Effectiveness of spinal manipulation in influencing the autonomic nervous system - a systematic review and meta-analysis

ABSTRACT

Background: Spinal manipulation (SM) has been hypothesized to influence the autonomic nervous system (ANS). Further, it has been proposed that the effects may vary depending on the segment manipulated. The aim of this systematic review was to synthesize the current level of evidence for SM in influencing the ANS in healthy and/or symptomatic population. Methods: Various databases (n = 8) were searched (inception till May 2023) and 14 trials (n = 618 participants) were included in the review. Two authors independently screened, extracted and assessed the risk of bias in included studies. The data were synthesized using standard mean differences and meta-analysis for the primary outcome measures. The Grading of Recommendations, Assessment, Development and Evaluation (GRADE) was used for assessing the quality of the body of evidence for each outcome of interest. Results: Overall, there was low quality evidence that SM did not influence any measure of ANS including heart rate variability (HRV), oxy-hemoglobin, blood pressure, epinephrine and nor-epinephrine. However, there was low quality evidence that cervical spine manipulation may influence high frequency parameter of HRV, indicating its influence on the parasympathetic nervous system. Conclusion: When compared with control or sham interventions, SM did not alter the ANS. Due to invalid methodologies and the low quality of included studies, findings must be interpreted with great caution. Future studies are needed which employ rigorous data collection processes to verify the true physiological implications of SM on ANS.

Introduction

Spinal manipulation (SM), or a high-velocity low amplitude thrust, is a precise joint technique that involves a quick thrust at the end of the joint limit, often resulting in cavitation [1]. The technique is performed to induce mechanical stimulation to the joint to improve mobility, reduce pain, and restore regular movements [2]. Despite the widespread use and documented benefits of SM, its underlying biological mechanism remains unclear [3]. Therefore, a more robust body of evidence investigating the mechanisms behind SM may increase acceptability in the healthcare community.

It is now commonly accepted that SM can influence various mechanisms at the peripheral, spinal, and supraspinal levels [1,4,5]. Further, evidence also demonstrates systemic effects of SM like anti-inflammation [6] and changes in biochemical markers such as substance-p, cortisol and neurotensin [5]. Importantly, there is consensus within manual therapy researchers/clinicians that SM may influence the autonomic nervous system (ANS) [7–9]. The reason for this consensus is due to the anatomy of the ANS – with its ganglions close to the spine [4], studies have demonstrated concurrent hypoalgesia and sympathetic nervous system (SNS) modulation following SM [10,11]. The ANS ganglion’s location has also led to hypothesis that the ANS response may vary depending on the segment of manipulation. Specifically, a cervical SM may result in parasympathetic nervous system (PNS) response and a thoracic SM might result in SNS response [12].

Studies exploring the effects of SM on the ANS use a variety of outcome measures, with heart rate variability (HRV) as the most common index of ANS status [13]. The two branches of the ANS, the PNS and SNS, dually innervate the heart and their neural pathways largely control cardiac functions, including heart rate [14,15]. HRV, the variation in the time interval between each heartbeat, is an indirect measure of the dynamic coordination of PNS and SNS neural effects on the heart during stress and recovery [16,17]. While HRV is believed to represent both sympathetic and parasympathetic influence, it is more strongly mediated by parasympathetic control [14,17]. There are over 25 measures of HRV that are derived and analyzed from resting heartrate data [15,18]. For a comprehensive review about HRV, please refer to [16]. Time and frequency-based measures of HRV are most often studied given their direct correlation with cardiac ANS activity [15,18]. Common time domain measures include the standard deviation of Normal-to-Normal (NN) intervals; number of consecutive NN intervals that differ more than 50 ms (NN50); and the root mean square of consecutive Time elapsed between two successive R Waves (RR) intervals (rMSSD). Frequency domain measures include high frequency (HF) and low frequency (LF) [14]. HF is a direct correlate of rMSSD, and both HRV indices are strong, reliable indicators of beat-to-beat changes at the heart primarily modulated by the PNS [16]. In contrast, the validity and reliability of the LF component remains unclear, with some studies suggesting its modulation to be a combination of parasympathetic and sympathetic activity [19]. HF is suggested to be more responsive to parasympathetic than sympathetic activity [20], and is more sensitive to changes in ANS function. For these reasons, researchers disagree that the LF/HF ratio reflects parasympathetic and sympathetic balance (also coined as ‘sympathovagal balance’) due to several confounding factors and questionable validity affecting the frequency-based measure of LF [16].

At rest, the autonomic regulation of skeletal muscle blood flow and/or tissue oxygenation is dominated by the SNS [21]. Skeletal blood flow, therefore, may represent an alternate and reliable method to measure SNS activity. One reliable, noninvasive technique to measure musculoskeletal blood flow is the use of Near Infrared Spectroscopy (NIRS) [22–24] which can measure oxygenated hemoglobin and tissue oxygenation index (TOI). Other approaches to quantifying SNS activity include blood pressure (BP) [25], pupillary reflexes [26], muscle sympathetic nerve activity (MSNA) [27], and blood concentration of epinephrine/nor-epinephrine [28].

Previously published systematic reviews have attempted to determine the effects of SM on ANS in humans (healthy and symptomatic) [20,29–32]. However, SM was not the sole intervention included, and the findings were largely inconclusive [20,29–32]. Additionally, a holistic look at the effects of SM on ANS must include measures of both PNS-dominated indices (e.g. HRV) and SNS-dominated measures (e.g. NIRS, MSNA). Therefore, the primary aim of this systematic review was to evaluate and synthesize existing evidence on the effects of SM on the ANS. The secondary aim was to determine whether evidence is available to support or refute a differential response of the ANS to manual therapy based on the region of the spine (e.g. cervical, thoracic, and lumbar) manipulated.

Methods

This review has been reported according to Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines [33]. The review protocol was prospective registered and can be found on the International Prospective Register of Systematic Reviews (CRD42019120152).

Eligibility criteria

Participants: Only human studies were eligible. There were no restrictions on age, type of condition, or gender.

