Cardiovascular phenotyping for personalized lifestyle treatments of chronic abdominal pain in Irritable Bowel Syndrome: A randomised pilot study

Abstract

Background:

Different physical exercise interventions for pain and other related symptoms largely follow non-personalized guidelines and show a high degree of variability in outcome. These interventions are considered to have different pathways towards improvement in autonomic regulation of energy metabolism. The current pilot study was conducted to assess the predictive value of individual cardiovascular (CV) activity markers at rest to predict clinical outcomes for two popular exercise-based interventions (walking and yoga) in patients with Irritable Bowel Syndrome (IBS).

Methods:

Twenty-seven adult participants with IBS were randomly assigned to a 16-biweekly Iyengar yoga or walking program. They completed pre- and post-treatment assessments on IBS symptom severity, affective and somatic complaints, and various measures of resting autonomic function including blood pressure (BP), heart rate and its variability, baroreceptor-sensitivity (BRS) to activations and inhibitions with gains of brady- and tachy-cardiac baro-responses, and BP start points for these spontaneous baroreflexes.

Results:

Pre-treatment BRS was differentially related to clinical response for the treatment groups. Specifically, a significant decrease in pain severity was found in response to yoga for those participants who had lower resting BRS to activations, but decreased pain severity was associated with higher resting BRS for those in the walking group. The effect was not related to affective symptom relief. Other CV measures showed similar associations with clinical outcomes for both groups.

Conclusions:

The data suggest therefore that CV based phenotypes may be useful in personalizing clinical interventions for IBS. They may also point to autonomic mechanisms that are targets for such interventions.

Graphical Abstract

Introduction

Chronic pain, or pain of neuropathic, somatic, or visceral origin persisting more than three months, causes significant disability, suffering and expense for every segment of the population. ^1,2 Chronic pain affects mood, sleep, physical activity, and is the most prevalent symptom driving medical visits. Various approaches have been developed to treat chronic pain. However, the outcome of such treatments is often unpredictable or satisfactory. Opioid medications once the primary recommended treatment for many patients with chronic pain have been found to be of limited benefit in long term, and frequently unsafe leading to addiction, mortality, and other complications including the narcotic bowel syndrome. ^3–6 Although not proven of general benefit across all individuals, healthy lifestyle interventions, like physical activity and yoga, have been offered as safer methods to reduce disease impact and lessen chronic pain severity with few adverse events. ^7–10 While attractive, physical exercise interventions as an alternative form of pain management may only be effective in subgroups of patients. For example, we have recently reported that physical exercise (e.g., walking and yoga) may provide modest improvement in visceral pain and other negative symptoms in patients with Irritable Bowel Syndrome (IBS), but also show variability in outcome. ¹¹ As such, markers useful for predicting outcomes from pain management interventions are a significant area of interest. ¹² In particular, identification and understanding of reliable biological markers that might help to predict who will benefit from a specific intervention is a key challenge in developing personalized medicine approaches to chronic pain. ^13,14

Among different autonomic biomarkers, baroreceptor sensitivity (BRS) and some related variables (e.g., heart rate variability, HRV, and blood pressure mean and variability) have been found to be the most reliable predictors of clinical outcomes and early indicators of clinical changes in many pathological conditions. ^15–19 In the first case, individual genotypes or endo-phenotypes associated with these cardiovascular (CV) variables have been proposed as moderators of health resiliency, both in general, and against pain, in particular. ^20–25 For example, prior studies have shown that a family history of hypertension in normotensive humans and presence of hypertension predicted lower pain intensity or higher pain thresholds. ^26,27 This corresponds with findings that a genetic predisposition to higher sympathetic activation in response to painful stimuli in people with the LL genotype of the serotonin transporter-linked polymorphic region (5-HTTLPR) was reliably associated with higher pain thresholds and more habituation to acute visceral and somatic pain compared with people with the s allele of the 5-HTTLPR and the association was stable at a period of 1 year. ²⁸ However, this endophenotype was probably related to an individual difference in the affectivity (trait anxiety or neuroticism) that could only impact on affective mechanism of pain chronification. In the second case, a significant increase in cardiac BRS during several weeks of physical exercise (e.g., endurance training) received health protection against cardiac death (10-year follow-up), and a similar result was seen in an animal model (i.e., reduced mortality in response to an ischemia test for those with a significant increase in cardiac BRS in a study of dogs). ^29,30 However, a simple improvement in exercise capacity without significant changes in BRS was not accompanied by a better prognosis in mortality rates. Moreover, an increase in BRS independent of training was associated with a reduced mortality risk. Other indicators (e.g., heart rate decrease, level of endurance) in the above-mentioned studies were only related to the exercise training interventions but not to the health benefits of the exercises.

