Eight weeks of device-guided slow breathing decreases sympathetic nervous reactivity to stress in posttraumatic stress disorder

Abstract

Posttraumatic stress disorder (PTSD) is characterized by increased risk for developing hypertension and cardiovascular disease. We recently showed that device-guided slow breathing (DGB) acutely lowers blood pressure (BP) and muscle sympathetic activity (MSNA) and improves baroreflex sensitivity (BRS) in PTSD. The aim of this study was to assess the long-term benefits of DGB on autonomic function at rest and during stress. We hypothesized that long-term DGB improves arterial BRS and lowers BP and MSNA in PTSD. Twenty-five veterans with PTSD were studied and randomized to either 8 wk of daily DGB (n = 12) or 8 wk of sham device (Sham; n = 13). BP, heart rate (HR), and MSNA were measured at rest and during mental math. Arterial BRS was assessed using the modified Oxford technique. Resting MSNA, BP, and heart rate (HR) remained comparable before and after 8 wk in both groups (DGB and Sham). Likewise, the change in sympathetic and cardiovagal BRS was not different between the groups. Interestingly, DGB significantly decreased MSNA reactivity to mental math when expressed as burst frequency (P = 0.012) or burst incidence (P = 0.008) compared with Sham, suggesting a sustained effect of DGB on sympathetic reactivity to stress in PTSD. Contrary to our hypothesis, long-term DGB did not lower systolic BP, diastolic BP, or HR responses to stress compared with Sham. Likewise, pulse pressure reactivity after 8 wk (P = 0.121) was also comparable. In summary, these data suggest that long-term use of DGB may lead to a sustained dampening of sympathetic reactivity to mental stress in PTSD.

INTRODUCTION

Posttraumatic stress disorder (PTSD) is a debilitating mental illness independently linked to a two- to threefold greater risk of developing hypertension and cardiovascular disease (CVD) (14, 15). PTSD is highly prevalent in both the military and general population. As many as 20% of post-9/11 veterans (46) are estimated to suffer from PTSD, and up to 8% of the general United States population (28, 29) will meet the diagnostic criteria for PTSD in their lifetime. Recent studies have reported that PTSD is associated with blunted arterial baroreflex sensitivity (BRS) and higher sympathetic nervous system (SNS) reactivity (16, 38), all independent risk factors for hypertension and cardiovascular disease (20, 21, 36). Additionally, patients with PTSD also demonstrate an exaggerated neurocardiovascular response to acute stress (38), another known risk factor for hypertension and CVD. Available pharmacological approaches targeting the SNS such as sympatholytic medications are often poorly tolerated due to side effects and complicated by long-term metabolic effects that can exacerbate the risk of CVD (8, 22). Therefore, safe and effective alternative treatments are needed in this high-risk population.

Nonpharmacological interventions targeting risk factors for hypertension such as SNS overactivity and blunted arterial BRS may have the potential to mitigate future risk of CVD in PTSD (17). Potential evidence-based nonpharmacological approaches include lifestyle modification such as diet, exercise, sleep, meditation, and relaxation. For example, increased physical activity has been shown to reduce CVD risk, although nonadherence to exercise may limit its clinical utility (45, 50, 52). Device-guided slow breathing (DGB) is another nonpharmacological approach (7), in which breathing is slowed to 5–6 breaths/min via an interactive biofeedback device (RESPeRATE), and is currently Federal Drug Administration (FDA)-approved for relaxation and the adjunctive treatment of hypertension (48). Slow breathing at these subphysiological rates is purported to decrease blood pressure (BP) by activation of pulmonary stretch receptors induced by the compensatory increase in tidal volume, leading to decreased central sympathetic outflow (47, 48). Although the evidence is conflicting (1, 12, 31, 32, 34), DGB has previously been shown to lower blood pressure (BP) and sympathetic nerve activity in humans with hypertension (23–26, 37). Prior studies have shown that DGB performed for at least 15 min per day for 8 wk led to sustained reductions in BP in hypertensive patients (26, 37).

We previously showed that DGB using the RESPeRATE device acutely lowers muscle sympathetic activity (MSNA) and improves sympathetic baroreflex sensitivity (BRS) in veterans with PTSD (17). However, it is unclear whether long-term DGB leads to sustained improvements in neurocardiovascular function or sympathetic reactivity to stress in PTSD. Therefore, with the use of a randomized, sham-controlled design, the goals of the present study were to determine whether 8 wk of DGB for at least 15 min daily in PTSD leads to sustained improvements in resting BP, heart rate (HR), sympathetic and cardiovagal BRS, and MSNA. In addition, our prior study showed that PTSD patients had greater MSNA reactivity during mental math (but not to other sympathoexcitatory stimuli such as the cold pressor test) (17); therefore, we also examined whether 8 wk of daily DGB in PTSD leads to sustained improvements in BP, HR, and MSNA reactivity to mental stress induced by mental math.

METHODS

We conducted a double-masked, sham-controlled, randomized clinical trial (NCT01627301) in which we randomized PTSD participants to 8 wk of daily at-home breathing with either DGB or a sham device, using a computer-generated blocked randomization scheme. The study coordinator administered the randomization, while the participants, study investigators, and data analysts remained masked to individual treatment assignments. Some PTSD participants (n = 14; 8 DGB and 6 Sham) that were part of our initial cross-sectional study (17) assessing the acute effect of DGB, were subsequently enrolled into the current longitudinal clinical trial.

