No Consistent Evidence For The Anti-Inflammatory Effect of Vagus Nerve Stimulation in Humans - A Systematic Review and Meta-Analysis
No Consistent Evidence For The Anti-Inflammatory Effect of Vagus Nerve Stimulation in Humans - A Systematic Review and Meta-Analysis
No Consistent Evidence For The Anti-Inflammatory Effect of Vagus Nerve Stimulation in Humans - A Systematic Review and Meta-Analysis
A R T I C L E I N F O A B S T R A C T
Keywords: Vagus nerve stimulation (VNS) has been identified as an innovative immunosuppressive treatment strategy in
Vagus Nerve Stimulation rodent studies. However, its’ clinical potential is still unclear. Therefore, we aimed to assess whether VNS can
Cholinergic Anti-inflammatory Pathway reduce inflammatory proteins and/or immune cells in humans, through a pre-registered systematic review and
Inflammation
meta-analysis according to PRISMA guidelines. The databases Cochrane, Pubmed and World of Knowledge were
Vagal
searched in duplicate up to the 3rd of March 2022 and publications from identified clinical trial registrations
Immune System
Cytokine were identified until 20th of August 2023. Studies were included if they provided peer-reviewed data for humans
Tumour necrosis factor who received VNS as short-term (<=1 day) or long-term (>=2 days-365 days) stimulation and reported at least
Interleukins one cytokine or immune cell after treatment. Screening of title, abstract, full text, and data extraction was
Inflammatory Reflex performed in duplicate by two independent reviewers. Data were pooled using a random-effects model and meta-
regression was performed for moderating factors. Reporting bias was assessed. The standardized mean difference
(Hedge’s g) was used to indicate overall differences of cytokine data (mean and standard deviation or median
and interquartile range at the study level) to test our a-priori hypothesis. The systematic review of 36 studies
with 1135 participants (355 receiving a control/sham condition and 780 receiving VNS) revealed anti-
inflammatory effects of VNS for cytokines in several reports, albeit often in subgroup analyses, but our meta-
analyses of 26 studies did not confirm these findings. Although most cytokines were numerically reduced, the
reduction did not reach statistical significance after VNS: not in the between-group comparisons (short-term:
TNF-α: g = − 0.21, p = 0.359; IL-6: g = − 0.94, p = 0.112; long-term: TNF-α: g = − 0.13, p = 0.196; IL-6: g =
− 0.67, p = 0.306); nor in the within-study designs (short-term: TNF-α: g = − 0.45, p = 0.630; IL-6: g = 0.28, p =
0.840; TNF-α: g = − 0.53, p = 0.297; IL-6:g = − 0.02, p = 0.954). Only the subgroup analysis of 4 long-term
studies with acute inflammation was significant: VNS decreased CRP significantly more than sham stimula
tion. Additional subgroup analyses including stimulation duration, stimulation method (invasive/non-invasive),
immune stimulation, and study quality did not alter results. However, heterogeneity was high, and most studies
had poor to fair quality. Given the low number of studies for each disease, a disease-specific analysis was not
possible. In conclusion, while numeric effects were reported in individual studies, the current evidence does not
substantiate the claim that VNS impacts inflammatory cytokines in humans. However, it may be beneficial
during acute inflammatory events. To assess its full potential, high-quality studies and technological advances
are required.
* Corresponding author.
E-mail address: [email protected] (S. Edwin Thanarajah).
1
These authors contributed equally to this work. Shared first authorship.
https://doi.org/10.1016/j.bbi.2023.12.008
Received 11 September 2023; Received in revised form 17 November 2023; Accepted 4 December 2023
Available online 7 December 2023
0889-1591/© 2023 Elsevier Inc. All rights reserved.
C. Schiweck et al. Brain Behavior and Immunity 116 (2024) 237–258
inflammatory response to endotoxemia was amplified by vagotomy and for articles with no year restriction, or any filters set, until the 3rd of
attenuated by stimulation of distal vagal nerve endings, a mechanism March 2022. No restriction was set for journals. Clinical trial registra
since known as the cholinergic anti-inflammatory pathway (CAP) tions and references of eligible articles were also searched, and experts
(Tracey, 2002). were asked to identify literature. For clinical trial registrations, related
The CAP acts as a negative feedback loop. Detection of inflammatory publications were searched and included if all data was obtained before
proteins by afferent VN endings activates the nucleus of the solitary tract 20th of August 2023. Two articles published after the search was
(Andersson and Tracey, 2012). Consequently, acetylcholine release completed were included following recommendation during the peer
mediated by VN efferents suppresses cytokine release by immune cells review process.
(in particular, splenic macrophages). This anti-inflammatory effect re
quires sympathetic release of norepinephrine, which binds to ß2- 2.3. Search strategy
adrenergic receptors on choline acetyltransferase positive T cells (Rosas-
Ballina, 2011; Vida, 2011). The acetylcholine synthesized by these T Details on the search strategy, selection process, data collection
cells binds to alpha7 nicotinic receptor subunits on splenic macrophages process, extracted data items and risk of study bias assessment can be
and in turn, suppresses the release of cytokines such as tumor necrosis found in the supplementary information. In brief, all articles were
factor alpha (TNF-⍺) (see Alen, 2022; Kelly et al., 2022). screened for title and abstract. Subsequently the full text was assessed,
Translating these findings to humans has been hampered by the and data was extracted by at least two independent reviewers (all in
difficult accessibility of the nerve. Invasive vagus nerve stimulation duplicate) and reviewed by a senior author.
(iVNS), targeting the cervical VN via a surgically implanted electrode, is
FDA-approved for treatment-resistant depression and refractory epi 2.4. Selection process
lepsy, but its mechanism of action remains unclear. With emerging ev
idence linking inflammatory processes with these conditions, the anti- Articles were extracted in a merged literature library (Mendeley
inflammatory potential of iVNS has been investigated in implanted pa Reference Manager 2.77.0 © 2022 Mendeley Ltd) and reviewed in
tients, and clinical studies were performed in autoimmune diseases (e.g., duplicate and independently by two authors (SS and LJ/TZ) each for the
Crohn’s disease, rheumatoid arthritis (Bonaz, 2016; Sinniger, 2020; title screen, the abstract screen, and the full-text screen. Decisions on
Koopman, et al., 2016). Technical advances in non-invasive, trans full-text screenings were recorded independently and separately by each
cutaneous VNS (tVNS) targeting the vagal trunk in the neck or its sen author (merged table can be found in supplementary information 1).
sory branch in the cymba conchae of the ear have recently enabled The inclusion of articles was discussed by SS and TZ and all finally
investigations in a diverse range of inflammation-associated conditions included articles were discussed with CS and SET. In case of disagree
(Tornero, 2022; Badran, 2022). ment between SS and TZ in any of the steps, CS or SET were consulted for
Targeting the CAP via neuromodulation is an intriguing therapeutic a decision. If essential information to decide on inclusion or exclusion
option that is better tolerated, more specific, and has fewer side effects was missing, authors of articles were contacted by email, with one
than pharmacological manipulation with anticholinergics. However, to reminder.
date, the anti-inflammatory efficacy of VNS is still unclear. Although
several excellent narrative reviews and commentaries (Van Maanen 2.5. Data collection process
et al., 2009; Kelly et al., 2022), as well as a systematic review (Kwan,
2016) with a focus on rodent data concluded that VNS is a promising Data of full texts was extracted into tables, in duplicate and inde
anti-inflammatory approach, to the best of our knowledge, a systematic pendently by SS and TZ/LJ and additionally verified independently by
investigation of currently available human data is still lacking. To close either CS or SET. In case of disagreement, CS or SET were consulted. In
this gap, we performed a systematic review and meta-analysis and case of missing data, authors were contacted with the email address
investigated whether VNS can reduce inflammatory cytokines after both published for the corresponding author. In case of no reply and/or an
short and long stimulation, encompassing all available human data, invalid email address, authors were additionally contacted on research-
irrespective of study design and population. gate. If no response occurred, an additional reminder was sent.