In IBS and other chronic pain conditions multiple autonomic mechanisms have shown associations with components (nociceptive, affective, and cognitive) of the chronic pain experience. ^20,22 For example increases in diastolic blood pressure (DBP) and decreases in HRV have been associated with decreased BRS, but clinically to both lower nociceptive or higher affective complaints related to pain, respectively. ^22,31,32 This suggests that both across individuals and interventions autonomic responses may serve as predictors or early indicators of chronic pain treatment response depending on individual phenotype and treatment targets. For example, walking and other methods stimulating aerobic physical activity seem to improve health through motor (efferent) division of the nervous system. ^21,33 Yoga and other techniques increasing body and mind relaxation seem to improve psychosomatic health through sensory (afferent) division of the nervous system. ^34–36 Non-responders to one kind of intervention could be responsive to another group of treatments or treatment regimens, but this hypothesis was not yet explored. ^37,38

Taken together, the findings in subjects with chronic pain and in other health-threatening conditions suggest that treatment efficacy for some interventions may be moderated by individual baroreflex activity measures like BRS that can be used to predict and to monitor health-related effects of interventions. Other CV measures, like BP, HRV, and HR, may have complementary prognostic and monitoring utility in addition to these baroreflex measures (a baroreflex activity phenotype) moderating the pathways through which the baroreflex predicts and regulates health. Thus, the current study was conducted to assess the value of individual CV activity-based phenotypes assessed at rest, as predictors of clinical outcomes for two popular exercise-based interventions (walking and yoga) in patients with IBS. The first main aim was to discover similar or different intervention impacts of walking and yoga on CV changes pre- to post-treatment as biomarkers of their effectiveness and on changes in somatic and affective symptoms. The second main aim was to discover whether (i) clinical outcomes in response to the different interventions would be predicted by the same CV phenotypes in a similar way (‘simple’ effects) or (ii) CV predictors of clinical outcomes would be different per group (‘group moderation’ effects).

Materials and methods

Participants

Participants were 18- to 65-year-old men and women who reported their primary medical complaint as chronic abdominal pain or discomfort and associated bowel habit changes consistent with ROME III criteria for IBS. ³⁹ Participants were recruited from flyers in the community, internet announcements, and referrals from UCLA and local physicians. Participants were initially screened for study inclusion over the phone. If determined to be interested and preliminarily eligible, participants were asked to come in to the study laboratory for medical screening by a study nurse practitioner. Inclusion criteria for participation in this study included the following: 18 – 65 years old, a diagnosis of or symptoms of IBS as determined by a study nurse practitioner, and physically able to engage in a movement or exercise program. Participants were excluded if they: drank two or more units of alcohol per day (1 unit = a 12 oz [350 mL] bottle of beer or 1 glass [150 mL] of wine or ½ oz [15 mL] of liquor); smoked a pack or more a day; were obese (body mass index greater than 32 kg/m²); had current use of coumadin, heparin products, or other potentially blood-thinning agents; had current use of beta blockers or daily use of narcotic or prescription pain medications; had previous abdominal surgeries; if taking any antidepressants, were not on a stable dose for at least 3 months; engaged in vigorous aerobic exercise for more than 1 hour per day, every day of the week; currently did yoga more than once a week; had a history of psychiatric hospitalizations or a diagnosis of a major psychiatric disorder (e.g., schizophrenia, bipolar disorder) or serious memory problems; had other major medical conditions or illnesses that were primary; and if 50 years old or older, had not had a colonoscopy or sigmoidoscopy in the last 5 years. Of the few cases of participants who reported doing occasional yoga exercises at study enrollment, none were doing Iyengar yoga nor regularly attending classes. During the course of treatment, for those in the yoga condition, only Iyengar yoga poses were performed as home practice. Informed consent was signed by all eligible participants prior to enrollment, and the institutional review board at UCLA approved the protocol.

A total of 35 participants were enrolled and randomly assigned to either the Yoga or Walking conditions. Participants were randomized into parallel groups using simple randomization using SPSS randomization sequences. The study originally aimed for 25 participants in each treatment condition, but due to difficulties with recruitment, the study was halted at 35 enrolled participants, resulting in uneven group sizes. The sample included equal numbers of those with constipation as their primary bowel habit and those with diarrhea as primary. Of the 35 participants who were enrolled, 8 dropped out of treatment, 5 from Yoga and 3 from Walking. Dropout rates for both conditions were 23%. Reasons for withdrawal were traffic, difficulties in getting to sessions, and changes in schedule. Twenty-seven participants completed the treatment and returned for the post-treatment assessment. Of these, 17 were assigned to Yoga, and 10 to Walking. For one participant from the yoga group, continuous blood pressure measurements were poorly assessed both before and after treatment, and therefore, this participant’s blood pressure and related outcomes data were excluded from the analyses.

Interventions

Participants enrolled in the Walking intervention participated in up to 16 group sessions of non-aerobic, moderate paced, outdoor walking led by physical trainers who set the pace and led discussion during each session. Walking sessions were offered on a biweekly basis and lasted about 60 min. Discussion during each walking session focused on health benefits of walking and physical activity. Participants enrolled in Yoga also attended up to 16, 60-minute biweekly group sessions led by an Iyengar yoga certified instructor. Sessions involved instruction in a progressive series of Iyengar yoga poses specifically selected to address abdominal pain and gastrointestinal (GI) function. Participants were also encouraged to practice select postures at home between sessions. Details of the treatment procedures have been published elsewhere. ¹¹

Laboratory procedure

Participants were tested Monday through Friday, anytime between 9am and 5pm, so there was not consistency across the timing of the laboratory testing. As a pilot study with small sample size also assessing feasibility of the design, it allows flexibility in time of day of testing. Participants were asked to refrain from a) drinking alcohol and strenuous exercise the morning of testing, b) smoking and ingesting caffeine (e.g., tea, coffee) during the three hours preceding the laboratory procedure, and c) any pain medications, such as Advil or prescription pain meds, before coming in. CV measures were obtained during two laboratory visits, before and within the two weeks following completion of one of the two interventions. Following 5 minutes of seated rest, three blood pressure (BP) readings were taken using a standard automated cuff. An initial 10-min baseline period was then begun and the participant was asked to sit still, try not to move, but not to fall asleep. Subjects were aware that a live feed camera was on and that further experimental procedures including a psychological stress test would take place after this initial baseline was complete. Physiological variables (described below) recorded during this baseline period were used for this paper.