Participants

Twenty-five post-9/11 veterans with elevated resting blood pressure in the prehypertensive range (resting BP 120–139/80–89 mmHg) as defined by the Joint National Committee 7 (11) and combat-related PTSD were enrolled over a period of 4 yr. All participants had a diagnosis of PTSD in their medical record. Exclusion criteria included pregnancy, hypertension, diabetes, heart disease, illicit drug use, alcohol use >2 drinks/day, hyperlipidemia, autonomic dysfunction, treatment with monoamine oxidase inhibitors, and any serious systemic disease. Patients treated with medications known to affect SNS activity including clonidine, β-blockers, angiotensin-converting enzyme inhibitors, and angiotensin receptor blockers were excluded, with the exception of selective serotonin reuptake inhibitors (SSRIs) and the alpha blocker prazosin for treatment of PTSD symptoms. Twenty-nine participants with PTSD entered the trial, but only 25 finished the trial; only data from those that completed the trial are presented in the current report. This study was approved by the Emory Institutional Review Board and the Atlanta Veterans Affairs Medical Center Research and Development Committee. All participants signed an approved informed consent document before study procedures.

Measurements

BP, HR, and respiratory rate.

All participants presented for one or two screening visits before entering the trial. During the screening visit(s), three seated BPs were taken using the standard ACC/AHA guideline technique (55) with an appropriately sized cuff placed on the upper arm, with the arm resting at heart level after at least 5 min of quiet rest, using an automated digital BP device (Omron, HEM-907XL; Omron Healthcare, Kyoto, Japan). Female participants underwent a urine pregnancy test to rule out pregnancy. During the experimental protocol, beat-to-beat arterial BP was measured continuously and noninvasively using digital pulse photoplethysmography (CNAP Monitor 500; CNSystems, Graz, Austria) as previously described (18). HR was measured using continuous electrocardiography (ECG) using a Bio Amp (model ML 132; ADInstruments, Colorado Springs, CO). Respiratory rate (RR) was continuously monitored via a respiratory belt pressure transducer placed around the upper abdomen (Pneumotrace II; ADInstruments).

Muscle sympathetic nerve activity.

Multiunit postganglionic sympathetic nerve activity directed to muscle (MSNA) was recorded from the peroneal nerve using microneurography (53). A tungsten microelectrode (tip diameter: 5–15 μm; Bioengineering, University of Iowa, Iowa City, IA) was inserted into the nerve, and a reference microelectrode was inserted in close proximity. The efferent nerve signals were amplified (total gain: 50,000–100,000), filtered (700–2,000 Hz), rectified, and integrated (time constant: 0.1 s) to obtain a mean voltage display of sympathetic nerve activity (nerve traffic analyzer, model 662C-4; Bioengineering, University of Iowa). The neurogram was recorded using LabChart 7 (PowerLab 16sp; AD Instruments, Sydney, NSW, Australia) along with concomitant ECG, BP, and RR. All MSNA recordings met previously established criteria (35) and were analyzed by a single investigator who was blinded to the device used by the participant during the trial. MSNA was ascertained by verifying entrainment to R waves, a 3:1 burst-to-baseline ratio and activity increasing with apnea and Valsalva. MSNA is expressed as burst frequency (bursts/min) and burst incidence (bursts/100 heart beats).

Arterial BRS testing using the modified Oxford technique.

The gold standard method for evaluation of arterial BRS is performed by measuring changes in MSNA and R-R interval during arterial BP changes induced by nitroprusside and phenylephrine (43). Sodium nitroprusside (100 μg in 10 mL of normal saline) was bolused through an antecubital intravenous catheter, followed 60 s later by an intravenous bolus of phenylephrine (150 μg in 10 mL of normal saline) during continuous MSNA, ECG, and hemodynamic monitoring. Medications were room temperature at the time of administration. These medications induce a decrease of ∼15 mmHg followed by an increase of ∼15 mmHg in arterial BP (5). Sympathetic BRS was defined as the slope of the linear regression between MSNA burst incidence and diastolic BP (DBP) binned in 3-mmHg pressure ranges (13). A steeper slope represents a greater increase in MSNA in response to low BP and a greater suppression of MSNA in response to high BP, i.e., increased sympathetic BRS, whereas a flatter slope represents decreased sympathetic BRS. Cardiovagal BRS was assessed as the slope of the linear regression between ECG R-R intervals and systolic BP (SBP; binned over 2-mmHg pressure ranges) (39). Similarly, a steeper slope represents increased cardiovagal BRS, whereas a flatter slope indicates blunted cardiovagal BRS. At least 2 min of total data were analyzed for each BRS assessment (sympathetic and cardiovagal).

Mental stress via mental math.

Participants were asked to serially subtract a one- or two-digit number from a three- or four-digit number and were urged to do so as quickly and accurately as possible, for a duration of 3 min. An investigator used flash cards with a three- or four-digit number and were urged by two additional study team members in white coats to answer “faster” and “get it right.” Importantly, the same investigator administered the mental stress at baseline and at the end of study within subjects. At the end of mental stress, participants were asked to rate their perceived stress during the mental math trial using a standard five-point scale: 0, not stressful; 1, somewhat stressful; 2, stressful; 3, very stressful; and 4, very, very stressful. This test has been shown to induce mental stress and increase MSNA, BP, and HR in PTSD (38), especially in those with elevated resting BP (18).

Device-guided breathing.

