Journal of Psychosomatic Research 148 (2021) 110569

Contents lists available at ScienceDirect

Journal of Psychosomatic Research

journal homepage: www.elsevier.com/locate/jpsychores

Investigating components of pranayama for effects on heart

rate variability☆
Erica Sharpe a, b, Alison Lacombe a, c, Adam Sadowski a, John Phipps a, Ryan Heer a,
Savita Rajurkar a, Douglas Hanes a, Ripu D. Jindal d, Ryan Bradley a, e, f, *
a
National University of Natural Medicine, Portland, OR, United States of America
b
State University of New York at Canton, Canton, NY, United States of America
c
United States Department of Agriculture, Produce Safety Microbiology Research Unit, Albany, CA, United States of America
d
Birmingham VA Medical Center, University of Alabama at Birmingham, Birmingham, AL, United States of America
e
University of California, San Diego, La Jolla, CA, United States of America
f
Australian Research Centre in Complementary and Integrative Medicine, Faculty of Health, University of Technology Sydney, Ultimo, NSW, Australia

A R T I C L E I N F O A B S T R A C T

Keywords: Objective: Traditional Indian breath control practices of Pranayama have been shown to increase indices of heart
Autonomic nervous system (ANS) rate variability (HRV) that are generally held to reflect parasympathetic nervous system (PNS) tone. To our
Blood pressure knowledge, individual components of pranayama have not been separately evaluated for impact on HRV. The
Heart rate variability (HRV)
objective of this study was to isolate five components of a pranayama practice and evaluate their impact on HRV.
Pranayama
Respiratory rate
Methods: In a crossover clinical trial, 46 healthy adults were allocated to complete five activities in random order,
Sheetali over five separate visits: 1) sitting quietly; 2) self-paced deep breathing; 3) externally-paced deep breathing; 4)
self-paced Sheetali/Sheetkari pranayama; and 5) externally paced Sheetali/Sheetkari pranayama
Results: Our final sample included 25 participants. There was a significant increase in a time-domain index of
HRV, the root mean square successive differences between RR intervals (RMSSD), during the five interventions.
The change in logRMSSD ranged from 0.2 to 0.5 (p < .01 in all conditions by paired t-test). Greater increases
were evident during externally-paced breathing than during self-paced breathing (mean pre-during logRMSSD
change of 0.50 vs. 0.36, p = .02) or sitting quietly (mean, 0.17 ms; p = .005 and 0.02 when comparing Activities
3 and 5 to Activity 1 by random intercept model with Tukey correction for multiple comparisons). Lastly, pre-
during increase in RMSSD was greater for Sheetali/Sheetkari vs. deep breathing, when controlling for respiration
rate, though not significantly different (p = .07 in random intercept model)
Conclusions: RMSSD increased with paced breathing, deep breathing, and Sheetali/Sheetkari pranayama, rein­
forcing evidence of a physiologic mechanism of pranayama.
Trial registration: NCT03280589
https://www.clinicaltrials.gov/ct2/show/NCT03280589?term=sheetali&draw=2&rank=1

1. Background conditions such as hypertension [2] and type 2 diabetes [3]. The health
benefits of these traditional Eastern practices have been posited to stem
In recent years, breathing exercises, yoga, and meditation have from regulating the autonomic nervous system [4], baroreceptor sensi­
become popular in the United States [1]. Beyond their perceived social tivity [5–7] and the hypothalamic-pituitary adrenal axis [8].
and spiritual benefits, these practices have been shown to provide at Among different indices of the parasympathetic nervous system, beat
least short-term benefit in patients with a broad spectrum of chronic to beat variability in heart rate, hereafter referred to as heart rate

Acronymns: HRV, heart rate variability; PNS, parasympathetic nervous system; HF-HRV, high frequency heart rate variability; RMSSD, root mean square of
successive differences between R-R intervals; SDNN, standard deviation of heart beat N-N intervals;; b/m, breaths per minute..
☆
This work was performed at the Helfgott Research Institute, at the National University of Natural Medicine, Portland, OR
* Corresponding author at: Helfgott Research Institute, 2220 SW 1st Ave, Portland, OR 97201, United States of America.
E-mail addresses: [email protected] (E. Sharpe), [email protected] (A. Sadowski), [email protected] (J. Phipps), [email protected] (S. Rajurkar),
[email protected] (D. Hanes), [email protected] (R.D. Jindal), [email protected] (R. Bradley).

https://doi.org/10.1016/j.jpsychores.2021.110569
Received 15 February 2021; Received in revised form 2 July 2021; Accepted 4 July 2021
Available online 8 July 2021
0022-3999/© 2021 Elsevier Inc. All rights reserved.
E. Sharpe et al. Journal of Psychosomatic Research 148 (2021) 110569