Measures

Primary self-report outcomes assessed before and after treatment included IBS severity (abdominal pain severity and overall GI symptom intensity) (see below). Extra measures related to primary outcomes were disease chronicity and usual symptom-free duration (1=Less than 1 hour; 2=Up to 24 hours; 3=Between 1 to 3 days; 4=Between 4 to 7 days; 5=Up to 1 month; 6=Up to 1 year). Secondary outcomes included measures of GI symptom-specific anxiety, negative and positive mood, and severity of non-GI somatic symptoms (see Supplementary materials 2). The secondary outcomes were added to the design of the study to distinguish specific clinical IBS and pain severity outcomes from symptom-specific distress and general health effects of the treatments. Findings related to the secondary outcomes are presented in Supplementary materials 2.

Overall GI and abdominal pain severity (primary clinical outcomes)

Participants rated severity of overall GI symptoms and of abdominal pain during the past week, from no symptoms to the most intense imaginable on a 21-point numerical rating scale (NRS) at each assessment. Validity of NRS scales for IBS severity has previously been established. ⁴⁰ Subjective assessment of changes in abdominal symptoms (i.e., a relative symptom severity assessment) was conducted both after treatment completion and at 6-month follow-up by a 7-point numerical rating scale: 1=Substantially improved; 2=Moderately improved; 3=Slightly improved; 4=No change; 5=Slightly worse; 6=Moderately worse; 7=Substantially worse.

Physiological measures

Continuous measures of electrocardiogram (ECG), beat-to-beat BP, and respiration were recorded during the laboratory session using Biopac MP100 hardware and Acq- Knowledge 3.8.2 software (Biopac, Goleta, California). A standard electrode configuration (right clavicle and precordial site V6) with three disposable Ag–AgCl electrodes (ConMed Corp.) was used to record the ECG. The signal was digitized at 2000 Hz, and RR-interval (RRI) series were derived using a QRS complex template detection algorithm to obtain R-peak localization as the apex of the interpolating parabolas. Beat-to-beat BP was measured non-invasively using a Finapres Continuous NIBP Monitor (Ohmeda, Englewood, Colorado) via a finger cuff attached to the third finger of the non-dominant hand and digitized at 2000 Hz with 12-bit resolution. The Finapres has been shown to be a suitable tool for reliable tracking of changes in BP. ⁴¹ Systolic (SBP) and diastolic (DBP) points were derived using two different template detection algorithms. All recordings were automatically and visually examined to verify ECG and BP wave classification and to correct for artifacts. Further details on the methods and algorithms can be found elsewhere. ^35,38,42 Respiration was recorded with the Biopac Respiratory Effort Transducer (Biopac, Goleta, California), a belt placed around the lower rib cage measuring changes in chest circumference.

Baroreflex sensitivity and heart rate variability

BRS and other baroreflex indexes were measured by the sequence method. ^23,38,42 SBP and RR interval (RRI) time series were scanned by custom software (see below) to identify sequences during which SBP and associated RRI both increased (“up” [+] sequences) or both decreased (“down” [–] sequences) successively in parallel over three or more beats (SBP and RRI ramps). We required a minimum change of 1 mmHg for SBP and 3 ms for RRI per beat and a minimum correlation of .80 between the parallel values to accept the pairs for calculating BRS (gain of baroreceptor sensitivity, ms/mmHg), and separately BPS (mmHg/beat, gain of SBP ramps, short-term SBP falls and rises triggering RRI baro-responses, −BPS and +BPS, respectively) and RRIS (ms/beat, gain of RRI ramps, short-term tachycardiac and bradycardiac baro-responses, −RRIS and +RRIS, respectively) as means of least squares linear regression slopes with lag 0, 1, and 2. ^38,42 Means of SBP start point or the first SBP reading of SBP ramp in the baroreflex sequences for initiation of down (−SPBP) and up (+SPBP) baroreflex reactions (the highest set-point for down sequences and the lowest set-point for up sequences or baroreflex down and up operational points) were also indicated to evaluate the effects of the baroreceptor resetting. ^23,38 A range between +SPBP and −SPBP values was also assessed as an indicator of the effectiveness of baroreflex regulation of the amplitude of spontaneous BP fluctuations. Means of Finapres-derived values of SBP and DBP were also used in the analyses.

Multistage band-pass linear filtering was adopted to suppress extraneous sources of RRI variation. ³⁸ This method is comparable to the Porges-Bohrer moving polynomial filter method of the assessment of respiratory sinus arrhythmia (HF-HRV) ^43,44 but extended the filtering procedure also to lower frequencies like the Traube-Hering-Mayer wave (LF-HRV). In this method, the RRI variances of residual time series (the filtered waveform) after a bandpass smoothing FIR (finite impulse response) filtering for alien frequencies and baseline trend are used to calculate HRV (RRI variability, ms²) in the very low frequency power band (VLF-HRV; 0.0075–0.075 Hz), the low frequency power band (LF-HRV; 0.075–0.125 Hz), and the high frequency power band (HF-HRV; 0.125–0.50 Hz). These bands of frequencies were selected to optimally adapt mathematical properties of the filtering method to the properties of the particular physiological processes. ^43–45 As the distributions of the HRV measures were skewed, natural logarithms (ln) of the LF and HF measures were used. Data processing for R-peak and BP-peak detection, artifact search, and baroreflex sequences was performed off-line using a custom computer program written by D.M.D. using the Spike2 system (Cambridge Electronic Design, Cambridge, England).