This review was registered on PROSPERO [CRD42022308664]. Raw data was extracted when possible. For quantitative analyses (e.
g., meta-analyses) only data providing an estimate of the mean or me
2.1. Eligibility criteria dian and a measure of variance were included. Medians were converted
to means and standard deviations using the method proposed by Wan
Any study in languages spoken by the review team, which reported and colleagues (Wan et al., 2014). If subgroup data was reported, the
on circulating cytokines and/or immune cells in any disease/condition data was combined into one group. Data was not imputed. We opted for
after an intervention with VNS in humans was eligible. No participant two separate designs: one short-term analysis including studies with a
restrictions were set. Study designs could include clinical trials, case- short stimulation duration (i.e., <=1 day) and long-term designs with
control, cross-sectional and longitudinal studies in healthy subjects studies longer than one day. The following data was extracted: general
and subjects with any pathological condition across all age groups, study design (e.g., clinical trial, case-control design, or longitudinal
where ethical approval was obtained. No year restriction was set, and study); the study population, e.g., the diagnosis (if applicable), type of
studies had to be peer-reviewed articles with full texts published in condition (autoimmune/mental health/other). Disease duration (mean
English, German, French or Dutch. If studies did not report on cytokines/ in years) was rarely given and therefore not extracted. We also extracted
immune cells at the post-intervention time point or for within-group the type of intervention (invasive VNS (iVNS)/transcutaneous auricular
designs a pre- and post-value, or if they did not mention vagus nerve taVNS)/ transcutaneous cervical VNS (tcVNS)/other, specified), the
stimulation, they were not considered in the meta-analysis. Animal treatment duration: in days for long-term interventions, minutes for
studies were also excluded. short term interventions; the stimulation parameters: stimulation site
and side, current intensity (mA), pulse width (μs), frequency (Hz); the
2.2. Information sources type of placebo control (if applicable). Additionally, sample character
istics that were extracted included the number of participants, the mean
The databases Cochrane, Web of Science and Pubmed were searched age, sex, and the mean BMI. For the variables of interest (e.g., cytokines
238
C. Schiweck et al. Brain Behavior and Immunity 116 (2024) 237–258
immune cells) we included pre and post intervention concentrations 10 studies, given that these factors may yield different results due to
(mean or median and standard deviation/range). Moreover, we recor methodological differences. Additionally, we conducted a subgroup
ded if an immune cell stimulation with Lipopolysaccharide (LPS) was analysis only for studies with an acute inflammatory event.
performed.
2.10. Reporting bias and sensitivity analyses
2.7. Study risk of bias assessment
Publication bias was assessed with Egger’s test and Funnel plots,
The risk of bias was assessed by the quality assessment tools for influential case analysis using the leave-one-out method, and calculating
controlled intervention studies and pre-post studies with no control influence diagnostics following Viechtbauer & Cheung (Viechtbauer and
group provided by the NHLBI (https://www.nhlbi.nih.gov/health-topi Cheung, 2010) dmetar (v0.1.0) (Harrer et al., 2019).
cs/study-quality-assessment-tools). Study quality was scored for each
study independently by two authors (SS and TZ). Studies were graded as 3. Results
poor, fair or good based on each 12, respectively 14 quality criteria by SS
and TZ and confirmed by either CS or SET. Criteria considered for After duplicate removal 1076 studies were screened for eligibility
controlled trials included randomization, treatment allocation, blinding, (title/abstract). In the full text screen, 42 studies were excluded with
group similarities at baseline, low drop-out rates, high adherence, sim reason (Supplementary Table 1), yielding 36 eligible studies (Koop
ilarity or avoidance of background treatments, good outcome assess man, et al., 2016; D’Haens, 2023; Bonaz, 2016; Andreas, 2019; Tarn
ment, sufficient sample size, subgroup analysis, and if all randomized et al., 2019; Aranow, 2021; Corrêa, 2022; Seitz, 2022; Brock, 2021;
participants were analyzed in the group to which they were originally Chaudhry, 2019; Mion, 2020; De Herdt, 2009; Salama et al., 2020;
assigned. The quality criteria for studies without a control group covered Aalbers, 2012; Boström, 2019; Stavrakis, et al., 2017; Corcoran et al.,
a clearly formulated study question and eligibility criteria, if the study 2005; Bremner, 2020; Wu, 2023; Addorisio, et al., 2019; Brock et al.,
participants were representative of the group of interest, and if all 2017; Barone, et al., 2007; Veiz, 2022; Stavrakis, et al., 2015; Zhou,
eligible participants were enrolled. Moreover, it was verified if the 2022; Merchant et al., 2022; Drewes et al., 2021; Kox, 2015; Baro et al.,
treatment was clearly described and applied consistently across the 2022; Sinniger, 2020; Stavrakis, 2020; Majoie et al., 2010; de Moraes,
study. Additional criteria included outcome measures, blinding of 2023; Tornero, 2022; Lerman, 2016; Stakenborg, 2017) with 1135
evaluators, low loss to follow-up, and statistical methods. For the risk of participants (355 receiving a control condition and 780 receiving VNS)
bias ratings, studies were deemed as poor, fair or good based on two for the systematic review and 27 studies for the meta-analysis (See
independent reviewer’s ratings, who both followed the guidance for Fig. 1). Six studies were used in the between-group short-term analysis
quality ratings provided by the NHLBI. If inconsistent ratings were ob (Addorisio, et al., 2019; Kox, 2015; Stavrakis, et al., 2017; Veiz, 2022;
tained, a third reviewer (CS or SET) made the decision. The tools do not Zhou, 2022; Stavrakis, et al., 2015), 7 for the within-group short-term
provide a list with items which adds up to a score, but rather provide analysis (Addorisio, et al., 2019; Kox, 2015; Stavrakis, et al., 2017; Veiz,
guidance for reviewer’s to critically appraise and consider nuances of 2022; Zhou, 2022; Stavrakis, et al., 2015; Brock et al., 2017; Drewes
study designs. et al., 2021; Brock, 2021), 8 in the between-group long-term analyses
(Stavrakis, et al., 2017; Aalbers, 2012; Bremner, 2020; Chaudhry, 2019;
2.8. Statistical analysis Salama et al., 2020; Stavrakis, 2020; Corrêa, 2022) and 20 studies for
the within-group long-term analysis (Sinniger, 2020; D’Haens, 2023;
Where appropriate control groups were present, between-group an Addorisio, et al., 2019; Stavrakis, et al., 2017; Brock et al., 2017; Drewes
alyses were conducted for 1) short (<=1 day) and 2) long-term (>=2 et al., 2021; Bremner, 2020; Salama et al., 2020; Stavrakis, 2020;
day) stimulation designs; additionally, within-group analyses were Corrêa, 2022; Baro et al., 2022; Barone, et al., 2007; Boström, 2019;
conducted including studies without appropriate control group. Within- Chaudhry et al., 2015; Corcoran et al., 2005; De Herdt, 2009; Koopman,
group analyses included pre-post values of analyses 1) or 2). Meta- et al., 2016; Majoie et al., 2010; Merchant et al., 2022; Mion, 2020; Wu,
analyses were performed for cytokines reported in minimally 5 2023).
studies. Additionally, we conducted a subgroup analysis only for studies
with an acute inflammatory event. Here, the lower threshold of at least 5 3.1. Study characteristics
studies to perform a meta-analysis was disregarded due to the limited
availability of appropriate studies. Most studies were conducted in adults (>90 %), open-label (>60 %)
and around 50 % were performed without a control condition. Eleven
2.9. Effect measures and synthesis studies used iVNS (of which 3 as double-blind randomized controlled
trial (DB-RCT)), 25 used tVNS (10 as DB-RCT). Sample sizes were low
For between-group analyses, the standardized mean differences be ranging from 5 (Baro et al., 2022) to 62 (Wu, 2023) per condition. Study
tween groups (VNS vs. sham) post-treatment was computed (Hedge’s g). characteristics are summarized in Supplementary Table 2; cytokine
For within-group analyses, Hedge’s g was derived from the difference data in Table 1.