future mechanistic studies on pranayama, which in turn aid creation of

targeted interventions for chronic diseases. These findings may also help
clinicians tailor interventions in the context of unique circumstances for
their patients, providing at least short-term improvements to an in­
dividual's autonomic nervous system.
The present study builds on a previous work conducted by the au­
thors of this study [30], which evaluated a pair of pranayama practices
called Sheetali and Sheetkari [27,33] (Fig. 1), traditionally refered to as
“cooling breaths”, and used for treatment of high blood pressure. In this
randomized trial, participants with hypertension completed 20 min of
Sheetali/Sheetkari per day for one month. Results demonstrated a 16
mmHg decrease in systolic blood pressure, decreased respiration rate,
and an increase in high frequency HRV (HF-HRV).
Fig. 1. Illustration of Sheetali and Sheetkari pranayamas, involving inhaling The 16 mmHg decrease in systolic blood pressure demonstrated in
through an open mouth (rolled tongue, Sheetali; exposed teeth, Sheetkari) and our previous work is particularly noteworthy since systolic blood pres­
exhaling through the nose. sure has been shown to be a greater predictor of cardiovascular risk than
diastolic blood pressure [34]. For instance, in the Systolic Blood Pressure
Intervention Trial (SPRINT), a mean systolic blood pressure difference of
Table 1 ~15 mmHg predicted a significant reduction in cardiovascular events
Combinations of modifiable components that form the activities tested. and improved mortality. This trial randomized 9361 individuals with a
Condition/ Activity Activity 2: Activity 3: Activity 4: Activity 5: baseline systolic BP of 130–180 mmHg to standard treatment (target
Component 1: Deep Deep Sheetali Sheetali systolic BP of <140 mmHg) or intensive treatment (target systolic BP of
Sitting breathing breathing and and
<120 mmHg), and used pharmacologicals alongside lifestyle in­
quietly self-paced externally- Sheetkari Sheetkari
paced self-paced externally- terventions in their treatment protocols [35]. Thus, a non-
paced pharmacologic intervention that provides a comparable treatment ef­
Component X X X X X
fect in blood pressure reductions, if confirmed in larger phase 3 trials,
1: Sitting will have high public health importance.
quietly The present study investigated, within a healthy population, the
Component X X X X impact of five distinct components of pranayama on HRV (sitting
2: Depth of
quietly, depth of breath, pace of breath, mouth shape, and duration of
Breath
Component X X practice). We hypothesized that an index of HRV, which is known to
3: Pace of reflect PNS tone at respiratory rates less than 9 breaths per minute (b/m)
Breath [14,36] the root mean square of successive differences in adjacent
Component X X heartbeats (RMSSD), will increase during the four active components of
4: Mouth
Shape
the practice, compared to sitting quietly, and the greatest increase would
Component X X X X X be evident during the Sheetali and Sheetkari practices. The choice of
5: RMSSD rather than blood pressure as the primary outcome stems from
Duration the fact that we were investigating the impact of pranayama in real time,
of Practice
and the breathing rates during some of the components of pranayama
would vary, and be less than 9 b/m, i.e. at rates in which RMSSD is a the
variability (HRV), has attracted much systematic investigation [7,9–14]. better index of cardiac parasympathetic tone [14,37,38].
For example, several studies have provided evidence that breathing
techniques, [15–18] yoga, [15] tai chi, [19] meditation, [20] and other 2. Methods
mind-body practices [21] impact HRV. Specifically, most evidence of
breathing and meditation practices have measured acute HRV change A crossover clinical trial with the order of experimental conditions
(pre/post or during a single practice [15,16], or pre/post the interven­ assigned by randomization was conducted at the Helfgott Research
tion period (weeks or months) [17,19]. Institute at the National University of Natural Medicine (NUNM) in
Indices of HRV are mathematically calculated in frequency and time Portland, Oregon. The NUNM Institutional Review Board approved this
domains [22] and are increasingly seen as predictors of health and life trial (approval number: AL2222017) and it was registered through
expectancy. For example, an index of HRV over 24 h, the standard de­ ClinicalTrials.gov (NCT03280589). All participants provided written
viation of heart beat N-N intervals (SDNN), has been shown to predict informed consent.
mortality in individuals with sepsis [23], congestive heart failure [24]
and cirrhosis [25]. Pranayama (a distinct limb of Yoga defined as 2.1. Aims
“controlled breathing” [26,27]) and other deep breathing practices
[28,29] have been shown to lower elevated blood pressure, [6,29–31] The study aimed to (1) evaluate a proposed mechanism of pra­
and to increase the HRV indices known to reflect PNS tone [15,30]. nayama by measuring HRV during pranayama vs. sitting quietly; (2)
Enhancing PNS tone has been demonstrated to enhance baroreflex- evaluate which component(s) of Sheetali/Sheetkari pranayama most
mediated reductions in blood pressure [6,32]. strongly contribute to physiological changes, by comparing breathing
Importantly, however, breathing practices such as pranayama are interventions that include strategic combinations (Table 1) of sitting
complex and passed down from generation to generation. To our quietly, deep breathing, pacing, and facial modifying components of this
knowledge, no systematic study has tried to deconstruct the impact of pranayama practice; and (3) to investigate the relationship between
individual components of pranayama on HRV. Identifying which com­ respiratory rate and HRV.
ponents of a pranayama practice produce the greatest changes in PNS
activity (possibly resulting in greater effects on blood pressure) could 2.2. Recruitment, enrollment, and randomization
offer several advantages. Researchers can use these findings to design
Participants were recruited from the general population in the

2
E. Sharpe et al. Journal of Psychosomatic Research 148 (2021) 110569

Fig. 2. Recruitment for our study, detailing numbers assessed for eligibility and enrolled, as well as those who completed visits and were analyzed.

Portland metro area using flyers, newspaper, radio, and web-based ad­ sinus arrythmia, e.g. beta-blockers, beta-agonists, calcium channel
vertisements. Healthy adults ages 25–55, naïve or only moderately blockers, analgesics, antidepressants, anticholinergics, or anxiolytics;
exposed to yoga and relaxation practices at the time of recruitment were had smoked tobacco products in the last 6 months; or had a current
invited to participate in the trial. Moderate exposure to yoga/relaxation diagnosis of diseases (e.g. significant heart disease, hypertension, hy­
practices was defined as having no formal certification in mind/body potension) that could impact baseline HRV indices and cause HRV
practices, and not practicing yoga more than once per week. Participants indices to be less responsive to our experimental procedures than they
were additionally excluded if they were unable to roll their tongue into a would be in healthy adults. [31]
tube, as this skill is an integral part of Sheetali breathing. Participants Interested participants were screened over the phone with the help of
were asked to avoid extreme physical exercise (heart beats over 120 a standardized script, which included the Nijmegen Questionnaire to
beats/min), and to refrain from alcohol, over the counter medications, screen potential participants for hyperventilation. [39] Those who were
and any recreational drugs 24-h prior to their scheduled visit. Addi­ eligible after the phone screen were invited for a screening visit when
tionally, participants had to agree to not begin a new diet or dietary anthropometrics, blood pressure, current medication use, and current
supplement regime during the course of participation in the trial. exercise practices were recorded. Those who met the inclusion/exclu­
Participants were excluded if they had conditions that may be sion were offered recruitment after providing informed consent through
adversely impacted by deep breathing, e.g. asthma, chronic obstructive signature. Participants were then assigned a unique alpha-numeric ID,
pulmonary disease, restrictive lung disease, etc.; were currently taking and allocated to all of five activities in random order, determined
medications that are known to impact HRV measurements or respiratory through a random sequence generator, (R software®, R foundation for