Statistical analysis

Descriptive and inferential analyses were performed with SPSS (SPSS Science, Chicago, IL) software using General Linear Models by the Type III method (GLM), SPSS built-in bootstrapping option for computing confidence intervals for regression estimates in GLM, and the SPSS macro command set ‘PROCESS’ to evaluate the significance of mediation effects. ⁴⁶ Values of p < .05 were regarded as statistically significant in all analyses. A percentile bootstrap procedure with 5000 bootstrap samples was used to generate 95% confidence intervals (CIs) of regression coefficients from empirical sampling distribution in GLM and mediation analyses. The bootstrap procedure was suggested as a robust alternative to inference based on parametric assumptions (such as normally distributed errors) to confirm findings obtained by parametric analyses and is recommended for reporting inferences in scientific reports. ^21,47 All parameter estimates were expressed as non-standardized (B) regression coefficients and their standard errors (SE) in the text. Where relevant, a partial η² was reported as a measure of strength of association (effect size), which is comparable to R² expressing the percentage of explained variance. Age (natural log(ln)-transformed value) and Body Mass Index (kg/m²) were included in all models.

Preliminary GLM Repeated Measures analyses were conducted to assess different effects of the walking/yoga interventions (Group effect) on CV changes from the pre- to post-treatment period to test their differential impact on energy-regulating mechanisms. These Group effects were also inspected for different mediation pathways within CV changes to inspect combined physiological mechanisms behind their simple associations (e.g., from Group to HR change through change in -BRS vs. from Group to -BRS change through change in HR: Group->-BRSΔ ->HRΔ vs. Group->HRΔ->-BRSΔ). All mediation effects were confirmed by a percentile bootstrap procedure included in the ‘PROCESS’ macro command set, as the bootstrapping technique more accurately captures the shape of the sampling distribution and therefore has greater power to detect mediation and to make speculations about probable causal mechanisms. ^48,49 Heteroscedasticity consistent standard errors were used in these analyses.

The next group of GLM Repeated Measures analyses was to test the hypothesis that individual pre-treatment differences in baroreflex and other CV measures (e.g., +BRS and −BRS in up and down sequences of RRI-SBP pairs) similarly or differentially predict response to the two interventions at post-treatment and 6-month follow-up. Group(2) simple and CV*Group(2) interaction effects with the CV measure (e.g., baseline +BRS value) as continuous and Group (Walking vs Yoga) as between-subject factors were assessed for the individual symptom outcomes. Symptom outcomes were assessed both as a simple difference score (baseline - post-treatment) and as a baseline adjusted difference score ([baseline - post-treatment]/baseline). The adjustment was used to control for absolute pain rating scales non-linearly and inter-individual inconsistency. ⁵⁰ For example, subjects may rate the present painful experience in comparison with prior ones or to a non-painful condition.

The bootstrap non-parametric procedure was used to inspect and confirm (validate) those effects that were found significant or with a tendency to significance by the parametric procedures. Final inferences of significance of the effects and relationships were based on results of the bootstrap procedure. The Johnson–Neyman (J-N) technique included in the same SPSS macro command set was used to detect regions of significant relationships in the cases of significant moderation (interaction) effects of pre-treatment CV values predicting different group effects on severity changes. Following the recommendations outlined by Rothman, ⁵¹ no compensations for the number of inferences (i.e., multiple testing correction like the Bonferroni test) were made in either model. Moreover, confidence intervals generated for each effect by the bootstrap procedure (a permutation test or a resampling-based method of inference) guaranteed family-wise type I error control associated with the multiple comparison problem. ⁵²

Results

Differences in demographic, medical, and psychological characteristics

Demographic details and descriptive statistics for the total sample and treatment conditions, as well as between- and within-group treatment effects on primary clinical outcomes are presented elsewhere. ¹¹ Shorter demographic details are presented in Table 1 of this paper ¹. Descriptive statistics of cardiovascular variables for the sample are presented in Supplementary materials 1 (Table 1S). Of the 27 participants who completed treatment, 20 completed 6-month follow-up questionnaires assessing IBS severity (Figure 1). Of these, 12 (60%) were from the yoga condition and 8 (40%) from the walking condition. ¹¹ No adverse events of the treatment procedures were reported.

Table 1.