between pre (t1) and post-treatment (t2). Here, the standard deviation
at t1 was used for standardization, with an assumed pre-post correlation 3.2. Qualitative synthesis
coefficient of 0.7 (if no correlation was retrieved). Statistical analyses
were performed in R v4.3.1. (R Core Team. R: A, 2023), following the In short-term studies, there were various populations, amongst
guide for R (Harrer et al., 2021). The esc (v0.5.1) (Lüdecke et al., 2019) which healthy controls (Addorisio, et al., 2019; Kox, 2015; Veiz, 2022),
and meta (v6.5.0) (Balduzzi et al., 2019) packages were used for participants undergoing surgery (Zhou, 2022; Salama et al., 2020),
random-effects meta-analyses, metafor (v4.2–0) (Viechtbauer, 2010) for rheumatoid arthritis (Addorisio, et al., 2019) and patients with parox
additional visualizations and comparison. The inverse variance method ysmal atrial fibrillation (Stavrakis, et al., 2015). Brief (i.e., 2 min- 60 min
with restricted maximum likelihood estimator as a between-study (Addorisio, et al., 2019; Kox, 2015; Veiz, 2022; Stavrakis, et al., 2015)
(heterogeneity) variance estimator, and Hartung-Knapp adjustment and intermediate stimulation protocols (e.g., >60 min-24 h (Zhou, 2022;
was applied. The Q-Profile method for confidence interval estimation of Salama et al., 2020) were applied. While some studies reported strong
τ2 and τ was used. Subgroup analyses/meta-regression was performed suppression of IL-6 or TNF-α at various timepoints or in subgroups (e.g.,
for treatment modality (tVNS or iVNS), LPS stimulation (yes/no), in a healthy population (Addorisio, et al., 2019); 2 h after ex-vivo LPS
stimulation duration and study quality, for cytokines reported in at least stimulation (Kox, 2015), in the femoral vein (Stavrakis, et al., 2017), 1-
239
C. Schiweck et al. Brain Behavior and Immunity 116 (2024) 237–258
Fig. 1. PRISM FLOW Diagram. The diagram depicts the results of the systematic review. Reasons for exclusion were defined as follows: 1: findings from animal
data, 2: no original research data, 3: no ethical approval, 4: studies that doń t fit the following criteria: case-control, cross-sectional and longitudinal studies, 5:
unpublished, 6: Protocol (without study data), 7: studies that doń t investigate the effect of VNS on circulating inflammatory markers in a pre-post design, 8:
Retracted Article.
day post-surgery (Salama et al., 2020)), the same or other studies re Mion, 2020; Aranow, 2021; Tarn et al., 2019). The total stimulation
ported no change or even increased levels of cytokines (e.g., (Veiz, 2022) duration varied largely (3 days-1 year). Elevated inflammatory param
for other subgroups or timepoints. To illustrate, while Stavrakis et al eters were mostly not a requirement for inclusion and the change of
(Stavrakis, et al., 2017) did find significant reductions under tVNS in inflammation was usually not the primary outcome. Open-label designs
blood taken from the femoral vein, no change was observed in blood were common (15/26). Due to the small sample sizes, some studies re
from the coronary sinus. The authors interpreted their results as evi ported modulatory effects of VNS on the single subject level, or in
dence for activation of the CAIP rather than a local suppression, given clinically defined subgroups, and found significant differences for cy
that cytokines were suppressed in the femoral vein (i.e., systemically) tokines (Drewes et al, 2021; Brock, 2021; De Herdt, 2009; Koopman,
and not in the coronary sinus (i.e., locally). et al., 2016; Majoie et al., 2010; Merchant et al., 2022); these changes
The 26 long-term intervention studies with small sample sizes (5–50 were however not evident across the whole sample. With a few excep
per group) included patient samples with a long medical history and tions, the change of single cytokines was not associated with improve
often, treatment resistance (Bonaz, 2016; Aalbers, 2012; Chaudhry, ment in clinical symptoms. In 7 studies VNS was performed before or
2019; Baro et al., 2022; Barone, et al., 2007; Boström, 2019; Corcoran immediately after the onset of an event that induced an inflammatory
et al., 2005; De Herdt, 2009; Majoie et al., 2010; Merchant et al., 2022; response (surgery (Stavrakis, et al., 2017; Salama et al., 2020; Andreas,
240
C. Schiweck et al.
Table 1
Qualitative synthesis of inflammatory parameters in identified studies. p value codes: <0.05*; <0.01**; <0.001:***, Abbreviations: d:day; ns: no significant change in concentrations (p ≥ 0.05), nr: not reported, na: not
applicable, IL: Interleukin; CRP: C-reactive Protein; TNF: tumor necrosis factor; LOD: level of detection; LPS: lipopolysaccharide, PBMCs: peripheral blood mononuclear cells, RA = Rheumatoid Arthritis; PSA: Psoriasis
Arthritis; sd = standard deviation; sub p = substance p;DAS-28: Disease Acitivity Score for Rheumatoid Arthiritis (<3.2 low activity, 3.2–5.1 moderate activity, >5.1 high activity) ASDAS:Ankylosing Spondylitis Disease
Acitivity Score, Quality: As assessed with the Quality Rating Scheme for Studies and Other Evidence (modified from the Oxford Centre for Evidence-based Medicine for ratings of individual studies): 1:Properly powered
and conducted randomized clinical trial; systematic review with meta-analysis; 2: Well-designed controlled trial without randomization; prospective comparative cohort trial; 3:Case-control studies; retrospective cohort
study; 4: Case series with or without intervention; cross-sectional study;5:Opinion of respected authorities; case reports. All cytokine concentrations are given in pg/mL, CRP was measured in mg/L.
Short
Stimulation
Duration
First author, Intervention Diagnosis LPS TNF-α, mean IL6, mean ± Direction and Comment/Special Design Comment Quality*
year ± sd or sd or median significance /Inflammatory
median (range) in for VNS status, Disease
(range) in pg/mL condition Severity at baseline
pg/mL and after
intervention
Addorisio, tVNS Healthy yes pre: 4796 ± pre: 4541 ± pre: 6137 ± pre: 5979 ± nr TNF-α ↓ *; IL-6 Whole blood, LPS induced No elevation of 1
et al., (ex 607; 624; 369; 480; ↓**; IL-1β ↓*** stimulation ex vivo. inflammatory
2019 vivo) post:4477 ± post:3625 ± post:5577 ± post:4342 Significant reduction pre- parameters reported
653 645 401 ± 597 post tVNS (up to 80 %) but at baseline.
not in sham condition
Addorisio, tVNS Rheumatoid yes nr nr nr nr nr CRP ↓ * Whole blood, LPS induced Inclusion of patients 4
et al., Arthritis (ex stimulation ex vivo. In RA, with moderate to
2019 vivo) CRP levels significantly severe disease
241
Brock et al., tVNS Healthy no na pre:1.95 ± nr nr na TNF-α ↓* (p = Reduction of TNF-α was No elevation of 4
2017 0.4; post:1.8 0.01); IL-10 = ns; significant at 24 h; no other inflammatory
± 0.6 IL-8 = ns; IL-4 = within-group differences parameters reported
ns; IFN-γ = ns for cytokines; increase in at baseline.
cardiac vagal tone and
slowing of heart rate
First author, Intervention Diagnosis LPS TNF-α, mean IL6, mean ± Direction and Comment/Special Design Comment Quality*
year ± sd or sd or median significance /Inflammatory
median (range) in for VNS status, Disease
(range) in pg/mL condition Severity at baseline
pg/mL and after
intervention
Drewes tVNS Rheumatoid no na pre: 9.95 ± na pre:5.8 ± na TNF-α = ns; IL-6 In high disease activity Inclusion based on 4
et al., arthritis 19.3106; 14.4323; = ns; IL-8 = ns; IL- group, reductions in disease severity.
2021 post:10.3 ± post:3.05 ± 12 = ns; subgroup DAS28-CRP, CRP, and IFN-
19.7858 4.6890 1: CRP↓ (p = γ; low disease activity High-disease activity
0.02); IFN-γ↓* (p group: no difference for group (DAS-28 >
= 0.02); IL-10 = DAS28-CRP, decrease in 3.2): clinical
ns; subgroup2: IL- cardiac vagal tone and improvement
10↓ * (p = 0.02); reduction in IL-10; Overall on days 2 and 5 after
CRP = ns; IFN-γ = cardiac vagaltone was tVNS
ns negatively associated with Low-disease acitivity
DAS28-CRP. Participants group (DAS-28 <
with high disease activity 2.6): no clinical
had lower baseline cardiac improvement after
vagal tone than tVNS.
participants with low
242
disease activity.