3
E. Sharpe et al. Journal of Psychosomatic Research 148 (2021) 110569

Table 2 sessions with a one-minute break every five minutes to allow the
Demographic and intake information for the 25 participants in our final data set. participant to rest, drink water, adjust posture, and ask questions: ten
Total sample (N = 25) minutes of baseline before the intervention, ten minutes of breathing
intervention (normal, deep, or Sheetali), a second ten minutes of
Age
Mean (SD) 34 ± 8 breathing intervention (normal, deep, or Sheetkari), and ten minutes
Range 22–55 post-intervention.
Gender Participants were compensated a modest amount for their time and
Male 6 (24%) travel expenses.
Female 18 (72%)
Other 1 (4%)
Ethnicity 2.3.1. Description of interventions
Hispanic or Latino 1 (4%) Intervention Activities 1–5, and the components of pranayama they
Not Hispanic or Latino 23 (92%) are used to assess, are outlined in Table 1, and described in detail below.
Not Reported 1 (4%)
Activity 1: Sitting quietly (control). In this control activity, par­
Race
Black or African American 1 (4%) ticipants watched a video demonstrating a seated posture in a chair with
White, Caucasian, or European-American 21 (84%) both feet flat on the floor, hands face down on thighs, eyes closed or
Asian or Asian-American 2 (8%) slightly open, and resting the back on the chair. They were then
More than one Race 1 (4%) instructed to release tension in the body, remain still, allow the chair to
Education
Some University 1 (4%)
support them, and breathe at their own natural pace.
2-Year University 5 (20%) Activity 2: Deep breathing (self-paced). Participants began with
4-Year University 10 (40%) the instructions detailed in the control activity. Deep breathing in­
Some Graduate School 2 (8%) structions were added with video-guidance to breathe through the
Graduate Degree or Higher 7 (28%)
nostrils, fully filling and emptying the lungs with each breath. Partici­
Major changes to exercise habitsa
Yes - pants were instructed to bring attention to each breath.
No 25 (100%) Activity 3: Deep breathing (externally-paced, positive control).
Strenuous exercise (frequency)a Participants began with the instructions detailed in the deep breathing
Never 11 (44%) activity. Pacing is added through audio cues (sound of a bell) indicating
1 time/week 3 (12%)
when to inhale and exhale, following a pace of 6 b/m.
2 times/week 4 (16%)
3 times/week 5 (20%) Activity 4: Sheetali and Sheetkari (self-paced). Participants began
4 times/week 2 (8%) with the instructions detailed in the deep breathing activity. Facial
Mind-body practicea manipulation was added through video guidance on how to perform
Yoga 9 (36%)
Sheetali and Sheetkari (Fig. 1). As in our previous work, Sheetali and
Meditation 4 (12%)
Qi Gong, Tai Chi, Martial Arts 1 (4%) Sheetkari were considered complementary parts of the same interven­
Breathing 3 (12%) tion, and were each performed consecutively in two five-minute seg­
Major diet changesa ments (totaling twenty minutes), with video instruction before each
Yes - breathing style. In Sheetali, the participants are instructed to follow a
No 25 (100%)
series of five steps, i.e. protrude their tongue a comfortable distance; roll
Servings of Caffeine per weeka
Coffee- drip 2.8 ± 3.3 into a tube; inhale through the tube; retract the tube back into the
Coffee- pressed 0.3 ± 1.6 mouth; and exhale through the nose. In Sheetkari, lips are kept apart,
Espresso 0.4 ± 1.2 and inhalation takes place through clenched teeth and a relaxed tongue.
Black tea 1.2 ± 1.6
Exhalation occurs through the nose.
Chai 0.2 ± 0.5
Green or white tea 1.4 ± 2.1
Activity 5: Sheetali and Sheetkari (externally-paced). Partici­
Energy drink 0.04 ± 0.2 pants began with the entire instructions outlined in Activity 4. Pacing
Soda pop (caffeinated) 0.2 ± 0.4 was added through audio cues indicating when to inhale and exhale,
Caffeine pill 0.04 ± 0.2 following a pace of 6 b/m.
Caffeinated candy or gum 0.04 ± 0.2
Body Mass Index (BMI, kg/m2)
Average 24 ± 3 2.3.2. Description of measures
Range 19–29 The Nijmegen Questionnaire. This validated screening tool has
a been shown to have a sensitivity of 91% and specificity of 95% for hy­
During the last four weeks.
perventilation [39] and give a broad view of symptoms associated with
dysfunctional breathing patterns. Participants are asked to record their
Statistical Computing, Vienna, Austria [40]) set for 5 × 5 crossover
responses regarding 16 conditions on a five-point scale, ranging from
design with a Williams Latin Square. [41]
‘never’ recorded as 0 to ‘very often’ counted as 4; thereby yielding a total
score ranging between 0 and 64. Those with scores greater than 20 were
2.3. Intervention procedure excluded at screening.
Heart Rate Variability. The ProComp8 Infiniti Encoder® (Thought
Each intervention activity was completed in the morning during one Technology Ltd., Montreal, Canada) was used to measure HRV, using a
of five separate study visits; morning was selected to reduce dirunal sampling rate of DC 512 Hz at 2048 samples/s, [42] which provides
variation in HRV. The five activies were: (1) sitting quietly with no adequate sensitivity for our outcomes. [43] Electrocardiogram (EKG)
external instructions on breath rate; (2) self-paced deep breathing; (3) (used to measure R-R intervals for calculation of HRV) was measured via
externally-paced deep breathing at a rate of 6 b/m, (4) self-paced electrodes placed on both wrists using adjustable bands. HRV data
Sheetali and Sheetkari pranayama for five minutes each; and (5) generally includes (1) the distance between adjacent “R” waves in an
externally-paced Sheetali and Sheetkari pranayama intervention at a EKG, and (2) the power density spectrum of the EKG waves. [44] HRV
rate of 6 b/m, for five minutes each. Activity 1, i.e. sitting quietly served data was analyzed in terms of “time domain” i.e. SDNN and RMSSD and
as the passive control, and Activity 3 was considered a positive control, “frequency domain” i.e. HF-HRV (0.15–0.4 Hz) and LF-HRV (0.04–0.15
based on evidence that deep breathing at 6 b/m increases HRV. [5] Hz). [22] Our primary HRV outcome was RMSSD because it is the most
Measurement periods during each visit consisted of four ten-minute appropriate HRV index for our intended respiratory rate range (< 9 b/