Intervention: SM was the intervention of choice and is defined as a high-velocity, low-amplitude (HVLA) thrust technique, commonly accompanied by a cavitation or ‘pop’ sound and performed manually by a practitioner and not a device [34].

Comparator: comparator can be any of the following: no intervention (control), sham, placebo, any other intervention (e.g. exercise).

Outcomes: Measures of ANS could include HRV, skin conductance, BP, skin blood flow, skeletal muscle blood flow, tissue oxygenation, catecholamine, adrenaline, nor-adrenaline, heart rate, pupil diameter, edge light pupil cycle time (ELPCT), MSNA, and salivary-α-amylase.

Types of Studies: Only randomized controlled trials (RCT) or controlled clinical trials (CCT) were included in the review. Studies involving humans (healthy or painful condition or disease) in which outcome measures of interest were used were included. The review only included studies published in the English language.

Information sources and search strategy

The following electronic databases were searched from inception to May 2023: Scopus, Cochrane Library, Cumulative Index to Nursing and Allied Health Literature, Allied and Complementary Medicine Database, PubMed, Excerpta Medica Database, and Physiotherapy Evidence Database An additional search was also undertaken on ProQuest (Dissertations and Theses) to include ‘grey literature.’ A comprehensive search strategy was established to identify studies relevant to the subject of this research question: interventions (SM) and outcome (ANS measures). Keywords such as ‘manual therapy’ and ‘orthopaedic physical therapy’ were used for this purpose (see Appendix A). The Boolean operators ‘OR’ and ‘AND’ were used between and within keywords as part of the search strategy.

Selection process

Articles generated by the search were saved into systematic review management software covidence^TM (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia; www.covidence.org). Covidence is a web-based collaboration software platform that streamlines the production of systematic and other literature reviews. The automation tool available in covidence^TM identified and removed duplicate articles. Two reviewers (KSK, SB) performed independent screening of the titles and abstracts of studies; followed by retrieval of full texts of studies that met the inclusion criteria. Any disagreements were resolved through discussion and consensus. A third reviewer (ASG) was available (though not required in this instance) in case of disagreements. The review team has more than 10 years (mean) of experience in undertaking systematic reviews with more than 15 years (mean) of clinical experience.

Data collection

Two reviewers extracted the data independently. A data extraction table/form was used to standardize this procedure. The data obtained from each included trial included: (1) Study aim; (2) Study population (including the number of participants, age, sex, and duration of disorder); (3) Details of interventions (SM including treatment dosage, the region of manipulation and the control group intervention); (4) Types of outcome measures (ANS measures); equipment used to measure the outcome measures; and measurement time points; (5) Study findings and author’s conclusions.

Risk of bias

The Risk of Bias was independently assessed by two reviewers (KSK and ASG) using the Cochrane Collaboration’s tool for assessing risk of bias [35] available as part of Covidence. Any disagreement between reviewers was resolved through consensus. A third reviewer (LW) was available but was not required in this instance. The criteria were scored as ‘yes,’ ‘no’ or ‘unclear’ and recorded in the Risk of Bias table. Studies that met at least six of the 12 criteria (categorized as ‘no serious flaws’) were considered to have a ‘low’ risk of bias.

Effect measures

In the event where pooling of data was appropriate, a meta-analysis was conducted. For this review, study data taken up to 2 h post-intervention were included. In the event where multiple measurements were taken, the last data within the above time period was used in analysis. In cases of missing data, every attempt was made to contact the author to obtain unpublished results. The mean and standard deviation were entered into the RevMan Web Software [36], where the data were compared to obtain an overall effect size of the intervention. All pooled data were analyzed with a random effect model. Standardised Mean Differences (SMD) between treatment groups were calculated with a 95% confidence interval. The effect size was generated (Cohen’s d; small ≥ 0.2 medium ≥ 0.5 and large ≥ 0.8) [37], for each outcome measure. The overall effect was also adjusted to account for the level of heterogeneity between studies, which was determined by comparing the relevant factors (outcome measures, type of intervention, participant characteristics) of each study.

Assessment of reporting bias

In the instance where more than 10 studies are to be included in a meta-analysis, a funnel plot should be used to test for publication bias. However, this was not performed as less than 10 studies were eligible to be included in meta-analysis in this study.

Certainty assessment

The overall methodological quality of evidence (very low, low, moderate, and high) was graded using the Grading of Recommendation, Assessment, Development, and Evaluation (GRADE) system [38].

Results

Results of the search

An initial search retrieved a total of 734 articles. The duplicates were removed, leaving 523 articles. After title and abstract screening, 78 articles were selected. Fifty-nine articles were excluded using the inclusion and exclusion criteria. After screening full texts of 23 articles, 14 articles were included in this systematic review with 10 eligible for inclusion in the meta-analysis (refer to Figure 1).

Figure 1.

PRISMA flowchart of included studies.

Included studies

The characteristics of included studies are elaborated in ‘Characteristics of Included Studies’ table (refer to Table 1).

Table 1.

Characteristics of included studies.