Demographic, clinical, and psychological characteristics of the samples¹

Treatment groups:	Walking	Yoga
Characteristics	N=10	N=17
Constipation/Diarrhea	6/4	9/8
Sex, Female/Male	9/1	15/2
Age (year)	39.0 (15.0)	34.6 (11.6)
Body Mass Index (kg/m2)	25.7 (5.3)	24.0 (4.2)
Smoke (Yes/No)	0/10	1/16
Antidepressants (NDRI or SSRI)	1/9	2/15
HADS-Anxiety, Mean (SD)	8.40 (5.68)	8.82 (4.25)
HADS-Depression, Mean (SD)	4.80 (3.26)	5.00 (4.33)
HADS-General, Mean (SD)	13.20 (8.13)	13.82 (8.10)
PHQ15 (4 wks), Mean (SD)	12.41 (3.67)	14.70 (5.70)
Positive affect, Mean (SD)	26.40 (9.68)	23.35 (9.55)
Negative affect, Mean (SD)	15.30 (4.76)	16.12 (6.96)
Overall GI symptom intensity (7d), primary outcome, Mean (SD)	11.20 (4.54)	11.47 (4.32)
Abdominal pain intensity (7d), primary outcome, Mean (SD)	9.70 (4.67)	11.18 (3.88)
Symptom-free duration (from minutes [1] to 1 year [6]), Mean (SD)	3.2 (0.7)	3.47 (1.13)
IBS chronicity (years), Mean (SD)	17.17 (13.64)	13.81 (10.33)
Other chronic pain (Yes/No)	4/6	6/11
Visceral Sensitivity Index, Mean (SD)	35.0 (14.1)	44.2 (19.3)

¹means and comparisons (Chi-square test, Mann-Whitney test, and one-way analysis of variance) presented for raw (not transformed and not adjusted) data and approved by bootstrap 95% CIs

^*- p < .05

Abbreviations: GI, gastro-intestinal, IBS, Irritable Bowel Syndrome

HADS-General, -Anxiety, and -Depression - Hospital Anxiety and Depression Scale General, and its anxiety and depression subscales, for details see Supplementary materials 2

Positive and negative affects - subscales of PANAS-X (positive and negative affect schedule), for details see Supplementary materials 4

PHQ15 - Patient health questionnaire-15, for details see Supplementary materials 2

NDRI - norepinephrine–dopamine reuptake inhibitor; SSRI - selective serotonin reuptake inhibitor

Figure 1.

Flow diagram

Comparison of effects of different physical treatment procedures on CV changes from pre- to post-treatment periods

Significant Group effects were found for pre- to post treatment changes in baroreceptor sensitivity and baroreflex effectiveness to spontaneous inhibitions (-BRSΔ and -BEIΔ), power of HF and LF bands of HRV, and HR (Bs[SEs] = 3.96[1.63], −0.10[0.04], 0.51[0.19], −0.82[0.38], and −5.73[2.60], ts[p] = 2.44[0.023], −2.54[0.019], 2.67[0.014], −2.13[0.044], and −2.20[0.038], η2 = 0.21, 0.23, 0.25, 0.17, and 0.17; confirmed by bootstrap 95% CIs: 0.62–7.42, −0.18 to −0.04, 0.12–0.87, −1.72 to −0.04, and −10.78 to −0.24). The Yoga program led to a decrease of baroreceptor sensitivity and an increase of baroreflex effectiveness as outcomes (M[SD] = 1.13[3.77] and −0.04[0.10]). The Walking program was associated with the opposite response, an increase in sensitivity and a decrease in effectiveness in the baroreflex measures (M[SD] = −2.44[4.31] and 0.06[0.09]). The Walking program showed a higher increase in HF-HRV power (M[SD] = −0.39[0.46]) compared with the Yoga program (M[SD] = −0.04[0.45]) but a decrease in LF-HRV power (M[SD] = 0.12[0.91]) compared to an increase after the Yoga program (M[SD] = −0.39[0.93]). The Yoga program led to an increase of HR (M[SD] = −3.17[5.97]) while Walking led to a decrease of HR (M[SD] = 2.16[6.56]). Respective mediation analyses showed that the Group effects on baroreceptor sensitivity was related to both phasic (Group->ΔHF-HRV power->-BRSΔ; B[SE] = −2.24[1.25], bootstrap 95% CI: −4.97 to −0.15) and tonic (Group->-BRSΔ ->HRΔ; B[SE] = 3.47[2.12], bootstrap 95% CI: 0.21–8.42) regulations of vagus activity. Other Group effects were independent.

Simple predictive effects of CV measures for changes in symptoms

All simple predictive effects were obtained in the same models with respective interaction (moderation) analyses (see below). Higher pre-treatment baroreceptor sensitivity to spontaneous activations (+BRS) and related heart rate responses (+RRIS) predicted greater decreases in abdominal pain severity by early post-treatment period for those in both groups (Table 2). Higher pre-treatment baroreceptor sensitivity to spontaneous inhibitions (-BRS) and related heart rate responses (-RRIS) predicted greater abdominal symptoms relief by later 6-month follow-up in both groups (Table 3). Wider pre-treatment RRIs (slower HRs) at times of spontaneous baroreflex activations (+SPRR) or inhibitions (-SPRR) predicted, respectively, increase in symptoms-free duration and decrease in severity of overall GI symptoms and abdominal pain by early post-treatment period in both groups (Table 2). Lower effectiveness of pre-treatment baroreceptor activations (+BEI, baroreflex effectiveness for regulating heart rate decelerations) predicted longer duration of the disease and chronicity of pain (η² = 0.31, B[SE] = −31.3[10.4], bootstrap 95% CI: −55.3 to −10.5).

TABLE 2.

Size (η²), B and its bootstrap 95% CIs of significant simple and interaction effects of Group (‘walk/yoga’ between-subject factor) and physiological baseline variables at t1 on symptom severity and intensity changes ([t1-t2]/t1).