4
E. Sharpe et al. Journal of Psychosomatic Research 148 (2021) 110569

Fig. 3. Respiratory rate (y-axis) during eight 5-min steps (x-axis) at each of five visits/breathing activities (Activities 1–5, indicated by colored lines). Activity 1 was
Sitting Quietly, Activity 2 was Deep Breathing (self-paced), Activity 3 was Deep Breathing (externally-paced), Activity 4 was Sheetali and Sheetkari (self-paced), and
Activity 5 was Sheetali and Sheetkari (externally paced). Steps 1 & 2 are baseline/ pre-intervention measurements, in Steps 3–6 the breathing activity for the visit is
administered per the randomization sequence, and Steps 7–8 are post-intervention measurements. For Sheetali and Sheetkari interventions, Sheetali took place
during Steps 3–4 and Sheetkari took place during Steps 5–6. Error bars represent ±1 SE.

Fig. 4. Mean respiratory rate (breaths per minute, b/min) over time (pre, during, and post breathing activities) for each intervention (Activities 1–5). Activity 1 was
Sitting Quietly, Activity 2 was Deep Breathing (self-paced), Activity 3 was Deep Breathing (externally-paced), Activity 4 was Sheetali and Sheetkari (self-paced), and
Activity 5 was Sheetali and Sheetkari (externally paced). *P < .05, ** P < .01, ***P < .001, NS P > .05 (not significant). Error bars represent standard error.

5
E. Sharpe et al. Journal of Psychosomatic Research 148 (2021) 110569

incomplete data (i.e., two completed only one visit, and the other only
two visits).

2.5. Statistical analysis

Prior to analysis of effects, we summarized respiratory rate and HRV

measures into three summary averages: “pre”, defined as the average
across the two 5-min segments occurring before the breathing exercise
(steps 1 and 2); “during”, defined as the average across the four 5-min
segments occurring during the breathing exercise (steps 3–6); and
“post”, defined as the average across the two 5-min segments occurring
after the breathing exercise (steps 7 and 8).
The main analyses comparing five conditions used random-intercept
models adjusted for the condition-dependent pre-intervention measure
of the outcome, using either pre-during change or pre-post change as the
dependent variable and condition as the main predictor. Significance of
contrasts between pairs of conditions was made using a Tukey's HSD
correction. For comparisons of different times within the same condi­
tion, we used paired t-tests. We separately tested two-way random-
intercept models for each outcome, with pacing (self- or externally-
regulated) and breathing style (deep breathing vs pranayama) as fac­
tors, using only data from Conditions 2–5.
Missing data were not imputed, as the random effects model is able
to use all available outcomes, without imputation. In order to test
sensitivity to missing data, we reran all analyses using only data from
participants who completed all activities. All analysis were completed
using R® version 3.6.0 (Vienna, Austria) and a threshold of P < .05 was
considered indicative of statistical significance.

Fig. 5. Box and Whisker plots of variability in mean respiratory rate (breaths 2.5.1. Sample size calculation
per minute, b/min) at baseline (A); and during each intervention, Activity 1–5 Applying estimates of baseline HRV parameters from our previous
(B). The grey box represents interquartile range. The dark, solid line represents research, [30] and utilizing a crossover design, we calculated that a
the median. The lowest and tallest lines represent minimum and maximum
sample size of n = 30 would provide 80% power with an alpha of 0.05 to
values, respectively.
detect a 50% increase in HRV parameters. Therefore, our targeted
enrollment was 45 participants, to allow for up to 12.5% attrition to
m), while HF-HRV, and standard deviation of N-N intervals (SDNN) account for incomplete follow-up, data artifacts, and/or inability of the
were tested as secondary outcome variables because they are also participant to complete study visits. We note that, with the resultant n =
associated with PNS response, yet HF-HRV is best for respiratory rates 25 participants obtained, our calculation would have indicated only
between 9 and 24 b/m and SDNN is the suggested index for 24-h heart 72% power to detect the same effect relative to control.
rate recordings [14,37].
Respiratory Rate. The ProComp8 Infiniti Encoder® (Thought 3. Results
Technology Ltd., Montreal, Canada) was also used to measure respira­
tory rate. Respiratory measurements (breaths per minute or b/m) were Two hundred and two participants were screened for this study, 46
acquired via thoracic and abdominal straps secured across the chest and were eligible and enrolled after applying inclusion/exclusion criteria,
abdomen. HRV and respiratory rate data was aquired and processed and 80% (37 participants) completed all five visits, as shown in Fig. 2.
from this device using Biograph Infiniti® software (Thought Technolo­ Although 37 participants finished the study, after artifact processing
gies Ltd., Montreal). by two blinded assessors, and removal of three participants who were
extreme biological outliers (with excessive HRV throughout all activ­
2.4. Data collection and management ities), data for 25 participants were finally used in our analyses.
Participant demographics are detailed in Table 2.
All physiological data acquisition was accomplished within the in­
strument software and then exported to an encrypted computer for 3.1. Physiological changes
analysis and uploaded to REDCap®. HRV data was processed within the
Cardio-Pro software (Thought Technology Ltd., Montreal, Canada), [45] 3.1.1. Respiratory rate
wherein each five-minute segment was subdivided into one-minute The relationships between respiratory rate, activity, and time course
epochs for examination and manual artifact correction by two study of the intervention period are diagramed in Figs. 3, 4, and 5. Respiration
investigators (AS, RH), blinded to the activity being tested. rate significantly decreased from pre-intervention to during the inter­
Error across R-R interval time series was measured and flagged vention for both types of pranayama (deep breathing and Sheetali/
digitally and confirmed manually. Potential outliers were identified by Sheetkari, Activities 2–5) while there was very little change in the
the software, and then visually examined by AS and RH for confirmation control activity of sitting quietly (Activity 1, Figs. 3 and 4). The decrease
and removal. Ultimately, participants that needed to be excluded from in respiratory rate for Activities 3–5 was significantly greater than
analysis due to error in their HRV measurements were determined using observed in Activity 1 (p < .05 for all comparisons of Activities 3–5 to
this process. Artifacts created by movement, coughing, or other in­ Activity 1, when using a Tukey HSD correction for five comparisons),
terferences were manually identified in the EKG spectra and normalized with a pre-during change of —4.29 ± 2.63 b/m for Activity 3 (pre: 11.17
using the Thought Technology software. There were three additional vs. during: 6.88), —3.83 ± 3.15 b/m for Activity 4 (pre: 11.17 vs.
participants that were removed before final data analysis due to during: 7.35), and —4.16 ± 2.3 for Activity 5 (pre: 11.57 vs. during:

6
E. Sharpe et al. Journal of Psychosomatic Research 148 (2021) 110569

Fig. 6. Log of the root mean square of successive normal RR intervals (RMSSSD, y-axis) during eight 5-min steps (x-axis) at each of five visits/breathing activities
(indicated by colored lines). Steps 1 & 2 are baseline/ pre-intervention measurements, in Steps 3–6 the breathing activity for the visit is administered per the
randomization sequence, and Steps 7–8 are post-intervention measurements. Breathing activities are defined as follows: Activity 1: Control/ Sitting quietly/ Normal
breathing; Activity 2: Deep Breathing (self-paced); Activity 3: Deep Breathing (externally-paced); Activity 4: Sheetali/Sheetkari Breathing (self-paced); Activity 5:
Sheetali/ Sheetkari Breathing (externally-paced). Error bars represent ±1 standar error (SE).

7.41). and 5, compared to a pre-during RMSSD increase of 0.30 ± 0.43 ms and

Fig. 5 shows variability in breath rate, giving us insight into the ef­ 0.41 ± 0.48 ms for Activities 2 and 4. Within pacing types (self-pacing
fects of self pacing vs. external pacing. At baseline (5A), breath rates in and external-pacing), breathing styles (i.e. Deep Breathing vs. Sheetali/
all conditions were similarly variable and in the same range (inter­ Sheetkari) were not associated with different magnitudes of HRV
quartile range, ~10–13 b/m). During the intervention period (5B), as change. Specifically, average pre-during change in RMSSD for Activities
designed, there was less variability in breath rates during externally- 2 vs. 4 were not significantly different from one another (p = .54), nor
paced pranayama (Activities 3 and 5) than self-paced versions of pra­ were those values for externally-paced practices, Activities 3 vs. 5: (p =
nayama (Activites 2 and 4), indicating adherence to pacing prompts. .98). Yet, when controlling for respiratory rate, Activity 4 (Sheetali/
Although average breath rate was lowest for externally-paced breathing Sheetkari) showed a greater pre-during increase compared to Activity 2
(closer to the targeted 6 b/m, with an interquartile range of ~6–8), (Deep Breathing) yet this difference was not significant (p = .07).
average breath rates produced during Activities 2–5 were not statisti­ Changes in RMSSD over the duration of each visit (Steps 1–8 during
cally different from one another. Activities 1–5) can be seen in Fig. 6. Baseline values taken during the
first 10-min of each visit (steps 1 and 2) were similar in all activities, but
3.1.2. Heart rate variability (RMSSD) there are notable differences between baseline and the intervention
When comparing pre-during changes between all interventions period (steps 3–6), as well as the intervention period and the post-
(Activities 1–5), the externally-paced pranayamas, Activities 3 (deep practice period (steps 7 and 8) for all activities, especially Activities
breathing) and 5 (Sheetali/Sheetkari), resulted in greater RMSSD than 3–5. Summary means of RMSSD during the “pre”-, “during”-, and “post”-
the control Activity 1 when adjusting for “pre” activity values for each intervention steps are shown in Fig. 7A. Increases in RMSSD from “pre”
intervention (p = .005 and p = .02, respectively, using a Tukey HSD to “during” were seen across all five activities (all p-values <.05). Ac­
correction for comparison of five estimates), with an average logRMSSD tivities 3 and 5 (externally-paced deep breathing and Sheetali/Sheet­
of 4.05 ± 0.4 ms during Activity 3, 4.01 ± 0.4 ms during Activity 5, and kari, respectively) showed significant differences in RMSSD between all
3.82 ± 0.49 ms during Activity 1. steps (pre-during, during-post, and pre-post, as shown in Fig. 7A), with
When comparing pre-during changes between pranayama in­ the largest change seen from pre to during these activities (logRMSSD
terventions (Activities 2–5) in a two-way model using breathing style change of 0.50 ms for Activity 3 and 0.49 ms for Activity 5), while
(normal vs. Sheetali/Sheetkari) and pacing (self- vs. externally-paced) as during-post change was 0.32 ms for both externally-paced activities, and
factors, we found no significant interaction. Pacing was associated with pre-post change was 0.19 and 0.17 ms for Activities 3 and 5,
a significant main effect on HRV, with external pacing (Activities 3 and respectively.
5) resulting in a greater increase in HRV from pre-intervention to during
the intervention than self pacing used in Activities 2 and 4 (effect on pre- 3.1.3. HF-HRV and SDNN (Secondary outcomes)
during logRMSSD change; b = − 0.17, p = .02), with a pre-during When comparing pre-during changes between all interventions
RMSSD increase of 0.50 ± 0.42 ms and 0.49 ± 0.41 ms for Activities 3 (Activities 1–5), Activities 3, 4 and 5 showed a significantly greater pre-