Author, Year	Methods/participants characteristics	Intervention	Outcome measure(s)/time point	Findings
Budgell and Hirano, 2001	Two group RCT, randomized crossover design; 25 young adults; 20 men five women; Age 28$.5\pm$6; Setting: Chiropractor Clinic	Upper Cervical Manipulation (vs) Sham Manipulation Upper cervical Manipulation group: C1–2 Manipulation Sham: Identical positioning without thrust	Heart Rate (bpm) and HRV Time: Pre-manipulation and 5 min post manipulation	Statistically significant decrease in HR (p = 0.0496), LF/HF ratio (p = 0.0037) and LF (p = 0.031) for the intervention group
Budgell and Polus, 2006	Two group RCT, randomized crossover design; 28 young adults; 23 men, five women; Age: 29 ± 7; Setting: Chiropractor clinic	Thoracic Spinal manipulation (vs) Sham Manipulation Intervention: Thoracic Spinal Manipulation (T1–4) Sham: Light push on the bilateral scapula	Heart Rate (bpm) and HRV Time: Pre-manipulation and 5 min post-manipulation	Statistically significant increase in normalized LF of the intervention group (p = 0.0201) Statistically significant decrease in normalized HF of the intervention group (p = 0.0043)
Gosling et al. 2005	Three group RCT; Randomized;30 adults; 10 participants each group; Mean age: 23.8; Setting: Controlled university setting	Upper cervical rotatory manipulation (vs) Sham Cervical Manipulation: Left C1/2 and Right C1–2 Sham: Identical positioning without thrust	Edge light pupil cycle time (ms) Time: Pre-manipulation and immediately post-manipulation	Statistically significant decrease in edge light pupil cycle time of right eye (t = 5.124, p = 0.001) and left eye (t = 3.067, p = 0.013) in intervention group compared to sham
Injeyan and Budgell, 2022	2 group RCT; Randomized; 31 young male and female adults; Age: 30 + 8.1 years, intervention group vs 34 + 9.9 years sham group; range = 20–40; Manipulation, 1 Sham Group Setting: Controlled university setting	Cervical Spinal Manipulation (vs) Sham Intervention: Cervical Manipulation (at potential site of joint dysfunction in mid-lower cervical spine) + Sham Sham Group: Sham + Sham; no thrust to spine	Heart Rate (bpm) and HRV Time: 5-min pre-intervention, intervention (several seconds), 5-min post-intervention	N = 29 (2 data sets were discarded because of frequent ectopic beats) No statistically significant effect on calculated measures of HRV. Post-intervention HR small, statistically significant decrease compared to pre-intervention HR at both early post-intervention (authentic 65.0 $\pm$ 8.3 vs sham 62.4 $\pm$ 6.6) and late post-intervention (intervention 64.8 $\pm$ 8.6 and sham 61.9 $\pm$ 6.8)
Kovanur Sampath et al. 2017	2 group RCT; Randomized; 24 men; Age: 18–45; 1 Manipulation, 1 Sham group Setting: Controlled laboratory	Thoracic Spinal Manipulation (vs) Sham Intervention: Thoracic spinal manipulation (T5) Sham group: Identical positioning without thrust	Salivary Cortisol and testosterone, T/C ratio, HRV, Oxyhemoglobin concentration in the right calf muscle Time: Measurements were taken at 5 and 30 min post-intervention	Significant between-group difference of salivary cortisol 5 min (0.35, p < 0.01) and T/C ratio 6 h (−0.09, p < 0.01) in the intervention group
*Kovanur Sampath et al. 2021	Randomized 2-sequence, 2-period crossover trial (washout 1 week); 24 participants; Age 48 $\pm$7; individuals with Achilles Tendonopathy Setting: University laboratory	Thoracic Spinal Manipulation (vs) Sham Intervention: thrust using ‘pistol grip’ in supine at level of T5 Sham: Identical position though no fixation of hand against thoracic spine nor thrust or rolling of the patient	Testosterone/Cortisol (T/C) ratio via salivary samples, HRV, and total oxygenation index (TOI) of the calf and Achilles tendon using near-infrared specstroscopy (NIRS) Time Salivary samples: pre-intervention and at 5-min, 30-min, 6-h post-intervention HRV: pre-intervention and 30 min post-intervention TOI: pre-intervention and a 5- and 30- min post-intervention	Statistically significant condition by time interaction was found for the T/C ratio (mean difference: −0.16; P < .05) and for TOI (mean difference: 1.35; P < .05) of calf muscle but not for Achilles tendon (P = .6); No statistical difference was found for HRV (P = .5).
Minarini et al. 2018	2 group RCT; Randomized and single-blinded; 73 participants; 37 in intervention, 36 in control; Age 18–45; Setting: University Clinic	Thoracic Spinal Manipulation (vs) Sham manipulation Thoracic manipulation: T2/3, T5/6, T11/12 Sham: Identical position with gentle compression of chest	Root mean square of successive difference (rMSSD) through Heart Rate Value Time: 60 s of continuous measurement pre and post manipulation	Statistically significant increase in rMSSD in the intervention group T2 24.34 ± 14.84 (p < 0.05); T5 23.09 ± 15.77 (p < 0.05); T11 14.54 ± 8.3 (p < 0.05)
Picchiottino et al. 2020	Randomized, cross-over (washout 48 h, sham-controlled trial; 51 healthy participants- 41 fully analyzed for HRV (age 19.9 $\pm$ 3.5); 30 fully analyzed for BP (age 20.2 $\pm 3.9)$ Setting: Controlled laboratory	Thoracic Spinal Manipulation (vs) Sham Maniipulation Intervention: spinal manipulation thrust at T5 level Sham: thrust in the plane of the scapulathoracic interface, not inducing spinal motion	HRV, Blood Pressure Variability, Pressure Pain Threshold Time: baseline and repeated 3 times (on average every 12 min) post-intervention	N = 41 fully analyzed No statistically significant differences found between spinal manipulation and sham and HRV nor BP post-intervention No relationship found between autonomic activity and pressure pain threshold post intervention
Puhl et al. 2012	2 group RCT; Randomized; 36 participants; 18 manipulation, 18 sham; Age: 26.2 ± 1.1; Setting: No described	Thoracic Spinal Manipulation (vs) Sham Intervention: Thrust manipulation at hypomobile segment between T1–6 Sham: Identical positioning without thrust	Plasma Concentration of epinephrine and norepinephrine Time: Immediately and 15-min post manipulation	No significant change in mean plasma concentration of epinephrine (p = 0.65) and norepinephrine (p = 0.36) in the intervention group compared to sham
*Rodrigues et al. 2021	3-arm, parallel, Randomized superiority trial, placebo-controlled, assessor-blinded trial; 59 patients with musculoskeletal pain; Age 40 $\pm$ 17 (spinal manipulation), 48 $\pm$ 11 (myofascial manipulation), 38 $\pm$13 (placebo) Setting: Physical Therapy Outpatient Clinic	Thoracic Spinal Manipulation (vs) Myofascial Manipulation (vs) Sham Intervention: Thrust manipulation focusing on upper thoracic spine Intervention: Myofascial manipulation to upper thoracic region Sham: Therapeutic ultrasound with no current output to upper thoracic spine	Resting HRV, Blood Pressure response to a sympathoexcitatory stimulus (cold pressor test-CPT) Time: Pre and Post interventions or sham	Only the spinal manipulation group demonstrated a significant difference, with an increase in pre-post intervention HRV (RMSSD, ms = 0.024) No significant differences in blood pressure pre and post intervention.
*Roy et al. 2009	5 group RCT; Randomized; 51 participants, 10 in each group; 1 control, two treatment, two sham; Setting: Chiropractor clinic	Treatment (vs) Sham (vs) Control Pain-free treatment group: Activator Method Chiropractor Technique Pain-free sham group: Identical positioning with a gentle push Pain treatment group: Traditional lumbar roll manipulation Pain sham group: Identical position with 5 s pressure without thrust	Heart Rate Variability Time: Measured 5-min pre- and 5- min post-treatment	There was a significant between-group change in VLF (p = 0.001) and HF(p = 0.002) in the pain treatment group compared to the pain sham group
*Sillevis et al. 2010	2 group RCT; Randomized; 100 volunteers with chronic cervical pain; Mean age 42.7 in the treatment group and 46.84 in the placebo group;	Thoracic Spinal Manipulation (vs) Control Intervention: Thoracic Spinal Manipulation (T3/4) Sham: Identical placement without thrust	Changes in pupil diameter Time: Measurements are taken just before manipulation, 1 min after manipulation and 4 min after manipulation	There was no statistically significant difference in pupil diameter post manipulation in the intervention group compared to sham There was a statistically significant within-group increase in pupil diameter in the sham group (p < 0.001)
Ward et al. 2013	3 group RCT; Randomized; 36 chiropractic student volunteers; Mean Age 26.8 ± 4.6; Setting: Controlled laboratory setting	Thoracic Spinal Manipulation (vs) Activator Method (vs) Sham Intervention group: T1–4 upper thoracic manipulation Activator: Activator based placebo Shame: Identical positioning without thrust	Changes in blood pressure variable and pulse oximetry Time: Measurements were taken before manipulation, 1 min after manipulation, 10 min and 24 h later	There was no statistically significant difference between group and within-group of both intervention and placebo group (p = 0.554)
*Ward et al. 2015	2 group RCT; 50 hypertensive participants; Mean age 45.5 ± 13.9; Setting: Controlled laboratory setting	Thoracic Spinal Manipulation (vs) Sham Treatment Group: T1–4 upper thoracic manipulation Sham: Identical positioning without thrust	Changes in blood pressure variable and pulse oximetry Time: Measurements were taken before manipulation, 1 min after manipulation, 10 min after manipulation	Statistically significant change within the group for PR interval in the control group (P < 0.047) and manipulation group (p < 0.046) Statistically significant within-group change in QRS duration in the manipulation group (p < 0.025)

