Effectiveness of non-invasive vagus nerve stimulation vs heart rate variability biofeedback interventions for chronic pain conditions: A systematic review

2.1 Search strategy

Electronic database searches were conducted through PubMed, Cochrane, MEDLINE, CINAHL, SPORTDiscus, and Google Scholar. The bibliographies of the collected studies were reviewed for further references applicable to this systematic review. The search was performed to identify all relevant published randomized controlled trials written in English, over the past 13 years (January 1, 2010, to December 31, 2023). An initial search was conducted on May 30, 2020, with a second search completed on October 29, 2024, during which the four review authors conducted a search with established search terms using a combination of keywords ranging from “chronic pain” or “fibromyalgia” or “headache disorder” or “migraine disorder” or “heart rate variability” or “biofeedback” or “vagus nerve stimulation” or “transcutaneous” or “pain assessment.” A full list of all search term strategies for each database are provided in the supplemental materials (Table A1).

2.2 Study selection

The protocol was developed a priori. All titles and abstracts were independently reviewed for inclusion or exclusion [14]. Inclusion criteria were expansively planned to capture all relevant literature, including unpublished (i.e., grey literature and dissertations), in order to minimize reviewer bias [15]. To further limit bias, quality assessment metrics were used, studies with plainly defined data management strategies with specific synthesis reporting metrics were included, and systematically appraising for risk of bias are mitigation strategies that were utilized.

Inclusion criteria included (1) ages of 18–65 years with chronic pain conditions more than 3 months, (2) studies with a randomized controlled design, (3) a treatment consisting of nVNS intervention, (4) a comparison utilizing HRVB, (5) a primary outcome using a validated pain intensity (PI) metric, and (6) a secondary outcome using a validated functional disability measure and HRV metric. Exclusion criteria included studies that were non-English, studies prior to 2010, studies related to non-human, where the participants had a progressive disease, were post-operative or had a surgically implanted vagus nerve stimulator, or had a traumatic brain injury, concussion, brain tumor, spinal cord injury, seizure disorder, and visceral pain. Titles and abstracts were screened for the inclusion and exclusion criteria and full text were retrieved and assessed if the initial screen was not clear. Duplicate articles were then removed. Discrepancies over trial inclusion were discussed between the authors. Trials were excluded if they failed to meet the inclusion criteria previously described. None of the review authors were blinded to each reviewed manuscripts authors, institutions, or results during assessment for inclusion or exclusion but were blinded to each other’s individual data extraction.

2.3 Data collection process

The results of the literature search and study selection were independently evaluated by three evaluators (K.P., J.H., and P.T.) examining article title, abstracts, and keywords during the initial screening according to the inclusion and exclusion criteria. Duplicate citations from the multiple databases were then removed. Afterwards, selected articles were obtained for full-text review. In cases of disagreement, a fourth reviewer (B.H.) was consulted. The complete data collection procedure is shown as a flowchart (Figure 1).

Figure 1

PRISMA flow diagram. *Consider, if feasible to do so, reporting the number of records identified from each database or register searched (rather than the total number across all databases/registers). **If automation tools were used, indicate how many records were excluded by a human and how many were excluded by automation tools. Source: Page MJ, et al. BMJ 2021;372:n71. doi: 10.1136/bmj.n71. This work is licensed under CC BY 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/.

2.4 Selected manuscript details

2.5 Outcome measures

2.6 Data extraction and management

All data extraction procedures were conducted according to the Cochrane Handbook for Systematic Reviews of Interventions version 6.0 [14]. The researchers independently extracted data from included articles. After independent review, the researchers then met to discuss any differences in the individual data extraction as a team, verifying each other’s review of the extracted data. Any disagreements were resolved through discussion amongst the authors and a fourth author, if needed. Qualitative data were extracted, compiled, and synthesized using a data extraction form that was designed by the authors following the “Checklist of items to consider in data collection” in the Cochrane Handbook for Systematic Reviews of Interventions [14] (Table 1). Meta-analysis was not performed due to the limited amount of quality evidence for HRVB for chronic pain. Differences in outcome reporting between studies prevented synthesis of p-values between included studies.