7
E. Sharpe et al. Journal of Psychosomatic Research 148 (2021) 110569

Fig. 7. HRV parameters over time (pre, during, post intervention) during each intervention (Activities 1–5) represented in terms of: (A) log of the root mean square
of successive normal RR intervals (RMSSD); (B) log of the standard deviation of all normal RR intervals (SDNN); and (C) log of high-frequency (HF) heart rate
variability. *P < .05, ** P < .01, ***P < .001, NS P > .05 (not significant). Error bars represent confidence intervals.

during increase in logSDNN (0.57 ± 0.36 ms, 0.48 ± 0.42 ms, and 0.50 We also see merit in investigating whether distinct oral morphology
± 0.33 ms) than Activity 1 (0.2 ± 0.31 ms) using the same model (p- used in some practices may increase physiologial benefits [28,48].
values: <0.001, 0.005, and 0.02, respectively, when comparing Activ­ Although our study was conducted in a healthy population, our results
ities 3–5 to Activity 1). are congruent with a recent randomized controlled trial of Sheetali
Supporting the effect of pacing on PNS response, similar results were pranayama in participants with essential hypertension, that demon­
seen for SDNN outcomes as were shown for RMSSD, indicating strated both relatively large increases in RMSSD and blood pressure
externally-paced Activities 3 and 5 produced a stronger change in PNS lowering effects compared to those not performing pranayama [53].
response than self-paced Activities 2 and 4, with an effect on pre-during However, unlike our study, this trial did not assess if one specific
logSDNN change of b = − 0.12, and p = .04. component of pranayama was able to demonstrate greater benefit over
Changes in SDNN and HF-HRV throughout the visit (pre-during-post others, and is limited by a lack of an active comparator.
intervention) are shown in Fig. 7B and C. Significant increases in SDNN The primary importance of this work lies in understanding the
from “pre” to “during” were seen across all five interventions (Activities mechanism by which various pranayama and other breathing practices
1–5), with all p-values <.05 (7B). There were no significant changes in may differentially affect the autonomic nervous system (i.e. through the
HF across any activity or time comparison (7C), yet a strong, positive simple act of sitting quietly, the pace of breathing, and/or the modifi­
correlation existed between HF and RMSSD (r > 0.80 for all activities cations made to the face and airways during practice). Of these com­
and time points; data not shown), which was hypothesized. [36,46,47] ponents, only pace of breathing has been isolated and investigated in
prior work, with physiologic mechanisms explaining the relationship of
4. Discussion slow-breathing pranayama and it's effects on autonomic modulation
proposed by Jerath et al. [54]. They suggest that pranayama breathing,
To our knowledge, this is the first study to separately evaluate the and inhalation specifically, enhances both the actions of stretch-induced
impact of the specific components of pranayama practice on HRV. Our inhibitory neural signals via activation of slowly adapting stretch re­
results demonstrate an increase in an index of cardiac parasympathetic ceptors, as well as, fibroblast-mediated hyperpolarization. These initial
tone during pranayama (both deep breathing and Sheetali/Sheetkari), responses led to downstream inhibitory effects within the hypothalamus
as expected, and that sitting quietly, deep breathing, and pacing were all and brainstem, lowering metabolic activity which upregulates para­
associated with that response. We found that externally-paced breathing sympathetic activity relative to sympathetic activity.
practices resulted in a greater pre-during increase in parasympathetic With targeted investigation of components beyond pace of breath,
tone than self-paced activities. Our results suggest that Sheetali/Sheet­ we can begin to understand the most important components of a prac­
kari pranayama merits further investigation in larger controlled trials. tice. These results can lead to more specific and targeted breathing