Independent Variables	Dependent variable changes
	Abdominal Pain Severity	Overall GI symptom intensity	Symptoms free duration
Model 11
Group	0.33√, 2.17, 0.03–3.48
+BRS	0.24√, 0.33, 0.04–0.60
Group*+BRS	0.34√, −0.24, −0.41 to −0.003	0.25√, −0.20, −0.34 to −0.01
Model 2
Group
+RRIS	0.26√, 0.04, 0.01–0.08
Group*+RRIS
Model 3
Group			0.14, −2.82, −6.88 to −0.24
+SPRR			0.16√, −3.34, −6.89 to −0.51
Group*+SPRR			0.13, 3.22, 0.26–8.33
Model 4
Group	0.42√, 8.67, 2.24–13.72	0.25√, 6.14, 0.55–10.37
−SPRR	0.42*, 8.59, 2.31–12.87	0.36*, 7.01, 2.70–10.02
Group*-SPRR	0.42√, −9.24, −14.62 to −2.33	0.26√, −6.64, −11.44 to −0.59

^√- bootstrap’s p < .05;

^*- bootstrap’s p < .01

¹- additionally adjusted for –BRS

Abbreviations:

GI, gastro-intestinal

+BRS - regression slopes (ms/mmHg) between changes of R-R intervals (RRI) and changes of associated beat-to-beat SBP when both increased (“+”, “up” sequences); a measure of baroreflex sensitivity

+RRIS - Regression slopes (gains) for up (+) and down (−) RRI ramps of cross-correlated SBP and RRI sequences (i.e., short-term bradycardiac baro-responses); a measure of gain of the efferent component of the baroreflex loop with baroreceptor activations

+SPRR, -SPRR - RR-interval (RRI) start points for baroreflex reactions to up (+) and down (−) SBP ramps; the first RRI value of RRI ramps in the baroreflex sequences: the highest RRI value in down sequences and the lowest RRI value in up sequences

TABLE 3.

Size (η²), B and its boostrap 95% CIs of significant simple and interaction effects of Group (‘walk/yoga’ between-subject factor) and physiological baseline variables at t1 on changes in abdominal symptoms relief (t2-t3).

Independent Variables	Dependent Variables
	Changes in Abdominal Symptoms Relief
Model 1
Group
−BRS	0.33, 0.15, 0.01–0.26
Group*−BRS
Model 2
Group
−RRIS	0.36√, 0.07, 0.003–0.13
Group*−RRIS
Model 3
Group	0.33, 10.11, 1.17–27.03
lnHF−HRV
Group*lnHF−HRV	0.32, −1.59, −4.38 to −0.14

^√- bootstrap’s p < .05

Abbreviations:

-BRS - regression slopes (ms/mmHg) between changes of R-R intervals (RRI) and changes of associated beat-to-beat SBP separately when both decreased (“-”, “down” sequences); a measure of baroreflex sensitivity

-RRIS - Regression slopes (gains) for down (−) RRI ramps of cross-correlated SBP and RRI sequences (i.e., short-term tachycardiac baro-responses); a measure of gain of the efferent component of the baroreflex loop with baroreceptor inhibitions

lnHF-HRV - natural log-transformed and non-transformed (ms²) high frequency heart rate variability power band

Moderation effects of pretreatment CV measures on specific outcomes for each treatment group.

Pre-treatment baroreceptor sensitivity to spontaneous activations was a significant moderator of post-treatment outcomes across the two groups for changes in severity of overall GI symptoms and abdominal pain (Table 2). The effects remained significant after an additional adjustment for the respective baroreceptor sensitivity to spontaneous inhibitions (-BRS). The Johnson–Neyman technique showed that participants with pre-treatment +BRS below 7 ms/mmHg had risk for pain aggravation in response to the Walking program, but participants with pre-treatment +BRS above 12 ms/mmHg were more sensitive to the Walking than to the Yoga program for decreasing overall GI symptoms, as well as visceral pain severity (Figure 2). This suggests that participants with extreme pre-treatment baroreflex values for spontaneous BP regulation could have opposite effect in response to the same intervention for managing these abdominal symptoms.

Figure 2.

Adjusted change in severity of chronic abdominal pain (a) and overall gastrointestinal (GI) symptoms (b) from pre- to post-treatment period ([t1-t2]/t1) depending on pre-treatment baroreceptor sensitivity to spontaneous activations (+BRS) and a treatment procedure (Yoga vs. Walking group).

Pre-treatment RRI set point associated with (i.e., RRI level at times of) spontaneous baroreflex inhibition was also a significant moderator of post-treatment outcomes across the two groups for changes in severity of overall GI symptoms and abdominal pain (Table 2). The Johnson–Neyman technique showed that participants with pre-treatment -SPRR<900 ms (i.e., with HR > 67 bpm at times of spontaneous baroreflex inhibitions) were more sensitive to the Yoga than to the Walking program for decreasing GI symptoms (Figure 3). However, participants with pre-treatment -SPRR>1000 ms (i.e., with HR < 60 bpm at times of spontaneous baroreflex inhibitions) were more sensitive to the Walking than to the Yoga program for decreasing GI symptoms. This suggests that the interventions may have opposite effects on GI symptoms and associated pain in people of slower or faster HR phenotypes at times of spontaneous baroreflex inhibition (i.e., at times for spontaneous baroreflex-related HR accelerations).