*Studies that enrolled symptomatic participants.

Study methods

All studies included in this review were RCTs. Of the 14 studies, 10 were two-group RCTs [7–9,26,28,39–43], three were three-group RCTs [44–46], and one was a five-group RCT [47].

Sample size

There were 618 participants included in the studies. The number of participants in the studies ranged from 24 to 100, with six studies recruiting 50 or more participants [9,26,41,43,45,47].

Participants

All but one study recruited both female and male participants; Sampath et al. (2017) [42] recruited only male participants. Of the 14 studies, 2 studies recruited patients with chronic neck pain [26,47], 1 study recruited hypertensive subjects [43], 1 study recruited patients with Achilles tendonopathy [40], 1 study recruited adults with musculoskeletal pain [45], while the remaining 9 studies only studied asymptomatic participants.

Interventions

The studies used four types of interventions: (1) Cervical spine manipulation (either on the cervical spine or the atlantoaxial joint); (2) thoracic spine manipulation (either by therapist discretion or fixed segment as per study protocol); (3) lumbar spine manipulation (therapist discretion), and (4), thoracic myofascial manipulation. Of the 14 studies, three performed cervical manipulation as their intervention [7,39,44], while 10 studies used thoracic manipulation [8,9,26,28,40–43,45,46]. Only one study performed a lumbar spine intervention as their treatment [47].

Outcome measures

The most common outcome measure was HRV, with 9 of the 14 studies measuring different parameters of HRV [7–9,39–42,45,47]. Frequency domain indices of HRV (LF/HF ratio and HF) were measured in eight studies [7–9,39,40,42,45,47], whereas four studies measured the time domain-based index, rMSSD [9,40,41,45]. No study used skin conductance as an outcome measure. Other methods to quantifying the ANS included ELPCT [44], pupil diamerter [26], and epinephrine/norepinephrine [28]. Four studies [9,25,43,45] utilized BP changes as an indirect measure of ANS.

Safety

No adverse events were reported by any study.

Excluded studies

After screening study abstracts and full texts, some studies were excluded [32,48–52]. Further details regarding why these studies were excluded can be reviewed in Appendix B.

Risk of bias in included studies

A summary of the methodological quality of all studies is presented in Figure 2. Except for two studies [7,8], randomization and allocation concealment were considered adequate in all included studies. Owing to the nature of manual therapy interventions, participant blinding was difficult. Therefore, the bias in participant blinding was rated ‘high risk’ for all studies. However, blinding of assessors was explicit in six studies [9,39,40,42,44,45]. There was minimal attrition in all studies, and data were mostly complete. There was no significant cause for concern in publication bias.

Figure 2.

Risk of bias in included studies.

Effects of intervention

HRV (frequency domain)

SM (vs) control/sham in influencing LF/HF ratio of HRV

Data extracted from eight studies [7–9,39,40,42,45,47] (345 participants) compared the effects of SM, against control, on the LF/HF ratio. ‘Low’ quality evidence (Table 2) demonstrated that SM did not influence the LF/HF ratio of HRV when compared with control intervention (SMD 0.11; 95% CI: −0.11 to 0.32) (Figure 3a).