The participants performed guided breathing daily in a seated position at home for a duration of 8 wk using the RESPeRATE (InterCure) device (DGB group) or an identically appearing sham device (Sham group) for 15 min daily after an initial orientation to the device and after baseline measurements in the laboratory. The device is composed of an elastic belt with a respiration sensor placed around the upper abdomen for biofeedback and earbuds used for auditory guidance (49). The DGB device monitors respiratory rate, calculates inspiration and expiration times, and generates a personalized melody of two distinct ascending and descending tones for inhalation and exhalation. Participants effortlessly entrain their breathing pattern with the tones, and the device gradually guides the user to a prolonged expiration time and subphysiological respiratory rate of ~5 breaths/min. The sham device is identical to the DGB device, using the same display, musical tones, and respiratory sensor belt. However, the sham device guides respiratory rates at a physiological rate of 14 breaths/min. The device automatically stores usage data, allowing for quantification of adherence and performance. Participants were randomized to the DGB device (n = 12) or the sham device (n = 13). Participants were advised to use the device at the same time of the day. In addition, the pretrial and posttrial study visits were performed at the same time of the day.

Clinician Administered PTSD Scale and PTSD Checklist–Military.

The diagnosis of PTSD was confirmed in all participants during the screening visit using validated and reliable measures of PTSD: the Clinician-Administered PTSD Scale (CAPS)‐IV for Diagnostic Statistical Manual for Mental Disorders, fourth edition (DSM-IV) (3) and/or the PTSD Checklist–Military version (PCL-M) (4). PCL-M is a validated and reliable self-report measure of PTSD symptoms (6). Both the CAPS-IV and PCL-M have excellent psychometric properties with high interrater reliability, test-retest reliability, and internal consistency (3, 4). The PCL-M and the CAPS scores have been shown to be highly correlated (r >0.80) (4, 16). CAPS-IV was administered by a single trained investigator specifically for the current study. Participants required a CAPS-IV and/or PCL-M severity score of ≥45 to confirm the presence of PTSD (54).

Experimental Protocol

On the day of both experimental sessions (pre and post the 8-wk trial), participants reported to the laboratory in the morning after abstaining from food, medications, caffeine, and alcohol for at least 12 h and from exercise for at least 24 h. The study room was quiet, semidark, and at a temperature of ∼21°C. Participants were placed in a supine position on a comfortable stretcher. A 20-gauge intravenous catheter was placed into the antecubital vein of the arm for the administration of medications during BRS assessments. Finger cuffs were fitted and placed on the fingers of the dominant arm for continuous beat-to-beat arterial BP measurements, and an upper arm cuff was placed for intermittent automatic calibrations with the finger cuffs. ECG patch electrodes were placed for continuous ECG recordings, and two belts with respiratory rate sensors were placed around the upper abdomen for monitoring of continuous respiratory rates and for biofeedback as part of the breathing intervention. The leg was positioned for microneurography, and the tungsten microelectrode was inserted and manipulated to obtain a satisfactory nerve recording. After 10 min of nonrecorded rest following microneurography, baseline BP, HR, and MSNA were recorded continuously for 10 min. After baseline measurements, arterial BRS testing was performed once by infusing intravenous boluses of 100 μg of nitroprusside followed 60 s later by 150 μg of phenylephrine, as described in Measurements. After 30 min of rest to ensure sufficient washout of BRS drugs and return to baseline conditions, participants underwent 3 min of mental math. This experiment was repeated 8 wk later after the participants completed the at-home breathing trial with either DGB or Sham. The breathing trial consisted of 15 min daily breathing exercise for 8 wk as used in previous studies that showed sustained improvements in BP with a similar regimen (12, 25) Participants were seen for interim visits every 2 wk to assess adherence by monitoring daily logs and usage times recorded on the device. Figure 1 gives the overview of the experimental protocol.

Fig. 1.

Summary of the experimental protocol. Pretrial and posttrial sessions included measurements of continuous blood pressure (BP), heart rate (HR), ECG, muscle sympathetic nerve activity (MSNA), and respiratory rate (RR) at rest, during baroreflex sensitivity (BRS) testing, and during mental math for 3 min. Pretrial studies were performed at baseline before randomization to device-guided slow breathing (DGB) vs. sham breathing daily for 15 min for a duration of 8 wk. Posttrial studies were performed at the end of the 8-wk trial period.

Data Analysis

Hemodynamics, MSNA, and BRS.

For all baseline analyses, we used the last 5 min of the 10-min baseline recording to ensure true resting state. MSNA, BP, and ECG data were exported from LabChart to WinCPRS (Absolute Aliens, Turku, Finland) for analysis as previously described (38). The continuous arterial BP waveforms were analyzed for beat-to-beat changes in SBP, DBP, and mean arterial pressure. R waves were detected and marked from the continuous ECG recording. MSNA bursts were automatically detected by the program using the following criteria: burst-to-noise ratio of 3:1 within a 0.5-s search window with an average latency of 1.2–1.3 s in burst occurrence from the previous R wave. After automatic detection, the ECG and MSNA neurograms were visually inspected for accuracy of detection by an investigator without knowledge of the group (DGB vs. Sham). MSNA was expressed as burst frequency (bursts per minute) and burst incidence (bursts per 100 heartbeats). Sympathetic BRS was quantified as the slope of the linear relationship between MSNA burst incidence and diastolic BP (DBP) during pharmacological manipulation of BP. Cardiovagal BRS was quantified as the slope of the linear relationship between R-R interval and systolic blood pressure (SBP). Only slope values with a correlation value >0.5 were included. At least 2 min of data, including both the decreases and increases in BP induced pharmacologically, were analyzed for each overall cardiovagal and sympathetic BRS assessment. Separate BRS analyses were also performed analyzing the change in MSNA and R-R interval during the decreasing BP phase (down) versus during the increasing BP phase (up) of baroreflex testing.