Kox et al. transvenous Healthy yes pre: 6.43 ± pre:5.37 ± pre: 3.2 ± 0; pre: 7.91 ± TNF-α = ns; nr Intravenous (in vivo) LPS Inflammatory 1
2015 VNS (in 2.365608; 1.16; post: 3.2 ± 0 14.13; IL6 = ns; IL-10 administration and LPS cytokines within
vivo post: 8.61 ± post:7.06 ± post:7.62 ± = ns; IL-8 = ns induced stimulation ex normal range at
and 3.51 1.63 13,98 vivo; No difference baseline. Sharp
ex between symptoms of in increase after LPS
vivo) vivo LPS induced stimulation, no
endotoxemia, nor cytokine change by VNS
levels between-groups; Ex- compared to sham.
vivo LPS induced cytokine
production was not
different between-groups
Leukocyte increase after
LPS administration
Salama tVNS Resectable non- no nr nr pre: 63.4 ± pre:53.2 ± IL-6↓ (p = IL-6↑* (p = 0.03); Comparison to controls From baseline to 1st 1
et al. small cell lung 29.0; post: 20.0; 0.02); CRP↓ (p CRP = ns; IL- without sham condition. POD sharp increase
2020◦ cancer 258.6 ± 110 post:154.8 = 0.01); IL- 10↑** (p < 0.008) Differences of tVNS on the in IL-6 levels, 4th
± 93 10↑ (p = 0.03) acute inflammatory POD: return to BL
response after lung surgery values.
(baseline = pre-operative
value, post-intervention = VNS group showed
4th postoperative d). On shorter
the 1st postoperative d, IL- hospitalization time
6 levels were lower with and lower incidence
tVNS compared to sham
(continued on next page)
C. Schiweck et al.
Table 1 (continued )
Short
Stimulation
Duration
First author, Intervention Diagnosis LPS TNF-α, mean IL6, mean ± Direction and Comment/Special Design Comment Quality*
year ± sd or sd or median significance /Inflammatory
median (range) in for VNS status, Disease
(range) in pg/mL condition Severity at baseline
pg/mL and after
intervention
Stavrakis tVNS Paroxysmal no pre:7.8 ± pre:9.1 ± 5.5; pre: 5.7 ± pre:5.5 ± TNF-α = ns; nr No significant between- No elevation of 1
et al. atrial 3.3; post: 6.3 post: 7.4 ± 5 2.9; post:5.9 2.7; post: IL-6 = ns; IL- group difference between inflammatory
2015 fibrillation, AF ± 2.8 ± 2.7 6.3 ± 3.3 10 = ns; CRP sham and tVNS in coronary parameters reported
(coronary catheter = ns sinus. tVNS decreased AF at baseline.
sinus) ablation duration
No elevation
reported by AF-
induction.
Stavrakis tVNS Paroxysmal no pre: 7.6 ± pre: 9.2 ± pre:3.9 ± pre:3.2 ± TNF-α↓** (p = nr A significantly greater No elevation of 1
et al. atrial 3.0; post: 6.8 4.8; post: 6.9 2.9; post: 5.7 2.0; post:4.3 0.006); IL-6 = reduction was achieved in inflammatory
243
2015 fibrillationAF, ± 3.0 ± 4.4 ± 3.0 ± 2.1 ns; IL-10 = ns; the intervention group parameters reported
(femoral catheter CRP↓*** (comparison of change at baseline.
vein) ablation scores); BL levels were non-
significantly higher in the No elevation
intervention group. reported by AF-
tVNS decreased AF induction.
duration
Stavrakis tVNS Patients no pre:46.2 ± pre: 46.7 ± pre:700.3 ± pre: 696.3 TNF-α: nr; TNF-α ↓ (p < Significant changes for all Inflammatory 1
et al. undergoing 1.5; 1.4; post:39.7 94.5; ± 181.3; IL-6:nr; IL10: 0.001)***; cytokines with time in both parameters were
2017 cardiac surgery post:39.9 ± ± 1.5 post:555.4 ± post:349.0 nr; CRP:nr IL-6:nr; IL-10: nr; groups, suggesting a increased at baseline
1.4 97.4 ± 54.4 CRP: nr significant decrease in immediately after
cytokine levels over time, surgery and
as part of the post- decreased over time.
operative natural course;
clinically significant Atrial fibrillation
First author, Intervention Diagnosis LPS TNF-α, mean IL6, mean ± Direction and Comment/Special Design Comment Quality*
year ± sd or sd or median significance /Inflammatory
median (range) in for VNS status, Disease
(range) in pg/mL condition Severity at baseline
pg/mL and after
intervention
Long
244
stimulation
duration
TNF-α, IL6, mean ± Direction and Comment/Special Quality*
mean ± sd sd or median significance Design
or median (range)
(range)
First Intervention Diagnosis LPS sham VNS sham VNS between within
author,
year
Aalbers iVNS Refractory epilepsy no nr nr pre: 1.16 pre: 1.16 IL-6 = ns; IL-10 IL-6 = ns; IL- Lower baseline IL-6 No elevation of 1
et al. (children) (0.23–9.57); (0.23–9.57); post: = ns; IL-1β = ns 10 = ns; IL- levels indicated inflammatory
2012 post 1.09 1.27 (0.30–10.13) 1β = ns more significant parameters reported
(0.25–17.39) seizure frequency at baseline.
reduction during
add-on phase [R2 =
No change in CRP,
ESR, white blood
cells, lymphocytes,
CD3+, CD4 + CD8+,
CD19+, CD56 + or
ratio of CD4+/CD8
+ cells; nor for IgG,
IgA,IgM or IgE..
Barone iVNS Refractory epilepsy no na pre:0.09 ± na pre:0.06 ± 0.05; na TNF-α = ns; Comparison of pre- No elevation of 4
et al. 0.01; post:0.06 ± 0.07 IL-6 = ns; implantation values inflammatory
2007 post:0.10 ± CRP = ns with values acquired parameters reported
0.03 3 months post- at baseline.
surgery. HF-HRV
and rmssd obtained
of a 24-h ECG non-
significantly
decreased after 3
months of VNS.
Boström tVNS healthy/migraine no na nr na nr na IL-1β↑ * Cytokine levels at No elevation of 2
et al. (refractory baseline and 3 inflammatory
2019 headache disorder) months post- parameters reported
245
VNS compared to
sham stimulation was
associated with a
reduction in severe
migraine attacks. No
246
De Herdt iVNS Refractory epilepsy yes na pre:1007 na pre:69445 na TNF-α = ns; PBMCs, LPS induced No elevation of 4
et al (ex (71–7658); (38251–184922); IL-6 = ns; IL- stimulation ex vivo. inflammatory
vivo) 10 = ns; In the 2 patients with parameters reported
(continued on next page)
C. Schiweck et al.
Table 1 (continued )
post:927 post: 49,869 CRP = ns,IL- generalized epilepsy, at baseline.
(22–3571) (32098–147871) 8 ↓* there was a marked
increase in IL-10 Seizure reduction
induction after 6 (>30 %) was reported
months of VNS, in 3/10 treated
compared to the patients.
patients with
localized epilepsy.
Drewes tVNS Rheumatoid no na pre:9.79 ± na pre:5.92 ± 15.14; na TNF-α = ns; See above Inclusion based on 4
et al., Arthritis 19.22; post: 4.10 ± 5.52 IL-6 = ns; IL- disease severity:
2021 post:9.92 ± 10: ↓* (low -high-disease activity
20.15 RA disease group (DAS-28 > 3.2)
activity)(p showed clinical
= 0.02),ns improvement
(high RA on days 2 and 5.
disease -low disease acitivity
activity); group (<2.6) showed
no clinical
improvement.
Koopman iVNS Rheumatoid no na pre:2900.00 na nr na TNF-α↓**, The reduction of Only participants 4
et al. Arthritis ± 566.00; CRP↓*** cytokine levels was with active disease
2016 post:1776.00 found in the state were included
± 342.00 combined cohort. (mean DAS-28 across
After turning off VNS both cohorts: 5.99).
on d 42, TNF-α
increased till d 56. Disease severity in the
And, when combined cohorts
stimulation was was substantially
247
intestinal
inflammation did not
change after VNS.