Figure 3.

Adjusted change in severity of chronic abdominal pain (a) and overall gastrointestinal (GI) symptoms (b) from pre- to post-treatment period ([t1-t2]/t1) depending on pre-treatment R-R interval set points associated with spontaneous baroreflex inhibitions (-SPRR) and a treatment procedure (Yoga vs. Walking group).

Pre-treatment RRI set point associated with (i.e., RRI level at times of) spontaneous baroreflex activations was a significant moderator of post-treatment outcomes across the two groups for a change in symptom-free duration (Table 2). The Johnson–Neyman technique showed that participants with pre-treatment +SPRR<830 ms and +SPRR>1000 ms (i.e., with HR>72 bpm and HR<60 bpm at times of spontaneous baroreflex activations) were more sensitive to the Walking program for decreasing or increasing symptom-free duration, respectively (Figure 4). This suggests that the Walking program may have a specific regulation effect that shortens periods without symptoms of disease in people with a faster HR phenotype at times of spontaneous baroreflex activations (i.e., at times for spontaneous baroreflex-related HR decelerations) but prolongs them in people with a slower HR phenotype.

Figure 4.

Adjusted change in symptom-free duration from pre- to post-treatment period ([t1-t2]/t1) depending on pre-treatment R-R interval set points associated with spontaneous baroreflex activations (+SPRR) and a treatment procedure (Yoga vs. Walking group).

Pre-treatment HF-HRV was a significant moderator of outcomes across the two groups for a change in abdominal symptom relief/aggravation from post-treatment to 6-month follow-up period (Table 3). The Johnson–Neyman technique showed that participants with pre-treatment HF-HRV>1300 ms² (i.e., with higher vagus activity) were more sensitive to the Walking than to the Yoga program for increasing abdominal symptom relief at 6-month follow-up (Figure 5). However, participants with pre-treatment HF-HRV<500 ms² (i.e., with very low vagus activity) were more sensitive to the Yoga than to the Walking program for increasing abdominal symptom relief at 6-month follow-up after the treatment completion.

Figure 5.

Change in abdominal symptom relief/aggravation from post-treatment to 6-month follow-up period (t1-t2) depending on pre-treatment power of the high frequency band of heart rate variability (HF-HRV, spontaneous vagus activity) and a treatment procedure (Yoga vs. Walking group).

Discussion.

This study had two primary aims. First, to examine cardiovascular changes associated with two different exercise-based interventions for IBS, and second, to determine if pre-treatment CV measures are predictive of clinical outcomes from these interventions. The results for the first aim showed the Walking and Yoga interventions had very different outcomes in terms of cardiovascular regulation. The Walking program was associated with increases in parasympathetic tone (increase of baroreceptor sensitivity to inhibitions coupled with decrease of HR) and spontaneous vagal reactivity (increase of baroreceptor sensitivity to spontaneous inhibitions coupled with the increase of power of high frequency component of heart rate variability), but inhibited central processing of baroreceptor signals associated with blood pressure fluctuations induced by respiratory and vasomotor effects (decrease of number of responses to spontaneous baroreceptor inhibition regulating heart rate accelerations and power of low frequency component of heart rate variability). In contrast, the Yoga program led to decreased parasympathetic tone (decrease of baroreceptor sensitivity to inhibitions coupled with increase of HR) but improvement of the effectiveness of central processing of baroreceptor signals (increase of number of responses to spontaneous baroreceptor inhibition regulating heart rate accelerations and power of low frequency component of heart rate variability). Thus, the Yoga program was an intervention with a constraining effect on vagus efferent responses in favor of a more released central processing of vagus afferent signals, while the Walking program had a constraining effect on vagus afferent signal processing in favor of more powerful vagus efferent responses. A similar effect of Iyengar Yoga on vagal reactivity (a decreased power of high frequency component but an increased relative power of low frequency component of heart rate variability) was found in a study of depressed patients showing symptom improvement. ³⁵ Phasic and tonic vagal or parasympathetic mechanisms are separated in further discussions below, because in healthy subjects at rest, the heart rate is usually under tonic parasympathetic influence, mainly to limit energy expenditure in general, but respiratory and baroreflex-related heart rate fluctuations are largely related to phasic parasympathetic regulations of energy expenditure challenged by spontaneous central and peripheral arousal fluctuations. ⁵³

Findings of the study also support the hypothesis that pre-treatment cardiovascular (CV) variables are predictive of clinical outcomes following the exercise-based interventions for IBS. While some CV variables were predictive of similar outcomes for both interventions, others predicted different outcomes for each intervention. Moreover, while some CV variables predicted mental health outcomes (see Supplementary materials 2), others predicted somatic or physical health outcomes. The latter findings were the main objective of the study. They showed that early and 6-month follow-up somatic severity relief including pain, as well as an increase in symptom-free duration in response to physical activity interventions in general, were greater in those with higher pre-treatment vagal reactivity and tone associated with baroreflex activations and inhibitions (baroreceptor sensitivity to spontaneous activations/inhibitions and related heart rate levels and responses), respectively. The differential relationship of baseline CV phenotype to IBS symptom changes from the two interventions was independent of CV phenotype prediction for affect outcomes. In contrast to somatic symptoms, a mental improvement (decrease in severity of depressive symptoms, general somatic complaints, and negative affect in general) in response to both interventions was predicted by a lower sympathetic tone (assessed by a lower pre-treatment DBP) (see Supplementary materials 2).