Heart rate variability analysis.

Heart rate variability (HRV) was quantified in the time domain as standard deviation of the NN (R-R) intervals (SDNN) and root mean square of the successive differences (RMSSD). HRV was analyzed in a 5-min segment at baseline, similarly to MSNA. During mental stress, HRV was analyzed in 1-min segments, including the 1 min before onset of the mental stress that was considered time 0.

Statistical Analysis

Data were analyzed using SPSS 22.0 (IBM SPSS, Armonk, NY) and the R Language and Environment for Statistical Computing (40a). A χ² test for independence was used to compare categorical variables [race, sex, smoking, selective serotonin reuptake inhibitor (SSRI) use], and an independent t test was used to compare continuous variables at baseline (pretrial DGB vs. pretrial Sham), including age, body mass index (BMI), resting hemodynamics, MSNA, HRV, and BRS. T tests were also used to compare the rate of the participants’ perceived mental math stress at baseline and pre- and post-breathing trial for both groups. Significantly different variables at baseline were accounted for in the analysis of the response to mental stress. Repeated measures ANOVA was used to compare the changes in baseline variables from pretrial to posttrial in both device groups (DGB vs. Sham). The response during 3 min of mental arithmetic was calculated as a mean reactivity for each 1-min block of the test period in two different conditions (pretrial and posttrial) for each device (DGB and Sham). A linear mixed model (LMM) analysis was conducted to assess the between-group differences and within-group difference (2). LMM analysis accounts for missing values and improves power when compared with the 2-way ANOVA for these analyses. The full description of the LMM model is described at the link provided in the appendix. Cardiovascular and MSNA data reported are paired (pre- and posttrial). As reported in Table 1 and Figs. 2–4, the term “condition” refers to the pre- and post-8-wk trial measurements, the term “device” refers to DGB and Sham groups, and “time” refers to the minute-by-minute response to mental stress. All P values are two tailed, calculated based on the initial hypothesis for different responses. P values <0.05 were considered significant.

Table 1.

Baseline characteristics pre- and postbreathing trial

Variable	DGB		Sham		P Value (Pre)	P Value (Pre vs. Post)
	Pre	Post	Pre	Post		Condition	Device	Device × Condition
Age, yr	36 ± 8		34 ± 6		0.434
BMI, kg/m2	30 ± 6		28 ± 7		0.491
Race (B/W)	6/6		12/1		0.030*
Sex (F/M)	2/10		4/9		0.645
CAPS-IV score	69 ± 23	67 ± 13	64 ± 23	52 ± 28	0.648	0.092	0.143	0.145
PCL-M score	67 ± 15	55 ± 24	57 ± 20	57 ± 19	0.297	0.204	0.790	0.493
Smoking (Yes/No)	7/5		9/4		0.688
Prazosin (Yes/No)	5/7		1/12		0.073
SSRI (Yes/No)	6/6		10/3		0.226
Depression (Yes/No)	6/6		8/5		0.695
SBP, mmHg	121 ± 11	121 ± 12	122 ± 11	118 ± 12	0.859	0.234	0.777	0.288
DBP, mmHg	78 ± 11	80 ± 13	78 ± 9	78 ± 12	0.980	0.521	0.806	0.548
PP, mmHg	44 ± 8	41 ± 9	44 ± 7	40 ± 8	0.815	0.046*	0.947	0.590
HR, beats/min	67 ± 13	70 ± 10	71 ± 17	69 ± 10	0.485	0.942	0.725	0.256
RR, breaths/min	18 ± 3	16 ± 3	17 ± 2	15 ± 4	0.192	0.004*	0.337	0.727
MSNA, bursts/min	22 ± 14	21 ± 10	19 ± 13	17 ± 13	0.942	0.666	0.365	0.924
MSNA, bursts/100 heartbeats	32 ± 21	30 ± 19	28 ± 17	25 ± 20	0.789	0.625	0.475	0.921
SDNN, ms	72 ± 22	71 ± 26	65 ± 28	68 ± 12	0.523	0.835	0.508	0.745
RMSSD, ms	63 ± 41	44 ± 24	59 ± 44	50 ± 24	0.816	0.097	0.921	0.545
CVBRS, ms/mmHg	11 ± 7	14 ± 8	14 ± 8	15 ± 12	0.243	0.417	0.378	0.663
SBRS, bursts·100 beats−1·mmHg−1	−2.2 ± 1.5	−1.2 ± 1.2	−0.8 ± 1.1	−0.5 ± 4.2	0.022*	0.369	0.212	0.532

Data are means ± SD measured before (pre) and after (post) use of device-guided slow breathing (DGB) or sham device (Sham) daily for 8 wk; n = 12 subjects in DGB group [5 moderate and 7 severe posttraumatic stress disorder (PTSD)] and n = 13 subjects in Sham group (7 moderate and 6 severe PTSD). Sixteen subjects have Clinician-Administered PTSD Scale (CAPS) pre (7 DGB and 9 Sham), and 12 subjects have CAPS post (5 DGB and 7 Sham). Eighteen subjects have PTSD Checklist for the military (PCL-M) pre (7 DGB and 11 Sham), and 16 subjects have PCL-M post (6 DGB and 10 Sham). B, black; BMI, body mass index; CVBRS, cardiovagal baroreflex sensitivity; DBP, diastolic arterial pressure; F, female; HR, heart rate; M, male; MSNA, muscle sympathetic nerve activity (n = 12 DGB and n = 11 Sham); RMSSD, root mean square of the successive differences; RR, respiratory rate; SBP, systolic arterial pressure; SBRS, sympathetic baroreflex sensitivity; SDNN, standard deviation of NN intervals; SSRI, selective serotonin reuptake inhibitor; W, white. Condition, pre- and posttrial; Device, DGB or Sham; P value (pre), t test comparison of DGB pretrial vs. Sham pretrial; P value (pre vs. post), repeated measures ANOVA comparison between devices and across trials.