Salama tVNS Patients undergoing No nr nr pre: 63.4 ± pre:53.20 ± 20.00; IL-6 = ns; CRP IL-6 ↓ * (p < Comparison to From baseline to 1st 1
et al. surgery for lung 29.0; post:93.70 ± 32.90 = ns; IL-10 = ns 0.05); CRP controls without POD sharp increase in
2020 cancer post:98.30 ± = ns; IL-10 sham condition. IL-6 levels, 4th POD:
39.70 ns Differences of tVNS return to BL values.
on the acute
inflammatory VNS group showed
response after lung shorter
surgery (baseline = hospitalization time
pre-operative value, and lower incidence
post-intervention = of pneumonia.
4th postoperative
day(POD)).
Stavrakis transvenous patients undergoing no pre:46.2 ± pre:46.70 ± pre:700.3 ± pre:696.30 ± TNF-α ↓* (p = TNF-α ↓ * * Stimulation was Inflammatory 1
et al. VNS cardiac surgery 1.5; 1.40; 94.5; 181.30; 0.01); IL-6 ↓* *; IL-6 ↓ * * initiated after parameters were
2017 post:38.30 post:33.00 ± post:275.30 post:147.10 ± (p = 0.03); IL- *; IL-10 ↓ * * cardiac surgery upon increased at baseline
± 1.40 1.40 ± 66.20 22.20 10 = ns; CRP = *; CRP ↓ * * * arrival the intensive immediately after
ns care unit. surgery and
decreased over time.
Blood samples were
drawn after arrival Atrial fibrillation free
the ICU as well as 24 survival was better in
h and 72 h later. the tNVS group.
Length of
249
See above
Stavrakis tVNS paroxysmal AF no pre:5.4 pre:6.9 pre: 2.0 pre:1.6 (0.5–3.7); TNF-α ↓** (p = nr The average burden No elevation of 1
et al., (4.3–9.1); (4.6–9.3); (0.9–3.7); post:1.5 (0.7–3.2) 0.009); IL-6 = of Atrial fibrillation inflammatory
2020 post:5.4 post:5.5 post:2.3 ns; IL-10 = ns, after 3 and 6 months parameters reported
(4.4–8.3) (3.9–8.3) (1.0–4.4) IL-1β = ns; IL- VNS was lower at baseline.
17 = ns compared to
Studies in systematic
Review only (not
included in meta-
analysis)
First Intervention Diagnosis LPS Assessed cytokines Direction and Findings Quality*
author, significance
year
between within between within
Andreas tVNS Patients after no CRP, IL-6 IL-6 = ns; CRP = ns nr -tVNS significantly reduced nr CRP increased 2 day 1
et al. cardiac surgery incidence of POAF (p = after end of surgery and
(2019) 0.022) returned to normal
-duration and time of onset ranger onday 7. IL-6
of arrhythmia remained was increased at the
uninfluenced-no difference end of surgery and
with tVNS for CRP or IL-6 returned to normal
levels range on day 7.
250
No effect of tVNS on
inflammation.
(2019) Sjörgeń s TNF-α, MIP-1α, MIP-1α↓; TNF-α↓; using PRO-F) and dtime elevated inflammatory
Syndrome IFNα, IL-10, IL-1β, (undetectable: IFN-γ, sleepiness (measured by parameters at baseline
IL-6, IP-10 IL12-p70, IFNα, IL-10) ESS) were significantly were not considered as
reduced across three TPs (p inlusion criteria.
< 0.0003 and 0.02)-tVNS
-> levels of MIP1α, IL-1β, LPS-induced release of
TNF-α, IL-6, and IP-10 were all inflammatory
significantly reduced across parameters decreased
the 5 TPs regardless of after 28 days.
responder grouping
Day time sleepiness and
fatigue were improved
after 7 days.
Tornero tVNS COVID-19 No CRP, PCT, IL-1β, IL- CRP↓; PCT↓;, IL-1β NA CRP and PCT levels nr No significant X
et al. 6, TNF-α, D-Dimer = ns; IL-6 = ns; decreased to a significantly differences for clinical
2019), acute COVID-19 infection (Corrêa, 2022; Tornero, 2022; Seitz, 3.3. Quantitative synthesis
2022), sepsis (Wu, 2023), psychological stress (Bremner, 2020)) and
except for one study (Andreas, 2019) all showed that VNS alleviated the Meta-analysis was possible for 27 studies, with pooled sample sizes
heightened inflammatory response. These findings indicate that the ranging from 85/88 (VNS/sham) in the short-term between analyses up
onset of stimulation in relation to the acute inflammatory event as well to 273 (VNS) in the short-term within analysis.
as the temporal dynamics of the cytokine levels during the inflammatory
response may be relevant for the efficacy of VNS. 3.3.1. Short-term between and within- group analysis
The predetermined minimum number of articles (n = 5) to conduct a
meta-analysis, was only reached for IL-6 and TNF-α. A random effects
model revealed no difference between sham and VNS groups for either
Fig. 2. Forest-Plots of Meta-analysis for short-term between and within-group comparison. Analysis was performed between verum and control intervention
for A. IL-6 and B. TNF-α. Within-group comparison of pre- vs. post- values were performed in the verum group for C. IL-6 and D. TNF-α. Concentrations for cytokines
were given in pg/mL, for CRP in mg/L.
253
C. Schiweck et al. Brain Behavior and Immunity 116 (2024) 237–258
Fig. 2. (continued).
cytokine (TNF-α: g = − 0.21, 95 % CI: − 0.74–0.32, P = 0.359; I2 = 61.5 3.4. Sub-group analysis/ Meta-regression: Stimulation duration, LPS,
%, 95 % CI: [43.9 %; 89.0 %]) or IL-6 (g = − 0.94, 95 % CI: [− 2.18; stimulation length and study quality
0.31], p = 0.112, I2 = 91.0 %, 95 % CI: [83.1 %; 95.2 %]) (See Fig. 2A-
B.) No within-group changes for VNS were found for TNF-α (g = − 0.45, For all cytokines where at least 10 studies were identified we per
95 % CI: [− 2.59; 1.70], p = 0.630, I2 = 97 %, 95 % CI: [95.4 %; 98.0 %]) formed a within-group analysis for stimulation method (iVNS vs non-
or IL-6 (g = 0.28, 95 % CI: [− 3.11; 3.67], p = 0.840, I2 = 97.6 %, 95 % invasive VNS), stimulation duration, LPS stimulation and study qual
CI: [96.3 %; 98.4 %]). Prediction intervals of all analyses indicated that ity. No difference between subgroups was found for any of the moder
future studies may find both positive and negative effects. Fig. 2C-D for ating variables: stimulation method all P > 0.05 and stimulation
forest plots. Data from (Stavrakis, et al., 2017) are for the femoral vein duration (all r2 = 0 % and P > 0.05), except for LPS, where IL-1β was
sampling, data from (Kox, 2015) were used for the 240-minute time lower in studies which had used an LPS stimulation versus those who did
point. not. While iVNS generally had stronger reductive effects than tVNS, the
difference between both treatments was not significant, and confidence
3.3.2. Long-term between-group analysis intervals for the separate effects per treatment always included 0. See
In the long-term analyses, no difference of VNS compared to sham Supplementary Table 3 for detailed results of the stimulation method,
stimulation was found for any cytokine (TNF-α: g = − 0.13, 95 % CI: Supplementary Fig. 3 and Supplementary Table 4 shows results for
[− 0.35–0.10], p = 0.196; IL-6: g = − 0.67, 95 % CI: [− 2.10; 0.76], p = the stimulation duration, and Supplementary Table 5 results for LPS.