In response to the Yoga program, an intervention with released afferent but restricted efferent vagal regulation, participants with lower pre-treatment baroreceptor sensitivity to spontaneous activations (i.e., a low vagal reactivity or phasic vagal activity) during rest demonstrated a somatic health improvement by the end of the intervention compared with the Walking program. This prediction of somatic improvement in response to the Yoga program did not correspond with relative mental health improvement (e.g., relief in general and gut-specific anxiety) that was indicated in response to the Walking program for the same CV phenotype with the restricted afferent regulation of baroreflex activity (see Supplementary materials 2). Participants with the opposite CV phenotypes (higher pre-treatment baroreceptor sensitivity to spontaneous activations, i.e., with a higher vagal reactivity or phasic vagal activity during rest) demonstrated a somatic health improvement in response to the Walking program. This prediction of somatic improvement in response to the Walking program did not correspond with mental health improvement (e.g., relief in general and gut-specific anxiety) that was indicated in response to the Yoga program for the same CV phenotype with released afferent regulation of baroreflex activity (see Supplementary materials 2).

Both Yoga and Walking programs also decreased somatic symptoms including pain severity in participants with another pair of opposite CV phenotypes related to respectively lower or higher vagal tone during rest at times of spontaneous vagally-mediated HR accelerations (higher or lower resting heart rate at times of spontaneous baroreflex inhibitions). A stability in health condition (i.e., a period without symptoms) could be maintained by the Yoga program in participants with CV phenotype also associated with a lower vagal tone (a higher resting heart rate at times of spontaneous baroreflex activations for HR decelerations). Moreover, a high long-term (6-month) effectiveness of the Walking and the Yoga programs for abdominal symptom relief including pain was also associated with respectively higher and lower vagus activity assessed by a high frequency band of heart rate variability immediately after the termination of the treatments. In general, the effectiveness of the Yoga program for somatic symptoms was less depended on CV phenotype difference compared with the Walking program, but only in short-term.

Thus, the improvement in somatic component of chronic pain requires such interventions like the Walking program (i.e., an intervention applying on efferent regulation) to be used in people with higher phasic vagal reactivity and higher vagal tone phenotypes, but such interventions like yoga (i.e., an intervention applying on afferent vagus regulation) to be used in people with lower phasic vagal reactivity and vagal tone phenotypes. However, the improvement in mental component of chronic pain requires other predispositions: the efferent vagus regulation associated with such interventions like walking in people with lower parasympathetic/sympathetic reactivity and tone phenotypes, but the afferent vagal regulation associated with such interventions like yoga in people with higher parasympathetic/sympathetic reactivity and tone phenotypes (see Supplementary materials 2). The same mood improvement (remission) in depressed patients with a CV phenotype of higher parasympathetic reactivity (high power of high frequency band of heart rate variability and high baroreceptor sensitivity) and tone (low heart rate) was indicated in response to the Yoga program in a previous study. ³⁵

The above findings suggest that the prescription of interventions, based on CV phenotypes might help improve outcome for predominant somatic or affective symptoms of this chronic pain syndrome. However, the findings also predict that incorrect predominant management of somatic or affective complaints in patients with mainly an affective or somatic mechanism of chronic pain, respectively, will not improve effectiveness of the intervention. Thus, these results are in accord with previous considerations that impacts of affective and somatic components of chronic pain on the patient’s condition should be correctly distinguished for their effective management. ^20,22,23

A primary limitation of the study is the relatively small sample size. This was addressed in part by relying of final inferences on effects obtained for both over-lapping outcome variables (abdominal pain and general GI complaints) using the bootstrap procedure (a resampling-based method of inference) protecting against false-positive results, as well as interpreting the results in the context of their relevance to proposed physiological mechanisms associated with yoga as an intervention with released afferent but restricted efferent vagal regulation and walking as an intervention with released efferent but restricted afferent vagal regulation. However, the findings should be replicated in larger samples of patients with IBS as well as other chronic pain syndromes before their implementation in clinic and thus they should be evaluated as pilot. Other limitations include the restriction of the Yoga treatment to one school, that of Iyengar yoga, the inability of the study to adequately examine sex differences in responses and the lack of a no treatment control condition.

Thus, the effectiveness of IBS and visceral pain treatment might be improved by personalizing interventions using CV phenotyping of patients and knowledge of specific effects of the interventions on CV regulation. Specifically, this study found that chronic pain severity management should be different in participants with different pre-treatment CV phenotypes associated with BRS regulation of BP. Understanding the molecular and signaling mechanisms that explain all these findings could provide ‘mechanistic’ insight into the health benefits of different physical activity practices and even propose new drugs or compounds for the prevention and treatment of chronic pain. At a speculative level, if the biomarkers obtained in the study are replicated and confirmed, it would be possible to develop exercise training programs matched to these biomarkers to maximize a given individual’s somatic or affective health responses to training.

Key Points.

• Identification of reliable biological markers that could help to predict who will benefit from a specific intervention is a key challenge in developing personalized medicine approaches to chronic pain.

• This study found that chronic pain management in Irritable Bowel Syndrome should be different in patients with different physiological regulation of blood pressure.

• Chronic pain is the most prevalent symptom driving medical visits. Effectiveness of chronic pain treatment could be improved by personalizing interventions using cardiovascular biomarkers.