^*P < 0.05 (two-tailed).

Fig. 2.

Change in systolic blood pressure (SBP; A) and pulse pressure (PP; B) during 3 min of mental math before (pre) and after (post) use of device-guided slow breathing (DGB) device or sham device (Sham) daily for 8 wk. Pretrial responses to math were comparable between DGB and Sham for SBP (time × device, P = 0.750) and PP (time × device, P = 0.415). During mental math, SBP (time × condition × device, P = 0.200) and PP (time × condition × device, P = 0.121) responses to mental math with the long-term use of DGB compared with Sham were comparable. Data are means ± SD; n = 12 participants for the DGB group and n = 13 participants for the Sham group. Additional P values: SBP (time, P = 0.026; condition, P = 0.096; device, P = 0.027; time × condition, P = 0.403) and PP (time, P = 0.677; condition, P = 0.253; device, P = 0.126; time × condition, P = 0.831). P values are two-tailed.

Fig. 4.

Change in muscle sympathetic nerve activity (MSNA) quantified as burst frequency (A) and burst incidence (B) during 3 min of mental math before (pre) and after (post) use of device-guided slow breathing (DGB) device or sham device (Sham) daily for 8 wk. Pretrial responses to math were comparable between DGB and Sham for MSNA frequency (time × device, P = 0.154) and MSNA incidence (time × device, P = 0.150). There was a significant difference in the MSNA reactivity to mental math after 8 wk of daily DGB compared with Sham when expressed as burst frequency (time × condition × device, P = 0.012) or burst incidence (time × condition × device, P = 0.008). Data are means ± SD; n = 12 participants for the DGB group and n = 11 participants for the Sham group. Additional P values: MSNA burst frequency (bursts/min; time, P < 0.001; condition, P = 0.550; device, P = 0.907; time × condition, P = 0.141) and burst incidence [bursts/100 heartbeats (hb)]; time, P < 0.001; condition, P = 0.351; device, P = 0.710; time × condition, P = 0.344]. P values are two-tailed and significant at <0.05.

RESULTS

Baseline Characteristics

A total of 44 post-9/11 veterans with prehypertension and PTSD were enrolled into the study. An excellent MSNA neurogram preintervention was unable to be obtained in 15 participants; therefore, 29 participants were then randomized into the trial after the initial study visit. Four participants dropped out of the trial due to time commitment (n = 1), inability or unwillingness to do the breathing treatment daily (n = 2), and other personal issues (n = 1). Data from the remaining 25 participants randomized to the DGB (n = 12) and Sham (n = 13) groups are presented.

Baseline pretrial.

The group randomized to Sham and the group randomized to DGB were well matched at screening, except for race. Age, BMI, sex, smoking status, and selective serotonin reuptake inhibitor (SSRI) use were not significantly different between the groups [Table 1, P value (Pre)]. The majority of participants were African-American men. However, there was a higher proportion (P = 0.030) of African-Americans in the Sham group compared with the DGB group. CAPS-IV and PCL-M scores were comparable between groups (Table 1). Resting SBP, DBP, pulse pressure (PP), HR, and RR were not significantly different between DGB and Sham groups. There was no difference in baseline MSNA between groups when expressed as burst frequency (22 ± 14 vs. 19 ± 13 bursts/min, P = 0.942) and burst incidence (32 ± 21 vs. 28 ± 17 bursts/100 heartbeats, P = 0.789). HRV measures (SDNN and RMSSD) were also not different between DGB and Sham groups at baseline. During BRS testing we found higher sympathetic BRS in the DGB group (−2.2 ± 1.5 vs. −0.8 ± 1.1 bursts·100 beats⁻¹·mmHg⁻¹, P = 0.022) compared with the Sham group at baseline [see Table 1, P value (Pre)]. However, baseline resting cardiovagal BRS (11 ± 7 vs. 14 ± 8 ms/mmHg, P = 0.243) was not different between the groups. The baseline differences in race and sympathetic BRS were controlled for in the linear mixed model analysis.

Effect of Breathing Intervention on Resting Hemodynamics, MSNA, BRS, and PTSD Symptoms

When pretrial resting hemodynamics, MSNA, HRV, and BRS were compared with posttrial resting measures respectively, no significant differences were observed between participants randomized to DGB vs. Sham (P > 0.05; Table 1, device × condition). There was no significant difference in the change in sympathetic BRS (condition, P = 0.369) or cardiovagal BRS (condition, P = 0.417) after intervention between groups. There was no improvement in PTSD symptoms as assessed by the CAPS-IV or PCL-M scores in either group. The results remained unchanged when controlled for race. The MSNA results in Table 1 are given for a subset of 12 DGB and 11 Sham participants who completed the trial in whom we were able to obtain adequate MSNA data. Of note, there was no difference in treatment adherence between the groups (DGB, 94.9% vs. Sham, 90.1%, P = 0.107).