0.306; IL-1β: g = − 0.02, 95 % CI: [− 0.06; 0.02], p = 0.270; IL-10 g = Next, we tested whether study quality had an influence on the results
− 0.08, 95 % CI:[− 0.30; 0.14], p = 0.387). Heterogeneity was large for using the quality scores derived with the instruments (NLHBI). For TNF-
IL-6 (I2 = 91.1 % CI: [84.8 %; 94.7 %]). Point estimates were low, but α, (4 poor, 11 fair, 2 good quality), IL-6, (3 poor, 9 fair, 3 good quality),
uncertainty intervals were large for TNF-α (I2 = 0.0 %, CI: [0.0 %-79.2 IL-10 (2 poor, 9 fair 3 good quality) and CRP (3 poor, 6 fair 2 good
%]), IL-10 (I2 = 0.0 % [0.0 %; 74.6 %]) and IL-1β (I2 = 0.0 %, CI: [0.0 quality), study quality did not influence results (TNF-α: Q = 1.79, P =
%-79.2 %]). Prediction intervals included both positive and negative 0.408; IL-6: Q = 0.31, P = 0.855; IL-10: Q = 1.63, P = 0.442, CRP =
effect sizes. The forest plots are depicted in Fig. 3. 0.19, P = 0.911). For IL-1β, only 1 study had poor and good quality
respectively, and analyses was not possible.
3.3.3. Long-term within-group analysis
Eight cytokines were reported in at least 5 studies and TNF-α in 17
studies for the long-term within-group analysis. We did not find a sig 3.5. Publication bias and sensitivity analysis
nificant difference after VNS for any cytokine. (Table 2), forest plots can
be found in Supplementary Fig. 1. Heterogeneity was high in all Inspection of funnel plots revealed some asymmetry for IL-1β, CRP
studies, with only two analyses below an I2 of 75 % (Table 2). and IL-10. No statistically significant results were observed for TNF-α (P
= 0.112), IL-6 (P = 0.944), IL-10 (P = 0.160), CRP (P = 0.095) using
3.3.4. Additional analyses within inflammatory subgroup Egger’s test. However, a significant Egger’s test was detected for IL-1β
All above analyses were repeated including only studies where acute (bias estimate: p = 0.017), with more studies showing a reductive
inflammation, as evidenced by increased inflammation scores at base change being published. Funnel plots are displayed in Supplementary
line or the induction of an inflammatory event (i.e., via LPS adminis Figure 4. Influential case analysis did not alter results regarding sig
tration, surgical intervention) took place. A significantly stronger nificance, but narrowed down prediction intervals for effect sizes and
decrease in the VNS than sham condition was found for CRP for the long- did change the direction of some effects (See Supplementary Table 7).
term between group results (g = − 0.58, 95 %CI: [− 1.09; − 0.061], p = A particularly prevalent outlier was the study by Stavrakis et al. (2017)
0.038), see Supplementary Fig. 2. Although all other results showed a (7/13 outliers in total). No more than 1 outlier per analysis was found.
decreasing effect of VNS on inflammatory parameters, they remained
non-significant. See Supplementary Fig. 2 for details. 4. Discussion
254
C. Schiweck et al. Brain Behavior and Immunity 116 (2024) 237–258
Fig. 3. Forest-Plots of Meta-analysis for long-term between-group analysis. Comparison of verum vs. control condition was performed after long-term treatment
for A. TNF-α, B. IL-6, C. IL-1β and D. IL-10. Concentrations for cytokines were given in pg/mL, for CRP in mg/L.
255
C. Schiweck et al. Brain Behavior and Immunity 116 (2024) 237–258
Table 2
Summary of meta-analytic test results with random effects model for long-term within-group comparison. Cytokines were measured in pg/mL, CRP in mg/L. Ab
breviations: k = Number of studies; g = Hedge’s g, 95 %CI: 95 % confidence Interval; t = t value, p = p value, PI = 95 % Prediction Interval, I2: Heterogeneity estimate,
I2 95 %CI: Confidence interval for heterogeneity.
Cytokine k g 95 %CI t P 95 %PI I2 I2 95 %CI
majority of studies in our systematic review reported a significant line, the subgroup analysis of four long term studies with an acute in
change of at least one inflammatory protein in at least one subgroup flammatory event showed lower CRP values after VNS compared to
analysis when exposed to VNS, we found no statistically convincing sham stimulation. However, due to the heterogeneity in experimental
effect of VNS on any inflammatory cytokine in our meta-analysis, neither designs, the results of these studies were not uniform (i.e., we did not
for long-term (between-group: IL-6, TNF-⍺, IL-1β, IL-10; within-group: find these effects for the other cytokines and the other subgroups), yet
TNF-⍺, IL-1β, IL-6, IL-8, IL-10, and CRP) nor short-term stimulation they clearly underscore the potential of VNS to mitigate acute inflam
(between- and within-group: TNF-⍺, IL-6). Only the subgroup analysis of matory responses rather than suppressing chronically elevated cyto
4 long-term studies with acute inflammatory events was significant: CRP kines. Moreover, the timing of VNS to interfere/ reduce distinct
values were lower after VNS compared to sham stimulation. Yet, no inflammatory processes seems to be pivotal for VNS efficacy. Although it
differences were evident for the other cytokines in any setting. In gen was not statistically possible to assess this effect in our analysis, this
eral, heterogeneity (I2) and/or confidence intervals for I2 were high. The should be considered in future studies. It is also plausible that VNS’s
inclusion of modifying factors such as stimulation duration, LPS stimu anti-inflammatory effect has a cumulative dynamic by inducing neuro
lation of cytokines, stimulation procedure (invasive vs non-invasive), or plasticity, comparable to its impact on clinical outcomes in epilepsy or
study quality as well as the exclusion of influential cases did not change depression treatment, as well as on brain signalling (Kibleur, 2018).
the results. When interpreting the present findings, it is crucial to note the variations
This result holds considerable significance given the attention elec across studies in key aspects, including patient samples (11 different
trical VNS has gained across medical disciplines following Borovikova disorders), proteins tested (>10), and the condition under which these
et al.’s landmark discovery of VNS’s capacity to decrease TNF-⍺ levels proteins were assessed (e.g., baseline with and without elevated in
and prevent septic shock in rodents during endotoxemia (Borovikova, flammatory levels, after systemic or ex-vivo LPS-stimulation, or intra
2000). Over the past two decades, an increasing number of studies have operatively). In addition, the quality of most studies ranged from poor to
explored the VNS’s anti-inflammatory potential, with renewed attention fair: The sample sizes were small (iVNS: 5–14, tVNS: 5–62) and only a
during the COVID-19 pandemic. The fact that the meta-analysis hardly limited number of studies employed appropriate control conditions. This
provided any significant findings, despite the promising results in rodent is due to 1) the invasive nature of iVNS, and 2) the lack of standardized
studies (Kwan, 2016) and a substantial number of positive findings in sham conditions. Particularly for iVNS sham conditions are difficult to
our own systematic review, calls for an exploration of the underlying realize: Current procedures such as leaving the device turned off (easily
reasons and a refocusing in the research field to advance this innovative detectable by the patients due to missing side effects such as a hoarse
technology. voice) or using low-frequency stimulation (1 Hz-stimulation impacts
In line with Borovikova et al. (Borovikova, 2000), rodent studies brain signaling and clinical symptoms) (Cao et al., 2021; Straube et al.,
mostly administered VNS prior to an inflammatory event (e.g., LPS- 2015) have been proven inadequate. Since the introduction of
infusion) (30 out of 32 studies (Kwan, 2016)). In contrast, most non-invasive VNS, more RCTs have been conducted. However, these
human studies have focused on VNS in chronic and often treatment- studies show a large variability in stimulation methodology (differences
resistant conditions. In these cases, chronic inflammation and prior in stimulation site, side, duration of stimulation, parameter settings and
(anti-inflammatory) therapies may have caused compensatory mecha control condition). The primary obstacle to a comprehensive investi
nisms (e.g. immune cell activation, altered cytokine production, gation of optimal stimulation settings is the lack of easily accessible
changes/dysfunctionality of T-and B-cell subpopulations, neuronal biomarkers: On the one hand, biological measures are required for pa
reconfiguration) which may confound the anti-inflammatory effects of tient stratification. For instance, identifying patients with elevated in
VNS (Cain et al., 2009; Li, 2022). Moreover, only a minority of studies on flammatory parameters and/or reduced “vagal tone” by assessing heart
chronic diseases (Koopman, et al., 2016; Bonaz, 2016; Addorisio, et al., rate variability has been discussed as useful options (Pellissier, 2014).