8
E. Sharpe et al. Journal of Psychosomatic Research 148 (2021) 110569

practices being tested for adoption to routine clinical conditions such as Declaration of Competing Interest
hypertension, type 2 diabetes and anxiety disorders. Moreover, the
findings can be extended to examine the impacts of components of other All authors have completed the Unified Competing Interest form. The
pranayama practices. authors have no competing interests to report.
The strengths of this trial include the novel comparison of several
individual components of a pranayama practice, the randomized Acknowledgements
crossover design, which enabled each case to act as their own control.
Our study demonstrates a technique through which components of a This project was supported by grant number 5R90AT00892403 from
proven traditional practice can be isolated and tested alongside one the National Center for Complementary and Integrative Health of the
another, within a relatively small sample size using a crossover design, National Institutes of Health, USA.
in order to elucidate which aspects of the practice may be most
responsible for its physiological, psychological or other measurable References
effects.
The study results needs to be viewed in light of its limitations. We [1] T.C. Clarke, et al., Use of yoga, meditation, and chiropractors among U.S. Adults
aged 18 and over, NCHS Data Brief. 325 (2018) 1–8.
had an attrition rate of 20% (losing nine participants from our starting [2] H. Cramer, The efficacy and safety of yoga in managing hypertension, Exp. Clin.
sample size of 46 at Visit 1 to 37 at Visit 5), indicating a larger drop-out Endocrinol. Diabetes 124 (2) (2016) 65–70.
rate than we predicted. We also did not expect such a high sample loss [3] H. Thind, et al., The effects of yoga among adults with type 2 diabetes: a systematic
review and meta-analysis, Prev. Med. 105 (2017) 116–126.
due to HRV artifacts and extreme outliers, leaving us with 25 partici­ [4] N.R. Prabhakar, Oxygen sensing by the carotid body chemoreceptors, J. Appl.
pants, rather our intended 30, and we did not evaluate for inter-reader Physiol. 88 (6) (1985) 2287–2295, 2000.
reliability between two blinded HRV assessors. The sample included [5] H.S. Song, P.M. Lehrer, The effects of specific respiratory rates on heart rate and
heart rate variability, Appl. Psychophysiol. Biofeedback 28 (1) (2003) 13–23.
healthy individuals without a history of hypertension or cardiovascular [6] C.N. Joseph, et al., Slow breathing improves arterial baroreflex sensitivity and
diseases, thus, it is unknown if the same mechanisms or magnitude of decreases blood pressure in essential hypertension, Hypertension 46 (4) (2005)
changes observed here would be consistent in a diseased population. 714–718.
[7] J.J. Goldberger, et al., Relationship of heart rate variability to parasympathetic
And lastly, each person only completed each breathing practice one
effect, Circulation 103 (15) (2001) 1977–1983.
time. It is possible there was some stress involved in learning a new skill [8] M.C. Pascoe, D.R. Thompson, C.F. Ski, Yoga, mindfulness-based stress reduction
(or being observed), and with more practice (in a longer duration study), and stress-related physiological measures: a meta-analysis,
greater benefits could be experienced. Psychoneuroendocrinology 86 (2017) 152–168.
[9] K.M. Edwards, et al., Effects on blood pressure and autonomic nervous system
Future studies involving larger sample sizes, among individuals who function of a 12-week exercise or exercise plus DASH-diet intervention in
suffer from hypertension are likely to offer more definitive results. It individuals with elevated blood pressure, Acta Physiol. (Oxford, England) 203 (3)
would be of interest to investigate whether other components contribute (2011) 343–350.
[10] R. Virtanen, et al., Reduced heart rate variability in hypertension: associations with
to the effectiveness of pranayama, including: varied pacing (slow, me­ lifestyle factors and plasma renin activity, J. Hum. Hypertens. 17 (3) (2003)
dium, fast, using external pacing); lung volume used (as a marker for 171–179.
depth of breath in order to predict a possible increase in tidal capacity, [11] E. Vaschillo, et al., Heart rate variability biofeedback as a method for assessing
baroreflex function: a preliminary study of resonance in the cardiovascular system,
which is one goal of pranayama [49]); inhalation to exhalation ratio (a Appl. Psychophysiol. Biofeedback 27 (1) (2002) 1–27.
1:2 breathing ratio is used in most pranayama practices [33,49,50] as [12] R. Cernes, R. Zimlichman, Role of paced breathing for treatment of hypertension,
well as in pursed-lip breathing, which has been proven to increase Curr. Hypertens. Rep. 19 (6) (2017) 9.
[13] J.F. Thayer, S.S. Yamamoto, J.F. Brosschot, The relationship of autonomic
oxygenation of the blood, triggering a PNS response [4] and slowing imbalance, heart rate variability and cardiovascular disease risk factors, Int. J.
respiration [48]); opening the mouth (during inhale to radiate moisture Cardiol. 141 (2) (2010) 122–131.
and heat, or during exhale to excrete more CO2 from the lungs as is done [14] Heart rate variability: standards of measurement, physiological interpretation and
clinical use. Task Force of the European society of Cardiology and the North
in Pursed-lip Breathing [28]); Sheetali vs. Sheetkari (comparison to one
American Society of pacing and electrophysiology, Circulation 93 (5) (1996)
another [33]). Additionally, variations on the practice that could be 1043–1065.
explored include: pausing between breaths (allowing time for gas ex­ [15] A. Tyagi, M. Cohen, Yoga and heart rate variability: a comprehensive review of the
change, increasing breathing efficiency, [50] and increasing HRV literature, Int. J. Yoga 9 (2) (2016) 97–113.
[16] C. Kumar, B.H. Krishna, N.M. Reddy, Modulation of sympathovagal balance after
[11,51–53]); and adding buzzing bee breath/ bhramari (humming) chandranadi pranayama in healthy volunteers, Int. J. Med. Res. Health Sci. 4 (4)
during exhalation (a traditionally recommended augmentation to be (2015) 783–785.
used with Sheetali/Sheetkari for enhanced relaxation [33]). [17] S. Singh, V. Gaurav, V. Parkash, Effects of a 6-week nadi-shodhana pranayama
training on cardio-pulmonary parameters, J. Phys. Educ. Sports Manag. 2 (24)
(2011) 44–47.
5. Conclusions [18] E.L. Phillips, P.D. Donofrio, Autonomic Disorders, in: L.R. Squire (Ed.),
Encyclopedia of Neuroscience, Academic Press, Oxford, 2009, pp. 799–808.
[19] L. Zou, et al., Effects of mind− body exercises (Tai Chi/Yoga) on heart rate variability
In this preliminary study, we completed the first investigation into parameters and perceived stress: a systematic review with meta-analysis of randomized
how several components of a pranayama practice can impact HRV. We controlled trials, J. Clin. Med. (2018) 7(11).
found that all five practices increased HRV as compared to the pre- [20] N. Markil, et al., Yoga Nidra relaxation increases heart rate variability and is
unaffected by a prior bout of Hatha yoga, J. Altern. Complement. Med. 18 (10)
intervention period. When using external pacing (at 6 b/m), both deep (2012) 953–958.
breathing and Sheetali/Sheetkari increased HRV (from pre- to during [21] Y. Ma, R. McCraty, Heart rate variability in mind-body interventions, Complement
the practice) more than sitting quietly while breathing normally. We Ther. Med. 29 (2016) A1–a2.
[22] K. Fujiwara, et al., Epileptic seizure prediction based on multivariate statistical
found external pacing (6 b/m) resulted in greater cardiac para­
process control of heart rate variability features, IEEE Trans. Biomed. Eng. 63 (6)
sympathetic tone compared to self-pacing. Additionally, both self-paced (2016) 1321–1332.
and externally-paced practices were associated with descreased respi­ [23] F.M. de Castilho, et al., Heart rate variability as predictor of mortality in sepsis: a
ratory rate, compared to sitting quietly. These findings suggest that prospective cohort study, PLoS One 12 (6) (2017), e0180060.
[24] D. Aronson, M.A. Mittleman, A.J. Burger, Measures of heart period variability as
sitting quietly, deep breathing, and slow pace of breath need to be tar­ predictors of mortality in hospitalized patients with decompensated congestive
geted for investigation as active components of the practice. Lastly, heart failure, Am. J. Cardiol. 93 (1) (2004) 59–63.
there was preliminary evidence from this study that the oral morphology [25] A.S. Bhogal, et al., Which heart rate variability index is an independent predictor of
mortality in cirrhosis? Dig. Liver Dis. 51 (5) (2019) 695–702.
used in Sheetali/Sheetkari may play a role in HRV changes, but this [26] D. Frawley, Yoga & Ayurveda: Self-Healing and Self-Realization., Lotus Press, 1999.
question merits more investigation with a larger sample size. [27] J. Mallison, Gherand Samhita, English Translation, Yoga Vidya, 2004.