Table 2.

Summary of findings (GRADE): effects of SM on HRV.

Certainty assessment							№ of patients		Effect		Certainty
№ of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	SM	Control/Sham	Relative (95% CI)	Absolute (95% CI)
HRV: Time domain: rMSSD (follow-up: upo to 2 h; assessed with: ECG/HRV Monitor)
3	Randomised trials	Nat serious	Seriousa,b,c	Seriousc	Seriousd	None	83	81	-	SMD 0.09 SD higher (0.22 lower to 0.40 higer)	⊕◯◯◯ Very low
HRV: Frequency domain (LF/HF) (follow-up: upto 2 h; assessed with: ECG/HRV Monitor)
8	Randomised trials	Not serious	Seriousb	Not serious	Seriousb,d	None	173	170	-	SMD 0.07 SD higher (0.14 lower to 0.29 higher)	⊕⊕◯◯ Low
HRV: Frequency domain (HF) (follow-up: upto 2 h; assessed with: ECG/HRV Monitor)
6	Randomised trials	Not serious	Seriousb	Not serious	Seriousb,d	None	137	137	-	SMD 0.12 SD lower (0.49 lower to 0.25 higher)	⊕⊕◯◯ Low

CI: Confidence interval; SMD: Standardised mean difference.

Explanations: a. Lack of consistent, statistically significant result; b. Heterogeneity; c. The outcome measure used can be influenced by other confounding factors; d. Small studies, total sample size < 200.

Figure 3a.

Forest plot of the effects of spinal manipulation on LF/HF ratio.

SM (vs) control/sham in influencing HF of HRV

A ‘low’ quality evidence (Table 2) from six studies (274 participants) [7–9,39,45,47] demonstrated that SM is not significantly better than control/sham in influencing HF of HRV (SMD −0.12; 95% CI: −0.49 to 0.25)

HRV (time domain)

SM (vs) control in influencing rMSSD of HRV

A ‘low’ quality evidence (Table 2) from three studies (n = 164 participants) [9,40,45] demonstrated that SM was not significantly better than control/sham in influencing HF of HRV (SMD 0.09; 95% CI: −0.22 to 0.40). One study [41] reported rMSSD as a measure of HRV, however, including the study in meta-analysis was problematic as the study investigated the effects of manipulation on three different thoracic segments and resulted in extremely high heterogeneity (90%). Hence, based on sensitivity analysis, we excluded the study from our meta-analysis.

SM (vs) control/sham in influencing BP

‘Low’ quality evidence from four studies [9,25,43,45] (n = 171 participants) (Table 3) demonstrated that SM is not better than control/sham in influencing the systolic BP (SMD −0.23, 95% CI −0.53 to 0.07) or diastolic BP (SMD −0.19, 95% CI −0.67 to 0.28) in neither normotensive nor hypertensive population.

Table 3.

Summary of findings (GRADE): effects of SM on BP.

Certainty assessment							No. of patients		Effect		Certainty
№ of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	SM	Control/Sham	Relative (95% CI)	Absolute (95% CI)
Systolic BP
4	Randomised trials	Serious	Seriousa,b	Not serious	Seriousc	None	85	86	-	SMD 0.23 SD lower (0.53 lower to 0.07 higer)	⊕⊕◯◯ Low
Diastolic BP
4	Randomised trials	Not serious	Seriousa,b	Not serious	Seriousc	None	85	86	-	SMD 0.19 SD lower (0.67 lower to 0.28 higher)	⊕⊕◯◯ Low

CI: Confidence interval; SMD: Standardised mean difference.

Explanations: a. Lack of consistent, statistically significant result; b. Heterogeneity; c. Small studies, total sample size < 200.

SM (vs) control/sham in influencing oxy-hemoglobin

‘Very low’ quality evidence based on two studies [5,40] showed that SM compared to control did not affect oxy-hemoglobin (SMD 0.35, 95% CI −0.13 to 0.82) levels (Table 4).

Table 4.

Summary of findings (GRADE): effects of SM on other markers.

SM compared to control/sham for oxy-Hb; blood pressure; pupil light cycle time and diameter; and epinephrine/nor-epinephrine
Quality assessment							No. of patients		Effect		Certainty
No. of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	Spinal manipulation	Control	Relative	Absolute
Oxy-hemoglobin
2	Randomised trials	Not serious	Seriousa	Seriousb	Very seriousc	None	35	34	-	SMD 0.35 higher (0.13 lower to 0.82 higher)	⊕◯◯◯ VERY LOW
Pupil Edge Light Cycle Time (ms)
1	Randomised trials	Not serious	Seriousa	Not serious	Very seriousd	None	20	10	-	SMD 0.54 SD lower (1.18 lower to 0.09 higher)	⊕◯◯◯ VERY LOW
Pupil diameter (mm)
1	Randomised trials	Not serious	Seriousa	Not serious	Very seriousd	None	50	50	-	SMD 0.55 SD higher (0.15 higher to 0.95 higher)	⊕◯◯◯ VERY LOW
Epinephrine
1	Randomised trials	Not serious	Very seriousa	Seriousb	Very seriousd	None	19	21	-	SMD 0 SD (0.62 lower to 0.62 higher)	⊕◯◯◯ VERY LOW
Norepinephrine
1	Randomised trials	Not serious	Very seriousa	Not serious	Very seriousd	None	19	21	-	SMD 0.29 SD lower (0.92 lower to 0.33 higher)	⊕◯◯◯ VERY LOW

CI: Confidence interval; SMD: Standardised mean difference.

Explanations: a. Lack of consistent, statistically significant result; b. The outcome measure used can be influenced by other confounding factors; c. Small studies, sample size < 100; d. Single study.

SM (vs) control/sham in influencing ELPCT and pupil diameter

A ‘very low’ quality evidence study [44] (30 participants) demonstrated SM was significantly more effective than the control in reducing the ELPCT (MD −21, p < 0.01) (Table 4). ‘very low’ evidence from a different study [26] (100 participants) showed that SM is not significantly more effective than control in influencing the ANS via pupil diameter (Table 4).