Effect of Breathing Intervention on Hemodynamic Responses to Mental Math

These experiments aimed to determine whether 8 wk of daily breathing with DGB compared with sham device improves BP, HR, and HRV responses during acute mental stress. Race was controlled for by including it as a covariate in these analyses. All of the following results depicts responses to 3 min of mental math (time effect), pre- and posttrial (condition effect), and between DGB and sham (device effect). Rate of perceived stress (RPS) was comparable (condition, P = 0.206; device, P = 0.269; condition × device, P = 0.664) across trials between the DGB group (3.4 ± 0.9 vs. 3.0 ± 1.0) and the Sham group (2.7 ± 1.1 vs. 2.5 ± 1.1). Figure 2 depicts the SBP and pulse pressure (PP) responses to mental math. When between-group differences in acute mental stress reactivity were examined, the change in SBP was not different between the groups (time × condition × device; P = 0.200; Fig. 2A). Likewise, DBP responses to mental math (time × condition × device; P = 0.441; result not graphed) and PP responses (time × condition × device; P = 0.121; Fig. 2B) pretrial and posttrial were also comparable between the groups.

As shown in Fig. 3, pretrial and posttrial HR (time × condition × device; P = 0.257; Fig. 3A) and HRV (SDNN: time × condition × device, P = 0.814; Fig. 3B) responses to mental math were not different between the DGB and sham device groups. Likewise, pre- and posttrial RMSSD reactivity were comparable between the groups (time × condition × device, P = 0.895; result not graphed).

Fig. 3.

Change in heart rate (HR; A) and standard deviation of NN intervals (SDNN; B) during 3 min of mental math before (pre) and after (post) use of device-guided slow breathing (DGB) device or sham device (Sham) daily for 8 wk. Pretrial responses to math were comparable between DGB and Sham for HR (time × device, P = 0.244) and SDNN (time × device, P = 0.152). During mental math, HR (time × condition × device, P = 0.257) and SDNN (time × condition × device, P = 0.814) responses to mental math were comparable in both groups. Data are means ± SD; n = 12 participants for the DGB group and n = 13 participants for the Sham group. Additional P values: HR (time, P = 0.140; condition, P = 0.862; device, P = 0.539; time × condition, P = 0.097) and SDNN (time, P < 0.001; condition, P = 0.652; device, P = 0.694; time × condition, P = 0.745). P values are two-tailed.

The hemodynamic and HRV responses to mental math were not significantly altered when the results were adjusted for race and sympathetic BRS. BP (SBP: time × condition × device, P = 0.284; DBP: time × condition × device, P = 0.481; PP: time × condition × device, P = 0.269), HR (time × condition × device, P = 0.308), and HRV (SDNN: time × condition × device, P = 0.914; RMSSD: time × condition × device, P = 0.944) responses remained comparable between the groups.

Effect of Breathing Trial on Sympathetic Responses to Mental Math

We also examined the impact of 8 wk of daily DGB versus sham breathing on MSNA reactivity to acute mental stress. There was a significant difference in the MSNA response during mental math between the groups, when expressed as burst frequency (bursts/min: time × condition × device, P = 0.012; Fig. 4A). The results remained similar when expressed as MSNA burst incidence (bursts/100 heartbeats: time × condition × device, P = 0.008; Fig. 4B). While there was no change in the MSNA response to 3 min of mental math within the Sham group (bursts/100 heartbeats: time × condition, P = 0.816), the DGB group demonstrated a blunted MSNA reactivity (bursts/100 heartbeats: time × condition, P = 0.031) to mental math posttrial compared with pretrial. The sympathetic responses to mental math were not significantly altered when the results were adjusted for race and sympathetic BRS. MSNA (bursts/min: time × condition × device, P = 0.012; bursts/100 heartbeats: time × condition × device, P = 0.009) responses remained blunted in the DGB group compared with the Sham group.

It is important to note that pretrial, MSNA burst frequency (time × device, P = 0.154; Fig. 4A) and incidence (time × device, P = 0.150; Fig. 4B) responses to math were statistically comparable between DGB and Sham, even when controlled for race.

DISCUSSION

This is the first randomized, double-masked, sham-controlled clinical trial in PTSD aimed at investigating potential benefits with long-term daily slow breathing using DGB on BP, HR, sympathetic, and parasympathetic activity at rest and in response to acute mental stress. This longitudinal study is a follow-up to the cross-sectional study in which we showed acute reductions in BP and MSNA and increased sympathetic BRS during a single session of DGB in PTSD patients (17).

The major findings of this study are as follows: 1) 8 wk of daily DGB lowered MSNA reactivity to acute mental stress in veterans with PTSD; 2) resting BP, HR, respiratory rate, MSNA, and HRV at rest were unchanged after 8 wk of DGB; and 3) there was no change in cardiovagal or sympathetic BRS following 8 wk of DGB. These results suggest that DGB may have long-term beneficial effects on sympathetic reactivity to acute mental stress in patients with PTSD.