2.7 Quality assessment

The Revised Tool to Assess Risk of Bias (RoB 2) in Randomized Trials is the recommended tool to assess risk of bias for Cochrane reviews [24]. The RoB 2 was used to collect and assess the risk of bias data [25]. It functions by answering a series of signaling questions with yes, probably yes, probably no, no, or no information. These answers are then interpreted through an algorithm to create a risk of bias judgement which is proposed on a judgement scale of “low,” “some concerns,” or “high risk” of bias. Five domains are assessed with a series of signaling questions in each domain. These domains include: “(1) risk of bias arising from the randomization process, (2) risk of bias due to deviations from the intended interventions, (3) risk of bias due to missing outcome data, (4) risk of bias in measurement of the outcome, and (5) risk of bias in selection of the reported result” [25]. Three authors independently reviewed the risk of bias of included trials as described in the Cochrane Handbook for Systematic Reviews of Intervention [26]. Any disagreements over the risk of bias assessment were discussed and resolved amongst the three review authors (KP, JH, and PT), and a fourth reviewer (BH), as necessary (Table 2).

Table 2

Assess risk of bias in randomized trials (RoB 2)

Green: Low Risk of bias (–); Yellow: Some Concern for risk of bias (?); Red: High Risk of bias (+).

3.1 Study selection

The search strategy returned 1,474 articles with 236 references from PubMed, 224 references from Cochrane, 3 references from websites (e.g., Google Scholar), and 1,011 references from MEDLINE, CINAHL, and SportDiscus. A total of 531 duplicates were removed from this initial search. The titles and abstracts were screened (n = 961) and 932 were excluded based on not meeting inclusion or exclusion criteria. A total of 29 articles were sought for retrieval and all received full-text review for eligibility. Of these, 19 were excluded, leaving 10 RCTs included for review with level 2 evidence according to Oxford Centre for Evidence Based Medicine (Figure 1). PRISMA Checklist is available in the Appendix (Table A2).

3.2 Characteristics of included studies

The systematic review consisted of ten RCTs. The total number of patients within the 10 articles yielded 813 subjects. There were n = 763 subjects in studies of nVNS and n = 50 subjects for one HRVB RCT. Six of the nine nVNS studies looked at headache disorders and migraines (n = 603) [21,27–31] with two investigating effects on fibromyalgia symptoms (n = 138) [32,36] and one for chronic low back pain (n = 22) [37]. Of the studies evaluating nVNS, three demonstrated significant results in episode frequency [21,27,29], six in PI reduction [27–30,32,37], and two in reduced medication use [27,29]. One study showed significant findings for reduced PI, depression scores, and increased HRV coherence [33]. Adverse effects of nVNS included skin irritation at the application site, facial twitching, tingling or pulling, and chest discomfort [30,31,37]. There were no negative effects reported in the use of HRVB [33]. Findings for the primary and secondary outcomes are listed in Figure 2.

Figure 2

Summary of results by outcomes. Note. (+), p value <0.05; (−), p value >0.05, NT = not tested; SA, serious attacks.

A detailed presentation of key elements for each included manuscript for each outcome measures and significant values are presented in Table 1. Furthermore, comprehensive information from each manuscript was extracted from each article and was recorded in detail. This detailed information for each included article is available in the supplemental materials (Table A3).

3.3 Summary of quality of results and risk of bias

The overall quality of the evidence extracted for use in this review was “moderate,” indicating that there is a moderate chance that the true effect has been reported [34]. A total of seven studies had “low risk” and three studies had “some concern.” A score of “moderate-quality” evidence for three nVNS studies were assigned “some concern” due to limited reporting on randomization [27,31,32] while HRVB intervention study had a “low” risk of bias [33]. There were zero studies that were assigned high risk. Results were recorded and individually summarized for RoB 2, which are presented in Table 2.

3.4 Overall completeness and applicability of evidence

Despite a number of pilot studies evaluating the effects of HRVB for the treatment of chronic pain conditions, only one RCT was identified and included in this review. This study included 50 veterans that completed the follow-up assessment, ages 43–65 [33]. The quality of this study was considered high due to low concerns in methodological quality and complete reporting of the data [34]. PI was extracted from this RCT as it is a subset of the brief pain inventory. Secondary outcomes in this study included pain medication use and the beck depression index. More research in HRVB for chronic pain conditions is warranted to identify clinical applicability, although the quality of evidence in this trial is high, no other RCTs were identified evaluating HRVB.

For nVNS, a majority of the RCTs included in this review investigated nVNS in the management of headache disorders [21,27–31]. Two RCTs explored the benefits of nVNS compared to exercise for fibromyalgia [32,36] and one for chronic low back pain [37].