SM (vs) control in influencing plasma concentration of epinephrine and norepinephrine

A ‘very low’ quality evidence based on one study [28] showed that SM compared to control did not affect epinephrine (SMD 0.00, 95% CI −0.62 to 0.62) and norepinephrine (SMD −0.29, 95% CI −0.92 to 0.33) levels (Table 4).

Subgroup analysis

Regional response

A subgroup analysis was undertaken to determine if changes in autonomic response (various outcome measures) differed based on the region of spine manipulated. Subgroups included SM to the thoracic, lumbar, or cervical spine.

Frequency domain of HRV

LF/HF ratio: There was no evidence for a differential response on LF/HF ratio and/or HF based on the region of applied SM (Figure 3b).

Figure 3b.

Regional response in LF/HF ratio following a SM.

HF: a subgroup analysis demonstrated that cervical spine manipulation (SMD −0.75; 95% CI: −1.21 to −0.29) resulted in reduction of HF compared to thoracic and/or lumbar SM (Figure 4a). This result should be interpreted with caution given the sample size and risk of bias of the subgrouped studies for cervical manipulation.

Figure 4a.

Effects of SM on HF of HRV.

Healthy vs painful population

Another subgroup analysis was undertaken to investigate whether ANS response to SM differed between people with pain and people without pain.

Frequency domain of HRV:

LF/HF ratio: a subgroup analysis demonstrated no significant difference in LF/HF ratio response between healthy and painful population following a SM (Figure 3c). There was significant heterogeneity (76%) in the studies that investigated participants with painful conditions.

Figure 3c.

Effects of LF/HF ratio response in healthy vs painful population following a SM.

HF: a subgroup analysis demonstrated that cervical spine manipulation (SMD −0.75; 95% CI: −1.21 to −0.29) resulted in reduction of HF compared to thoracic and/or lumbar SM (Figure 4b). This result should be interpreted with caution given the sample size and risk of bias of the subgrouped studies for cervical manipulation.

Figure 4b.

Regional response of SM on HF of HRV.

Search (Science Direct)

(“spinal manipulation OR ‘spinal Manipulative Therapy’ OR ‘HVLA’) AND (‘Autonomic nervous system’ OR ‘Sympathetic Nervous System’ OR ‘parasympathetic nervous system’ OR ‘Adrenaline’ OR ‘Noradrenaline’)

Number of search results: 157

Search (PubMed)

Search: (‘Manipulation, Spinal’[Majr]) AND ‘Autonomic Nervous System’[Mesh] Filters: Clinical Study, Clinical Trial, Clinical Trial, Phase I, Clinical Trial, Phase II, Comparative Study, Controlled Clinical Trial, Randomized Controlled Trial, English, Humans

Results: 18

Other outcome measures

A subgroup analysis based on region of SM and/or population was not possible for any other outcome measures (HF, rMSSD, BP, epinephrine/nor-epinephrine, ELPCT and oxy-hemoglobin) due to the lack of studies that enabled these comparisons.

Discussion

Summary of main results

This review included 14 studies [7–9,26,28,39–47] (618 participants) comparing SM against control in influencing the ANS. This was a comprehensive review which summarized and analyzed the influence of SM on a variety of proxy measures for the ANS (HRV, BP, epinephrine/nor-epinephrine) in a wide range of participants (healthy volunteers and those with painful conditions). This review was also first of its kind to investigate whether regional approach to SM had a differential effect on the ANS. The findings from this review established ‘low’ level evidence that SM is not effective compared to control in significantly influencing the included indices of ANS activity. Specifically, there was no evidence that SM may influence LF/HF ratio, HF, and rMSSD, several well-researched indices of HRV. No significant effects of SM on BP, pupil diameter, and plasma concentration of epinephrine and norepinephrine were found. However, a subgroup analysis revealed that cervical SM reduced HF component of HRV compared to thoracic and lumbar spinal manipulation.

Overall completeness and applicability of evidence

The data from this review can be considered relevant to current clinical practice and research as we did not find evidence in favor of SM influencing HRV, which is in contradiction to the current consensus amongst professions applying manual therapy techniques that believe the opposite. However, these findings must be considered with considerable caution given the methodological concerns regarding HRV data collection and overall quality of the included studies. We hypothesize the following reasons for the lack of effect observed in this review: (1) six out of eight studies [7–9,39,42,47] that investigated the effects of SM on HRV studied changes in healthy volunteers. The effects of SM may not be as pronounced in healthy controls versus a painful population (noted in meta-analysis); (2) best practices for monitoring and collecting HRV data were not followed in any of the included studies; (3) there was considerable difference in patient characteristics across studies, especially age and gender. These covariates have been shown to influence HRV [53] and could have influenced results; (4) most studies had small sample sizes, with less than 20 participants in each intervention group. Hence, the possibility of type-2 error cannot be ruled out, suggesting the need for future studies with larger sample size.

Overall, we believe that the lack of effect could be due to paucity of robust research in this area, especially regarding the performance of poor HRV data collection practices. For example, seven of the studies included in the meta-analysis pooling the effects of SM on the LF/HF ratio. Generally, studies did not comment on whether breathing rate was controlled during data collection. Controlled breathing is a critical factor when interpreting frequency domain-based indices of HRV and has been shown to drastically influence the reliability of results, so much so that researchers strongly advise against using frequency based HRV indices in isolation [54]. More research is recommended to verify the clinical implications of any changes or lack of in HRV (and ANS) following application of a SM technique. However, based on the current review, the clinical applicability of HRV/ANS changes following a SM must be reconsidered.

Quality of the evidence

As reflected by the GRADE ratings, the overall quality of the evidence in this review was ‘low’ to ‘very low’ for all outcomes. This is because there was considerable heterogeneity in included trials in terms of interventions, outcome measure and data collection techniques. Therefore, we were unable to pool data, especially for BP changes. The key issues resulting in downgrading of evidence related to imprecision (wide CIs) and inconsistency (heterogeneity and low sample size). We did not downgrade the risk of bias for blinding therapists as this is not possible to achieve when applying manual therapy techniques in studies. We also did not downgrade the evidence for ‘other bias’ for methodological limitations within included studies (refer to the limitations section); specifically pertaining to lack of baseline HRV data of participants in included studies.