2019; Drewes et al., 2021; Sinniger, 2020) defined disease activity and/ Second, we need efficacy measures for VNS. In this context, traditional
or elevated inflammatory parameters as inclusion criteria and demon measures such as heart rate variability and pupillary reactivity have
strated a beneficial effect of VNS. In most of the studies, however, failed to provide reliable information about VN activation thus far
stratification was not performed, resulting in samples which may or may (Vespa, 2022; Wolf et al., 2021; Keute et al., 2019) most probably due to
not have had elevated baseline levels of inflammatory proteins. methodological limitations. Current advances in the field are showing
In contrast, clinical studies that performed VNS during or immedi first promising results and provide new insights into the mechanisms of
ately after inflammatory events - such as surgeries (Stavrakis, et al., vagal signalling. For instance, phasic burst stimulation in contrast to
2017; Salama et al., 2020; Stakenborg, 2017), acute COVID-19 infection tonic VNS is able to modulate pupillary reactivity during the pupillary
(Corrêa, 2022; Tornero, 2022; Seitz, 2022), sepsis (Wu, 2023) and light reflex (Wienke et al., 2023; D’Agostini, 2023). Moreover, explo
emotional stress (Bremner, 2020)– did confirm the anti-inflammatory ration of more direct measures such as somatosensory evoked potentials
potential of VNS, with only a minority of small-scale studies showing (Fallgatter, 2003; de Gurtubay et al., 2021) or microneurography of the
contrary outcomes (Kox, 2015; Corrêa, 2022; Andreas, 2019). Particu VN (Ottaviani et al., 2020) could provide a crucial step in closing this
larly striking were the findings of peri- or postoperative VNS, which gap.
significantly reduced cytokine levels amid surgical inflammation. In It is still unknown how electrical VNS interferes with afferent and
256
C. Schiweck et al. Brain Behavior and Immunity 116 (2024) 237–258
efferent signals transmitted via the VN to induce its anti-inflammatory curation, Writing – original draft. Leona Jacobsen: Investigation, Data
effect. Except for tVNS at the cymba conchae (Burger et al., 2020; curation, Writing – original draft. Andreas Reif: Writing – original draft,
Warren, 2019) that exclusively stimulates the afferent sensory fibers of Writing – review & editing. Sharmili Edwin Thanarajah: Conceptu
the auricular branch of the VN (Kaniusas, 2019), current devices non- alization, Methodology, Validation, Investigation, Data curation,
selectivity engage afferents and efferents as well as fibres of different Writing – original draft, Writing – review & editing, Visualization, Su
diameter and myelination status. Efferent fiber stimulation is believed to pervision, Project administration.
activate the canonical CAP (anterograde) and the efferent brainstem
nuclei (retrograde). Afferent fiber stimulation, conversely, activates the Declaration of competing interest
NTS (anterograde), leading to a.) CAP activation through efferent nuclei,
b.) initiation of the vagus-sympathetic reflex via C1 neurons in the The authors declare that they have no known competing financial
medulla oblongata, and c.) engagement of the Hypothalamic-Pituitary- interests or personal relationships that could have appeared to influence
Adrenal (HPA) axis through hypothalamic nuclei (Tanaka, 2021). the work reported in this paper.
Additionally, retrograde afferent stimulation can prompt the release of
anti-inflammatory neuropeptides at nerve endings (Chavan et al., 2017), Data availability
that can have a direct impact on intra-tissue cytokines, that are not
considered in the meta-analysis due to the lack of data (evidence on Data will be made available on request.
intratissue cytokines was only provided by one study (Sinniger, 2020)..
In conclusion, non-selective stimulation of afferent and efferent circuits Acknowledgments
may have a potentiating effect. Moreover, the spatial arrangement of
different morphological types of fibre within the nerve is organ- and CS and SET received intramural funding (Fokus Grant) to investigate
function-specific changing along the nerve (Jayaprakash, 2023) and Vagus nerve stimulation. SET was funded by the Else-Kröner Frisenius
different between the right and left nerves (Han, 2018; Jayaprakash, Stiftung and by the Leistungszentrum Innovative Therapeutics (Ther
2023). In line with this, selective stimulation of distinct fascicles has aNova) funded by the Fraunhofer Society and the Hessian Ministry of
been shown to achieve organ-specific stimulation in pigs (Jayaprakash, Science and Art.
2023). Hence, we urgently need selective stimulation techniques to
achieve a breakthrough with this technology. Appendix A. Supplementary data
257
C. Schiweck et al. Brain Behavior and Immunity 116 (2024) 237–258
Cain, D., Kondo, M., Chen, H., Kelsoe, G., 2009. Effects of acute and chronic Merchant, K., Zanos, S., Datta-Chaudhuri, T., Deutschman, C.S., Sethna, C.B., 2022.
inflammation on B-cell development and differentiation. J, Invest. Dermatol. 129, Transcutaneous auricular vagus nerve stimulation (taVNS) for the treatment of
266–277. pediatric nephrotic syndrome: a pilot study. Bioelectron Med 8, 1.
Cao, J., et al., 2021. Different modulation effects of 1 Hz and 20 Hz transcutaneous Mion, F., et al., 2020. Transcutaneous auricular vagus nerve stimulation for the
auricular vagus nerve stimulation on the functional connectivity of the treatment of irritable bowel syndrome: a pilot, open-label study. Bioelectron Med
periaqueductal gray in patients with migraine. J. Transl. Med. 19, 354. (lond) 3, 5–12.
Chaudhry, S.R., et al., 2019. Inter-ictal assay of peripheral circulating inflammatory Ottaviani, M.M., Wright, L., Dawood, T., Macefield, V.G., 2020. In vivo recordings from
mediators in migraine patients under adjunctive cervical non-invasive vagus nerve the human vagus nerve using ultrasound-guided microneurography. J Physiol 598,
stimulation (nVNS): A proof-of-concept study. Brain Stimul 12, 643–651. 3569–3576.
Chaudhry, S., Morrissey, D., Woledge, R.C., Bader, D.L., Screen, H.R.C., 2015. Eccentric Pellissier, S., et al., 2014. Relationship between Vagal Tone, Cortisol, TNF-Alpha,
and concentric loading of the triceps surae: an in vivo study of dynamic muscle and Epinephrine and Negative Affects in Crohn’s Disease and Irritable Bowel Syndrome.
tendon biomechanical parameters. J Appl Biomech 31, 69–78. PLoS One 9, e105328-.
Chavan, S.S., Pavlov, V.A., Tracey, K.J., 2017. Mechanisms and therapeutic relevance of R Core Team, 2023. R: A Language and Environment for Statistical Computing.
neuro-immune communication. Immunity 46, 927–942. R Foundation for Statistical ComputingVienna, Austria.
Corcoran, C., Connor, T. J., O’Keane, V. & Garland, M. R. The effects of vagus nerve Rosas-Ballina, M., et al., 2011. Acetylcholine-Synthesizing T Cells Relay Neural Signals in
stimulation on pro- and anti-inflammatory cytokines in humans: A preliminary a Vagus Nerve Circuit. Science 1979 (334), 98–101.
report. 12, 307–309 (2005). Salama, M., Akan, A., Mueller, M.R., 2020. Transcutaneous stimulation of auricular
Corrêa, F.I., et al., 2022. Transcutaneous Auricular Vagus Nerve Stimulation Improves branch of the vagus nerve attenuates the acute inflammatory response after lung
Inflammation but Does Not Interfere with Cardiac Modulation and Clinical lobectomy. World J Surg 44, 3167–3174.
Symptoms of Individuals with COVID-19: A Randomized Clinical Trial. Life 12, Seitz, T., et al., 2022. Percutaneous auricular vagus nerve stimulation reduces
1644. inflammation in critical covid-19 patients. Front Physiol 13, 897257.
D’Agostini, M., et al., 2023. Short bursts of transcutaneous auricular vagus nerve Sinniger, V. et al. A 12-month pilot study outcomes of vagus nerve stimulation in Crohn’s
stimulation enhance evoked pupil dilation as a function of stimulation parameters. disease. 32, (2020).
Cortex 159, 233–253. Stakenborg, N., et al., 2017. Abdominal vagus nerve stimulation as a new therapeutic
D’Haens, G., et al., 2023. Neuroimmune Modulation Through Vagus Nerve Stimulation approach to prevent postoperative ileus. Neurogastroenterol. Motil. 29, e13075.