The included trials on nVNS and migraines were three studies that comprised a total n = 327 subjects between the ages of 18 and 75 [21,29,31]. Two of these studies [29,31] had a moderate quality of evidence due to “some concerns” regarding methodological quality, and complete reporting of data [34]. Primary outcome data extracted for this review included PR, PI, and pain episode frequency. While improvements in PI were reported to be significant by Grazzi et al. [29] (p = 0.002), improvements in PI were not significant in the other two studies (p > 0.05). While frequency of pain episodes was reported to be not significant by Straube et al. [31] (p = 0.094), improvements in pain frequency (PF) were reported by Grazzi et al. [29] (p = 0.03). Chaudhry et al. [21] reported a significant improvement in the frequency of severe episodes only (p = 0.049). Of the three studies, the study of Grazzi et al. [29] was the only study to report on the secondary outcome of rescue medication use, finding a statistically significant reduction in medication use (p = 0.008).

Three RCTs evaluating the effect of nVNS on cluster headaches (CH) studied n = 276 subjects between the ages of 18 and 75 [27,28,30]. There were “some concerns” regarding methodological quality and complete reporting of data for one study [27], resulting in moderate-quality [34]. Primary outcome assessment information extracted for this review included PR, PI, and pain episode frequency. Extracted secondary outcome data included changes in medication consumption.

The study by Gaul et al. was the only research on CH to report on frequency of pain episodes, noting 3.9 fewer attacks per week (95% CI: 0.5, 7.2; p = 0.02). In the studies by Gaul et al. [27], Goadsby et al. [28], and Silberstein et al. [30], significant reductions were reported in PI (8.93 [95% CI: 0.47, 17.39 points], p = 0.039; 42%, p < 0.01; and p = 0.008, respectively), and PR (32%, p = 0.001; 15% greater than controls p = 0.05; and 2.1 [95% CI: 1.8, 2.3], respectively). Conflicting results were reported for the secondary outcome measure of medication consumption by Gaul et al. [27] (−15 [95% CI: −22.8, −7.2] p < 0.001) and Silberstein et al. [30] (p = 0.15).

The included trials on nVNS and fibromyalgia examined n = 138 subjects between the ages of 18 and 65 [32,36]. The methodological quality of one trial had “some concerns” and reporting was complete, resulting in moderate-quality data [32]. Primary outcome assessments included PI, PR, the Beck Depression Inventory, symptom severity, and pain index [32,36].

3.5 Migraines

For the treatment of migraines, Straube et al. [31] used nVNS at the auricular branch of the vagus nerve with a frequency of 25 Hz for the treatment group and 1 Hz for the active controls. The study included both males and females, 18–70 years of age. The RCT found that the control group at 1 Hz was significant for reducing >50% of headache frequency compared to the 25 Hz group. No statistically significant changes in PI were found in the control or intervention group. Both groups (1 Hz and 25 Hz) significantly reduced depression scales and use of medication, with no statistical significance between groups.

Chaudhry et al. [21] assessed the use of nVNS at the cervical branch of the vagus nerve for migraine intervention, compared to age matched controls. The study included both males and females, 27–66 years of age. The nVNS group only had significant results for reduced PF of severe migraine attacks. There were no significant findings for changes in PI and PR. Disability and medication usage were not assessed. A limitation of this study was that all but one participant were female, and literature shows males and females have different pain and HRV responses [17].

Grazzi et al. [29] performed a post-hoc analysis of nVNS at an unspecified frequency to the cervical branch of the vagus nerve in migraines compared to a sham group. The study included both males and females between 18 and 65 years of age. Significant findings were reduced PI, reduced PF, and reduced use of rescue medications in the nVNS group. It appears that the duration of nVNS (30, 60, 120 s), respectively, correlated with the increased significance of results.

Reported adverse events related to the intervention were ulcers at the auricular cite [31]. No device or intervention related adverse events were reported by Chaudhry et al. [21] or Grazzi et al. [29].

3.6 CH

Silberstein et al. [30] and Goadsby et al. [28] had similar studies with comparable outcomes. Silberstein included both male and females, 18–65 years of age. Goadsby included both males and females, ages 33–55 years of age. Both studies looked at nVNS for treatment of episodic cluster headaches (eCH) vs chronic cluster headaches (cCH). Both RCTs found significant effects for eCH, but not cCH, in reduced PI, with a frequency of 25 Hz to the cervical branch of the vagus nerve.

Gaul et al. [27] also looked at nVNS in CH in both male and female subjects, 18–70 years of age. Gaul found reduced PI, PF, improved PR, reduced depression, and reduced medication use with a treatment frequency at 5 Hz to the cervical branch of the vagus nerve. No adverse events related to the device or intervention were reported.