Potential biases in the review process

We expect minimal biases in the review process. A minimum of two reviewers acted independently through the various phases of the review and a third reviewer was available to resolve any disagreements if required. The search strategy was developed in consultation with an experienced librarian, furthering the robust process of appropriate study retrieval. We did not downgrade the risk of bias based on ‘publication’ bias. Further, only publications done in English language were included in the review, thereby, raising the possibility of language bias [55]. However, we expect minimal impact due to this limitation and did not have access to reviewers who had suitable knowledge or skill to translate published studies from other languages than English.

Agreements and disagreements with other studies or reviews

Cardiovascular autonomic activity

HRV was the most common outcome measures of cardiovascular autonomic activity. Our findings demonstrate that SM does not significantly influence any parameters (LF/HF ratio, HF and rMSSD) of HRV. This is in contrast with previous findings [13,29,31,56]. While our review was confined to SM, other studies/reviews have included different techniques such as myofascial release and mobilization that may explain the contrast in the findings. A subgroup analysis in the current review did demonstrate that a cervical spine manipulation may influence HF component of HRV. This is in agreement with previous findings [13,31] and may provide evidence to a differential autonomic response based on the region of spine manipulated. However, this must be interpreted with caution as this is based on findings from just two studies, one of which was rated as ‘high’ risk of bias. Also, our findings demonstrate that HF may reduce (that is, reduced parasympathetic tone) after cervical manipulation. This is in contrast with long held notion that a cervical spine manipulation may increase parasympathetic activity [12,32]. These results must be interpreted with great caution due to the fact that the studies included in analysis failed to control breathing rate during data collection and did not obtain a baseline of at least two weeks of HRV data preceding the chosen intervention. Both factors have been shown to adversely influence HRV data and thereby greatly influencing the interpretation of results [54]. Overall, we did not find evidence for a differential response following a thoracic or lumbar spine manipulation, which is consistent with previous findings [30,31,57]. Hence, the debate on regional response continues to require verification through rigorously designed studies (see methodological limitations).

Two issues with regards to HRV need to be clearly acknowledged and discussed. (1) While some indices of HRV (e.g. rMSSD and HF) can be directly correlated with cardiac ANS activity, there is no singular measure of the global drive of the vagus nerve (ANS) throughout the body, which includes peripheral sites such as the thoracic spine. Therefore, observed changes in cardiac ANS activity as measured via HRV following SM do not provide a direct quantification of the changes of the ANS at the specific/direct spinal site [58] (2) The exact relationship between cardiac autonomic activity and the particular branch of the ANS (SNS or PNS) remains debated [16]. As mentioned above, the LF/HF ratio is interpreted as an indicator of ANS activity with an assumption that the LF component is an index of SNS activity and the HF component is an index of PNS activity. However, accumulating body of evidence indicates that this assumption oversimplifies the complex non-linear interactions between the SNS and the PNS divisions of the ANS [59]. For example, studies have shown that interventions known to increase SNS activity such as acute exercise reduced the LF power instead of increasing it [19,60]. Similarly, SNS activity has been shown to modulate the HF component. Further, concurrent increase (co-activation) or decrease (co-inhibition) in the SNS and the vagal outflows have also been demonstrated raising serious questions regarding the validity of LF/HF ratio as an indicator of cardiac autonomic activity [61]. Despite the ongoing debate, the LF/HF ratio is still widely used to assess autonomic cardiovascular regulation, limiting interpretation of results [62].

Pupillary reflex

From this review, we have contradictory results as to whether SM affects the ANS as measured by the pupil, which is consistent with previous findings [13]. In Gosling et al., (2005), they performed an atlantoaxial joint manipulation while measuring the pre- and post-intervention effect on ELPCT. The results showed that there was a significant decrease in ELPCT post SM, which conclude that SM does have a parasympathetic effect on the ANS. On the contrary, Sillevis et al. (2010) [26] showed no significant change in pupil diameter between groups 5 min post SM. However, Sillevis et al. (2010) [26] recruited patients with chronic cervical pain. It is known that chronic pain is often associated with autonomic dysfunctions, which could affect the generalizability of the results.

Other markers

The review found that there was no evidence that SM affects the plasma concentration of epinephrine and norepinephrine, which is in line with previous findings [13,62]. It is worth noting that the plasma concentration of epinephrine and norepinephrine can be affected by multiple factors such as the level of arousal and the amount of sleep during the night [63]. Therefore, it may not be a reliable measure of the function of ANS. Furthermore, it has been hypothesized that SM can influence cardiovascular physiology by lowering the BP. However, our findings suggest that SM is no more effective than control/sham in influencing either the systemic BP or diastolic BP. This is consistent with previous findings [25,43,46]. Finally, SM did not affect oxy-hemoglobin (a measure of skeletal muscle blood flow) indicating that SM does not result in SNS changes. However, this is based on findings from only two studies [5,40] and requires further investigation.

