. Kemp, A.R. Brunoni, I.S. Santos, M.A. Nunes, E.M. Dantas, R. Carvalho de Figueiredo, A.C. Pereira, A.L. Ribeiro, J.G. Mill, R.V. Andreão, J.F. Thayer, I.M. Benseñor, P.A. Lotufo, Effects of depression, anxiety, comorbidity, and antidepressants on resting-state heart rate and its variability: an ELSA-Brasil cohort baseline study, Am. J. Psychiatry 171 (12) (2014 Dec 1) 1328–1334.