The differences in outcomes in the study by Gaul et al. [27] compared to that by Silberstein et al. [30] and Goadsby et al. [28] may be related to the selection of treatment frequency. The frequency of nVNS appears to have different effects at the cervical and auricular branches of the vagus nerve in headache disorders. One possible explanation may be due to the myelination of the cervical branch and no myelination of the auricular branch of the vagus nerve. [31]. Further investigation should be done to understand these differences.

Reported adverse events related to nVNS were facial twitching, tingling or pulling, and skin irritation, redness, or soreness at the application site [30]. Goadsby et al. [28] reported no severe adverse events related to the device or intervention, however it is unclear as to what non-severe adverse events occurred.

3.7 Chronic pain

Winstead [33] investigated the effects of HRVB compared to a sham group in veterans with chronic pain. The study included males (66%) and females (33%) within 43–65 years of age. HRVB performed six times per week for 25 min over the course of 8 weeks followed by a 12-week booster session and 16-week follow up. Significant positive outcomes were reported for reduced PI and disability for the HRVB group compared to the sham group. No adverse events were reported.

3.8 Fibromyalgia

Kutlu et al. [32] investigated the use of nVNS in females with fibromyalgia. nVNS with a frequency of 10 Hz, to the auricular branch of the vagus nerve, combined with exercise, was compared to exercise alone (control group). There were significant benefits in both the intervention group and the control group for reducing PI and reducing disability scores; however, there were no significant differences between the groups. Limitations of this study were generalizability as only females were included, and literature shows males and females have different pain and HRV responses [17]. The exercises in both groups were generally stated; therefore, it is unclear as to which specific exercises may be beneficial. No adverse events were reported.

Pacccione et al. [36] studied both males and females with fibromyalgia separated into four treatment groups: receiving active nVNS (25 Hz), sham nVNS, meditation centered diaphragmatic breathing, and sham breathing technique. No significant differences were identified between groups. There are two primary limitations of this study. First, the participants were primarily female (94.8%) and based on the individual and sex differences in HRV responses, the results may have been affected. Second, the sample size for each group may not have been adequate to identify significant differences between groups. One adverse event occurred with active nVNS resulting in chest discomfort and pain with the intervention, resulting in the participant withdrawal from the study.

3.9 Chronic low back pain

Uzlifatin et al. [37] compared two groups: one group received nVNS (25 Hz) and exercise while the control group received only exercise. Both groups reported significant improvements in perceived disability; however, no differences were identified between groups. Limitations included a small sample size and retesting occurring only 1 day after the treatment which may not have allowed adequate time to measure for changes which might have occurred over time if long-term follow up assessment was completed. No adverse events were reported.

4.1 Limitations of this review

Limitations for this systematic review include study type, where only full-text RCTs in English were included, including one doctoral dissertation [33]. Third-party participants, such as an academic librarian, for settling discrepancies in article selection were not involved in this review. The sample size was low among the included studies. While the largest sample size was N = 285 participants, [29] the mean sample size among the remaining 9 studies was N = 62.8, and the overall mean sample size was N = 173.9. A lack of heterogeneity among pain conditions is also noted across the studies, with all but four studies evaluating subjects with migraine and CH conditions. The remaining two studies [32,33,36,37] evaluated subjects with chronic pain and fibromyalgia, respectively. Other potential limitations include the varied use of placebo/sham and blinding, limited study duration, and follow-up duration. Due to variability in observed conditions, outcome measures, and limited reported statistical data, meta-analysis was not performed.

5.1 Implications for practice

In summary, impaired vagal tone is related to pain in individuals who experience chronic pain conditions. Individuals with chronic pain conditions including fibromyalgia, migraine, and CH, who received nVNS or HRVB, reported an improvement in PI, PF, medication use frequency, and depression. More “high-quality” research is needed that looks at nVNS and HRVB in chronic pain conditions. Negative outcomes included adverse events reported at the site of application of nVNS resulting in skin irritation, redness, soreness, or muscle twitching, pulling or dropping, and ulceration. No severe adverse effects were related to the use of nVNS.

5.2 Implications for research

Currently, more research is needed to investigate the specific mechanisms behind impaired vagal activity in chronic pain patients [13]. Additionally, research that explores the underlying mechanisms of HRV biofeedback on pain and determining the optimal treatment parameters for nVNS are required.