Limitations

Of the 14 studies reviewed, only 5 studies included symptomatic population (4 with pain) and 1 looked at the effects of SM on hypertensive subjects. Although the results of this study found that SM did not bring about significant change for this symptomatic population, previous studies have found differences in response between the healthy and symptomatic, the causes of these contradictions need further investigation [64]. Not enough is known about possible changes in the symptomatic population who are commonly seen in clinical practice [13]. Hence, more studies on symptomatic population are warranted to understand the clinical transferability of the findings [5,13,31]. This may require studies to include patient related outcome measures and how changes in ANS activities correlate with these measures. It is worth noting that the ANS is often a quick responder to stimulation, which kickstarts a cascade of downstream responses in the endocrine or central nervous system [65]. Perhaps, a more longitudinal study that assesses multiple sessions of SM could be conducted to investigate the summative impact of SM with time rather than the influence from a single session. This would greatly supplement the body of evidence regarding the clinical applicability of SM as well as to appreciate the cumulative effect of multiple sessions of SM, which could better reflect actual clinical practice [13]. Similarly, a longer duration longitudinal study would allow an increased number of HRV data collection points. All of the included studies performed two or less HRV measurements. While this is understandably a pragmatic approach to data collection (i.e. pre-, post-measurements), it greatly reduces the reliability and applicability of HRV interpretation. One study illustrates the significant impact of singular HRV data collection points, where interpretation of a few monthly HRV data points would have led to inaccurate training prescription without an understanding of the participant’s normal variability of HRV preceding the race/event [66]. At minimum, studies investigating the influence of an experimental condition on HRV should establish a two-to-three-week minimum baseline of HRV preceding the interventional period and, HRV should be collected daily, ideally five days a week for greatest reliability [54]

Recommendations

Resources such as the Guidelines for Reporting Articles on Psychiatry and Heart rate variability (GRAPH) and previously published articles [67] are readily available to guide researchers on best practices when collecting and interpreting HRV. Researchers should know that two of the biggest changes they can make to immediately improve reliability and validity of HRV data collection and interpretation are: 1) avoid investigation of frequency domain indices in isolation (stick to time domain) and, 2) collect at least five measurements of HRV a week, preceding the chosen intervention. Ideally a baseline measurement of HRV is taken at least two-to-three weeks ahead of the intervention to understand a participant’s normal variability in HRV [66]. These simple but impactful changes in data collection methods can improve one’s ability to discern patterns in change of HRV due to physiological responses versus changes due to study error. Finally, additional studies are needed that monitor changes after SM with extended timelines beyond immediate effects, and that select appropriate sham interventions.

Conclusion

In conclusion, the current review found that based on the best available evidence, SM does not significantly alter any parameters of ANS when compared with sham manipulation or controls. The quality of the included evidence is ‘low’ which may reflect the low number of included studies and their small sample sizes. Further, every included study which investigated changes in HRV to quantify the influences of SM on ANS did not follow best practices for HRV data collection and analyses. This lack of appropriate methodology to HRV data collection practices has profound effects on the reported results and therefore interpretation of our findings. Future studies should ensure that robust, evidence-based methodology is practiced in clinical trials to enable meaningful conclusions.

Biographies

Kesava Kovanur Sampath is a Principal Academic Staff Member in the Centre for Health and Social Practice at Waikato Institute of Technology in New Zealand. He is the chair of the osteopathic research steering committee in New Zealand. He is part of the leadership team at the Duke Centre of Excellence in Manual and Manipulative Therapy. He is also a visiting research fellow at the University of Technology, Sydney. Kesava’s research specialties include neurophysiological outcomes, spinal manipulation, manual therapy practice and education.

Steve Tumilty is one of the few clinician scientists in Physiotherapy Worldwide and one of the few Registered Physiotherapy Specialists in New Zealand. The majority of his clinical experience has been in the outpatient musculoskeletal practice setting in UK, Germany and New Zealand. He also has experience in professional sports and Occupational Health Physiotherapy. In 2002 he came to work at the School of Physiotherapy, Otago University and he has developed and coordinated the specialist Masters degree for Sports and Orthopaedic Manipulative Therapy for which he provides teaching and clinical expertise. He is the author or co-author on 125 research outputs in the field of Physiotherapy and Orthopaedic Manipulative Therapy for musculoskeletal conditions and has been invited to present at international conferences. He has taught manual therapy seminars in New Zealand and Japan. Associate Professor Tumilty’s research interests are in Tendinopathy, modulation of the Hypothalamus-Pituitary Axis using manual interventions, and the influence of the autonomic nervous system on musculoskeletal pain and healing.

Liana Wooten is an Assistant Professor and Assistant Director Admissions at Tufts University, Phoenix. Her research/Areas of Interest includes Applied Physiology, Clinical Exercise Physiology, Scholarship of Teaching and Learning.

Suzie Belcher is an Academic Staff Member and a physiotherapist with a passion for sports rehabilitation, injury prevention, biomechanical analysis and performance conditioning. Suzie has worked with international athletes from the UK and NZ, in Winter and Summer sports up to Olympic/Paralympic level. She recently completed a PhD ‘Improving the Design and Implementation of New Zealand’s NetballSmart Injury Prevention Programme’.

Gerard Farrell is a PhD candidate at the Center for Health, Activity and Rehabilitation Research, School of Physiotherapy at the University of Otago. His research interests include the neuroendocrine mechanisms of manual therapy, the part a dysfunctional stress response has to play in the development and maintenance of persistent post-concussion symptoms, and the role manual therapy plays in the treatment of persistent post-concussion symptoms.

Angela Spontelli Gisselman is an assistant professor at Tufts University School of Medicine. Her main research interests are in the role of health metrics, such as heart rate variability (an index of the autonomic nervous system), and their ability to influence decision making in rehabilitation, such as post-concussion rehabilitation. In addition to this line of research, other areas of interest include heart rate variability and temporomandibular disorders; the use of thermal imaging in tendinopathy; examination and management of shoulder injuries; load monitoring technology and post-operative rehabilitation.

Funding Statement

The author(s) reported there is no funding associated with the work featured in this article.

Abbreviations

ANS

autonomic nervous system

BP

blood pressure

CCT

controlled clinical trial

ELPCT

edge light pupil cycle time

GRAPH

Guidelines for Reporting Articles on Psychiatry and Heart rate variability

HF

high frequency

HRV

heart rate variability

HVLA

high velocity, low amplitude

LF

low frequency

MSNA

muscle sympathetic nerve activity

NIRS

near infrared spectroscopy

O2Hb

oxygenated hemoglobin

PNS

parasympathetic nervous system

RCT

randomized controlled trial

rMSSD

root mean square of consecutive RR intervals

SDNN

standard deviation of NN intervals

SM

spinal manipulation

SNS

sympathetic nervous system

TOI

tissue oxygenation index

Disclosure statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/10669817.2023.2285196