9
E. Sharpe et al. Journal of Psychosomatic Research 148 (2021) 110569

[28] Association, A.L., Pursed Lip Breathing, Available from: https://www.lung.org/lun [40] R.C. Team, R: A Language and Environment for Statistical Computing, Available
g-health-diseases/lung-disease-lookup/copd/patient-resources-and-videos/purse from: https://www.R-project.org/, 2020.
d-lip-breathing-video, 2020. [41] H. Patterson, E. Williams, E. Hunter, Block designs for variety trials, J. Agric. Sci.
[29] R. Cernes, R. Zimlichman, RESPeRATE: the role of paced breathing in hypertension 90 (2) (1978) 395–400.
treatment, J. Am. Soc. Hypertens 9 (1) (2015) 38–47. [42] T. Tech, ProComp Infiniti System w/BioGraph Infiniti Software - T7500M,
[30] P. Shetty, et al., Effects of sheetali and sheetkari pranayamas on blood pressure and Available from: https://thoughttechnology.com/procomp-infiniti-system-w-biogr
autonomic function in hypertensive patients, Integr. Med. 16 (5) (2017) 32–37 aph-infiniti-software-t7500m/, 2020.
(Encinitas, Calif.). [43] O. Kwon, et al., Electrocardiogram sampling frequency range acceptable for heart
[31] Y. Wu, et al., Yoga as antihypertensive lifestyle therapy: a systematic review and rate variability analysis, Healthc. Inform. Res. 24 (3) (2018) 198–206.
meta-analysis, Mayo Clin. Proc. 94 (3) (2019) 432–446. [44] S. Chen, et al., Effects of heart rate variability biofeedback on cardiovascular
[32] C. Li, et al., Effects of slow breathing rate on heart rate variability and arterial responses and autonomic sympathovagal modulation following stressor tasks in
baroreflex sensitivity in essential hypertension, Medicine (Baltimore) 97 (18) prehypertensives, J. Hum. Hypertens. 30 (2) (2016) 105–111.
(2018), e0639. [45] J.M. Del Pozo, et al., Biofeedback treatment increases heart rate variability in
[33] V. Mandlik, Yoga Sopan (Easy Way to Health: Physical and Mental), Yoga patients with known coronary artery disease, Am. Heart J. 147 (3) (2004) 545.
Chaitanya Prakashan, Naskih, Maharashtra, India, 2008. [46] F. Shaffer, J.P. Ginsberg, An overview of heart rate variability metrics and norms,
[34] T.E. Strandberg, K. Pitkala, What is the most important component of blood Front. Public Health 5 (258) (2017) 258.
pressure: systolic, diastolic or pulse pressure? Curr. Opin. Nephrol. Hypertens. 12 [47] S. Laborde, E. Mosley, J.F. Thayer, Heart rate variability and cardiac vagal tone in
(3) (2003) 293–297. psychophysiological research - recommendations for experiment planning, data
[35] A.P. Bress, et al., Cost-effectiveness of intensive versus standard blood-pressure analysis, and data reporting, Front. Psychol. 8 (2017) 213.
control, N. Engl. J. Med. 377 (8) (2017) 745–755. [48] S. Sakhaei, et al., The impact of pursed-lips breathing maneuver on cardiac,
[36] G.G. Berntson, et al., Heart rate variability: origins, methods, and interpretive respiratory, and oxygenation parameters in COPD patients, Open Access Maced. J.
caveats, Psychophysiology 34 (6) (1997) 623–648. Med. Sci. 6 (10) (2018) 1851–1856.
[37] M.E. Russell, et al., Inclusion of a rest period in diaphragmatic breathing increases [49] V. Mandlik, Ashtang Yoga: Teachers Training Course, 2008.
high frequency heart rate variability: implications for behavioral therapy, [50] P.W. Dorle, K. Woodworth, in: Y.V. Gurukul (Ed.), Human Anatomy & Physiology
Psychophysiology 54 (3) (2017) 358–365. Handbook, 2008.
[38] J. Penttilä, et al., Time domain, geometrical and frequency domain analysis of [51] G. Grassi, et al., Baroreflex activation therapy in congestive heart failure: novel
cardiac vagal outflow: effects of various respiratory patterns, Clin. Physiol. 21 (3) findings and future insights, Curr. Hypertens. Rep. 18 (8) (2016) 6.
(2001) 365–376. [52] P.M. Lehrer, et al., Heart rate variability biofeedback increases baroreflex gain and
[39] E.P. Grammatopoulou, et al., Hyperventilation in asthma: a validation study of the peak expiratory flow, Psychosom. Med. 65 (5) (2003) 796–805.
Nijmegen Questionnaire - NQ, J. Asthma 51 (8) (2014) 839–846. [53] P.M. Lehrer, R. Gevirtz, Heart rate variability biofeedback: how and why does it
work? Front. Psychol. 5 (2014) 9.

10