Reduces Inflammatory Activity in Crohn’s Disease Patients: A Prospective Open- Stavrakis, S., et al., 2020. TREAT AF (Transcutaneous Electrical Vagus Nerve Stimulation
label Study. J Crohns Colitis jjad151. https://doi.org/10.1093/ecco-jcc/jjad151. to Suppress Atrial Fibrillation): A Randomized Clinical Trial. JACC Clin
de Gurtubay, I.G., Bermejo, P., Lopez, M., Larraya, I., Librero, J., 2021. Evaluation of Electrophysiol 6, 282–291.
different vagus nerve stimulation anatomical targets in the ear by vagus evoked Stavrakis, S. et al. Low-Level Transcutaneous Electrical Vagus Nerve Stimulation
potential responses. Brain Behav 11, e2343. Suppresses Atrial Fibrillation. (2015).
De Herdt, V. et al. Effects of vagus nerve stimulation on pro- and anti-inflammatory Stavrakis, S. et al. Focus on neuromodulation of cardiac arrhytmias Low-Level Vagus
cytokine induction in patients with refractory epilepsy. 214, 104–108 (2009). Nerve Stimulation Suppresses Post-Operative Atrial Fibrillation and Inflammation A
de Moraes, T.L., et al., 2023. Brief periods of transcutaneous auricular vagus nerve Randomized Study. (2017).
stimulation improve autonomic balance and alter circulating monocytes and Straube, A., Ellrich, J., Eren, O., Blum, B., Ruscheweyh, R., 2015. Treatment of chronic
endothelial cells in patients with metabolic syndrome: a pilot study. Bioelectron Med migraine with transcutaneous stimulation of the auricular branch of the vagal nerve
9, 7. (auricular t-VNS): a randomized, monocentric clinical trial. J Headache Pain 16, 63.
Drewes, A. M. et al. Short-term transcutaneous non-invasive vagus nerve stimulation may Tanaka, S., et al., 2021. Vagus nerve stimulation activates two distinct neuroimmune
reduce disease activity and pro-inflammatory cytokines in rheumatoid arthritis: circuits converging in the spleen to protect mice from kidney injury. Proc. Natl.
results of a pilot study. 50, 20–27 (2021). Acad. Sci. 118 e2021758118.
Fallgatter, A.J., et al., 2003. Far field potentials from the brain stem after transcutaneous Tarn, J., Legg, S., Mitchell, S., Simon, B., Ng, W.-F., 2019. The effects of noninvasive
vagus nerve stimulation. J Neural Transm 110, 1437–1443. vagus nerve stimulation on fatigue and immune responses in patients with primary
Han, W., et al., 2018. A neural circuit for gut-induced reward. Cell 175, 665–678. Sjögren’s syndrome. Neuromodulation: Technol. Neural Interface 22, 580–585.
Harrer, M., Cuijpers, P., Furukawa, T., Ebert, D., 2021. Doing meta-analysis with R: A Tornero, C., et al., 2022. Non-invasive vagus nerve stimulation for COVID-19: results
hands-on guide. (chapman and Hall/CRC. from a randomized controlled trial (SAVIOR I). Front Neurol 13.
Harrer, M., Cuijpers, P., Furukawa, T. & Ebert, D. D. dmetar: companion R package for Tracey, K.J., 2002. The inflammatory reflex. Nature 420, 853–859.
the guide ‘doing meta-analysis in R’. R package version 0.1.0. 2019. Preprint at Van Maanen, M.A., Vervoordeldonk, M.J., Tak, P.P., 2009. The cholinergic anti-
(2019). inflammatory pathway: towards innovative treatment of rheumatoid arthritis. Nat
Jayaprakash, N., et al., 2023. Organ-and function-specific anatomical organization of Rev Rheumatol 5, 229–232.
vagal fibers supports fascicular vagus nerve stimulation. Brain Stimul 16, 484–506. Veiz, E., et al., 2022. Increased concentrations of circulating interleukins following non-
Kaniusas, E., et al., 2019. Current directions in the auricular vagus nerve stimulation I–a invasive vagus nerve stimulation: results from a randomized, sham-controlled,
physiological perspective. Front Neurosci 854. crossover study in healthy subjects. Neuroimmunomodulation 29, 450–459.
Kelly, M.J., Breathnach, C., Tracey, K.J., Donnelly, S.C., 2022. Manipulation of the Vespa, S., et al., 2022. Characterization of vagus nerve stimulation-induced pupillary
inflammatory reflex as a therapeutic strategy. Cell Rep Med 3. responses in epileptic patients. Brain Stimul 15, 1498–1507.
Keute, M., Demirezen, M., Graf, A., Mueller, N.G., Zaehle, T., 2019. No modulation of Vida, G., et al., 2011. β2-Adrenoreceptors of regulatory lymphocytes are essential for
pupil size and event-related pupil response by transcutaneous auricular vagus nerve vagal neuromodulation of the innate immune system. FASEB J. 25, 4476.
stimulation (taVNS). Sci Rep 9, 11452. Viechtbauer, W., 2010. Conducting meta-analyses in R with the metafor package. J Stat
Kibleur, A., et al., 2018. Electroencephalographic correlates of low-frequency vagus Softw 36, 1–48.
nerve stimulation therapy for Crohn’s disease. Clin. Neurophysiol. 129, 1041–1046. Viechtbauer, W., Cheung, M.W., 2010. Outlier and influence diagnostics for meta-
F.A. Koopman et al. Vagus nerve stimulation inhibits cytokine production and attenuates analysis. Res Synth Methods 1, 112–125.
disease severity in rheumatoid arthritis. doi:10.1073/pnas.1605635113. (2016). Wan, X., Wang, W., Liu, J., Tong, T., 2014. Estimating the sample mean and standard
Kox, M., et al., 2015. Transvenous vagus nerve stimulation does not modulate the innate deviation from the sample size, median, range and/or interquartile range. BMC Med
immune response during experimental human endotoxemia: A randomized Res Methodol 14, 1–13.
controlled study. Arthritis Res Ther 17, 150. Warren, C.M., et al., 2019. The neuromodulatory and hormonal effects of transcutaneous
Kwan, H., et al., 2016. Vagus nerve stimulation for treatment of inflammation: vagus nerve stimulation as evidenced by salivary alpha amylase, salivary cortisol,
systematic review of animal models and clinical studies. Bioelectron Med 3, 1–6. pupil diameter, and the P3 event-related potential. Brain Stimul. Basic Transl. Clin.
Lerman, I., et al., 2016. Noninvasive transcutaneous vagus nerve stimulation decreases Res. Neuromodul. 12, 635–642.
whole blood culture-derived cytokines and chemokines: a randomized, blinded, Wienke, C., Grueschow, M., Haghikia, A., Zaehle, T., 2023. Phasic, Event-Related
healthy control pilot trial. Neuromodulation: Technology at the Neural. Interface 19, Transcutaneous Auricular Vagus Nerve Stimulation Modifies Behavioral, Pupillary,
283–291. and Low-Frequency Oscillatory Power Responses. J. Neurosci. 43, 6306.
Li, H., et al., 2022. Macrophages, chronic inflammation, and insulin resistance. Cells 11, Wolf, V., Kühnel, A., Teckentrup, V., Koenig, J., Kroemer, N.B., 2021. Does
3001. transcutaneous auricular vagus nerve stimulation affect vagally mediated heart rate
Lüdecke, D., Lüdecke, M. D. & Calculator’from David, B. W. Package ‘esc’. R Package variability? A living and interactive Bayesian meta-analysis. Psychophysiology 58,
Version 0.5 1, 2019 (2019). e13933.
Majoie, H.J.M., et al., 2010. Vagus Nerve Stimulation in Refractory Epilepsy: Effects on Wu, Z., et al., 2023. Transcutaneous auricular vagus nerve stimulation reduces cytokine
Pro- and Anti-Inflammatory Cytokines in Peripheral Blood. production in sepsis: An open double-blind, sham-controlled, pilot study. Brain
Neuroimmunomodulation 18, 52–56. Stimul 16, 507–514.
Zhou, Q. et al. Effect of transcutaneous auricular vagus nerve stimulation on delayed
neurocognitive recovery in elderly patients. 34, 2421–2429 (2022).
258