We previously showed that slow breathing using DGB acutely lowers BP and MSNA and increases HRV and sympathetic BRS in prehypertensive patients with PTSD (17). The purported mechanisms by which slow breathing exerts its physiological effects is via activation of pulmonary stretch receptors due to increased tidal volumes and modulation of the baroreflex leading to reduced SNS activation and subsequent reductions in BP (47). In the previous study, we showed that DGB in PTSD patients (17) lowers BP and MSNA and increases HRV and sympathetic BRS acutely while the participant is actively slow breathing, compared with measures taken during breathing at normal respiratory rates with the sham device. However, the current study shows that 15 min of daily DGB for 8 wk did not improve baseline hemodynamics, lower respiratory rate, or improve autonomic function in PTSD measured at rest, similar to the findings of de Barros et al. (12) in hypertensive patients and results from a recent meta-analysis (32). The decrease in MSNA and improvements in sympathetic BRS during acute DGB that our laboratory (17) and others (33, 37) have reported does not appear to translate into lower steady-state MSNA when participants are not actively using the DGB device.

Our current findings contrast with some prior trials demonstrating that months of daily voluntary slow breathing exercises can reduce resting SBP, DBP, and HR in patients with hypertension (23, 25, 44) and chronic heart failure (30). Therefore, DGB may have beneficial effects on resting hemodynamics in some patient populations, but not in others. It is important to note that, unlike in the current study, many of the prior studies used music or spontaneous breathing without the use of a sham device for the control intervention (12, 25, 31, 34). A strength of the present study is the use of an identically appearing sham device with the same biofeedback functionality, which allows for masking and a better control intervention compared with calm music (12) or spontaneous breathing (25) as used in prior studies. The sham device guided breathing to a rate of 14 breaths/min. This rate was slightly lower than the participants’ spontaneous normal breathing rate (which ranged between 16 and 19 breaths/min) and ensured that the sham group did not hyperventilate or breathe above their physiological respiratory rates. Importantly, the comparable resting respiratory rate pre and post 8 wk of DGB highlights the fact that the intervention did not lower the overall resting RR of our PTSD participants. Therefore, it is possible that the physiological effects of DGB are only observed when individuals are actively slow breathing and reversed once the participants return to their natural breathing rate. Together, these findings suggest that for a sustained decrease in MSNA, a long-term subphysiological RR might be required, and perhaps interventions that entrain participants to lower their resting RR throughout the day may have more beneficial effects on autonomic regulation.

PTSD is also associated with exaggerated neurocardiovascular reactivity during stress, an independent risk factor for hypertension and cardiovascular disease (10, 20, 27) in this population. Only a few studies (25, 30) have explored the effects of long-term slow breathing on cardiovascular reactivity to acute mental stress. Contrary to our hypothesis, long-term daily DGB did not lower BP or HR reactivity to mental stress in veterans with PTSD. These results contrast with a recent study from Lachowska et al. (30) reporting that 12 wk of slow breathing significantly reduced BP and HR responses to mental stress in 21 patients with heart failure. Similarly, Hering et al. (25) found reduced SBP and HR responses to mental stress in 10 hypertensive subjects vs. 12 controls. It is important to note that these tests were conducted in a diseased population and did not include a control group or sham intervention. It is possible that the observed decreases in cardiovascular reactivity after slow breathing might be apparent only in patients with established hypertension or heart disease.

Interestingly, in contrast to BP reactivity, we found that in prehypertensive participants with PTSD, 8 wk of daily DGB significantly reduced MSNA reactivity to mental stress induced via mental math. The mean MSNA reactivity to mental math in both PTSD groups (DGB and Sham) ranged from ~7 to 13 bursts/min in the current study, which is lower than previously reported data from our laboratory (38). However, the current study includes a larger sample size, and MSNA response to math is very variable between individuals (9), despite comparable resting MSNA. Therefore, more variability is expected with a larger sample, despite comparable resting MSNA. Our observation that DGB and not Sham lowers MSNA reactivity during mental math suggests that long-term daily slow breathing may lead to physiological changes that lead to reduced neural sympathetic reactivity during stress even when the individual is not actively slow breathing and that this reduction in SNS reactivity occurs despite no significant change in BRS. The only other study to explore SNS reactivity to acute stress with slow breathing found that in hypertensive patients, long-term slow breathing did not affect SNS reactivity, despite a dampened BP reactivity (25). Sympathoinhibition during mental stress after long-term DGB in our PTSD cohort was not accompanied by an inhibited pressor response, suggesting a dissociation between the BP and MSNA responses. This dissociation between MSNA and BP reactivity to mental stress has previously been described (9, 19) and indicates that other mechanisms may be involved in the integrated pressor response associated with mental stress in PTSD, such as elaboration of angiotensin II or nitric oxide that were unmeasured in the currents study, but may not be affected by long-term DGB. In addition, PTSD is characterized by increased inflammation (16, 38, 40) that may contribute to heightened SNS reactivity during mental stress; whether DGB modulates inflammation in PTSD should be investigated in future studies. Finally, although we observed an amelioration in the SNS response to mental stress after 8 wk of DGB, we did not observe a change in the RMSSD and SDNN response, suggesting that long-term DGB does not modulate parasympathetic reactivity to mental stress in PTSD. At the onset of mental stress, parasympathetic nervous system (PNS) withdrawal is the initial driver of the increase in HR and subsequent increase in BP, while SNS activity kicks in later to drive the subsequent peak in blood pressure (41, 42, 56). The initial PNS withdrawal at the onset of mental stress contributes to the HR and overall BP reactivity to mental stress; therefore, the lack of change in the HRV response may also contribute to lack of change in HR and BP responses despite a reduction in SNS reactivity and the dissociation between the MSNA and BP reactivity during mental stress after long-term DGB.

Limitations

We recognize several limitations to our study. First, the study population was mainly composed of men and African-American veterans with PTSD; therefore, the results may not be generalizable to women, other racial groups, or the non-veteran population. In addition, we did not stratify randomization by race and there were more Black participants randomized to Sham versus DGB; however, adjusting for race in the analyses did not change the results. Second, menstrual cycles were not controlled for in female participants, therefore, ovarian steroid hormones might have differentially influenced the outcomes in women. Third, we did not measure tidal volume or end-tidal CO₂ in this study, which could have added insights into the possible mechanisms contributing to MSNA modulation with DGB. Fourth, only SNS activity directed to the muscle (MSNA) was measured during each intervention. The effects of DGB on sympathetic innervation to other organs such as the heart and kidneys remain unknown. Fifth, 12-h abstention from medications may not have been sufficient to eliminate the effect of SSRIs on MSNA. However, we did not observe a sympatholytic effect of SSRIs in our cohort. Finally, although the overall adherence was excellent, there might have been some variability in the use of the device at home (time of day, depth of breaths) by the participants.

Perspectives and Significance

In summary, 8 wk of DGB lowered sympathetic reactivity to mental stress in PTSD without changes in BP reactivity or resting MSNA, BP, HR or BRS. These findings advance our understanding of the complex relationships between MSNA and hemodynamic responsiveness to acute mental stress and the potential beneficial effect of slow breathing on sympathetic reactivity in PTSD. More studies to assess the long-term benefits of DGB on autonomic reactivity and cardiovascular risk in PTSD are needed.

GRANTS

This work was supported by United States Department of Veterans Affairs Clinical Sciences Research and Development Program Merit Review Award I01CX001065; American Heart Association National Affiliate, Collaborative Sciences Award 15CSA24340001; National Heart Lung Blood Institute Grant R01 HL135183; National Center for Complementary and Integrative Health Grant R61 AT010457; National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health (NIH) under Award Numbers UL1TR002378 and KL2TR002381; NIH Training Grant T32 DK-00756; Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development and the Clinical Studies Center of the Atlanta Veterans Affairs Health Care System, Decatur, Georgia; and Foundation for Atlanta Veterans Education and Research (FAVER).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

J.P. conceived and designed research; I.T.F., T.J., M.V., and J.P. performed experiments; I.T.F., Y.H., T.J., M.V., and J.P. analyzed data; I.T.F. and J.P. interpreted results of experiments; I.T.F. and Y.H. prepared figures; I.T.F. drafted manuscript; I.T.F., Y.H., T.J., M.V., J.D.S., B.R., and J.P. edited and revised manuscript; I.T.F., Y.H., T.J., M.V., J.D.S., B.R., and J.P. approved final version of manuscript.

APPENDIX

Linear mixed model analysis.

Estimation and inference of linear mixed model (LMM) was performed with the lme4 (2) and lmerTest (29a) packages in the R statistical programming environment (51). LMM was used to estimate the within-group difference after long-term intervention, e.g., “slopes” difference in device-guided slow breathing (DGB) and Sham groups, which indicates the difference in rates of change of the long-term intervention for different devices. Moreover, it was used to produce an estimate of between-group difference in “slopes” difference, which was our primary interest. The fixed effect model of the LMM can be specified as follows:

$$E\left(Y_{ijk}\right)=b_0+b_1\times \text{condition}+b_2\times \text{condition}\times \text{device}+\left(b_3+b_4\times \text{condition}+b_5\times \text{condition}\times \text{device}\right)\times t_{ij}$$ (1)

where i is the index for observations, j is the index of conditions (0 for Pre and 1 for Post), and k is the index of subjects. t_ij is the time of the i th observation during the intervention (time = breathing time elapsed) at j th condition, and E(Y_ijk) is the expected value of observation at the i th level of time at j th condition of k th subject. The binary variable condition represents different conditions: condition = 0 for any observation in the Pre, and condition = 1 for those in the Post. The binary variable device represents the device group: device = 0 for any subject in the Sham group and device = 1 for those in the DGB group. In Eq. 1, coefficient b₃ can be interpreted as the rate of change over time before the intervention, coefficient b₄ can be interpreted as the Pre versus Post difference of rate of change over time in the Sham group, and coefficient b₅ can be interpreted as an estimator of between-group difference in “slopes” difference of two groups. In other words, b₃ would be the rate of change over time for all subjects before the intervention, b₃ + b₄ would be the rate of change over time for subjects after long-term Sham use, and b₃ + b₄ + b₅ would be the rate of change over time for subjects after long-term DGB use. When modeling the observed data Y_ijk, subject-condition-specific zero-mean Gaussian random variables were added to the fixed effect model where z_0jk served as random intercept and z_1jk served as random slope, so that we would control for individual variability in different conditions. Random error ϵ_ijk was also added, which follows a zero-mean Gaussian distribution. We assumed that all random effects and random errors were jointly independent. In general, our final LMM model can be written as

$$Y_{ijk}=b_0+b_1\times \text{condition}+b_2\times \text{condition}\times \text{device}+z_{0jk}+\left(b_3+b_4\times \text{condition}+b_5\times \text{condition}\times \text{device}+z_{1jk}\right)\times t_{ij}+{ϵ}_{ijk}$$ (2)

Coefficient b₄ and coefficient b₅ can be interpreted in the same way as Eq. 1. For each response, we compared random intercept model (only z_0jk) and random slope model (both z_0jk and z_1jk), and we chose the model with smaller Bayesian information criterion (BIC) (45a), a commonly used model selection criterion. T test statistics with a Satterthwaite approximation to degrees of freedom were used to evaluate the statistical significance in the LMM analysis, with the significance level set at α = 0.05.