Splenic neuromodulation as monotherapy and in combination with biologic or targeted synthetic DMARDs for difficult-to-treat rheumatoid arthritis (ConsiderRAte): a randomised, sham-controlled, early feasibility trial

Summary

Background

Preclinical and early human evidence indicate that splenic neuromodulation modulates immune activity either independently or in conjunction with biologic or targeted synthetic (b/ts) disease modifying antirheumatic drugs (DMARDs). This study evaluated the safety, tolerability, and preliminary effectiveness of splenic nerve stimulation (SpNS) as monotherapy and in combination with b/tsDMARDs in adult patients with difficult-to-treat rheumatoid arthritis.

Methods

ConsideRAte was an early-feasibility, multicentre, randomised sham-controlled trial of an implantable device (Galvani System) delivering SpNS conducted across six centres in the USA and Netherlands. Adults with difficult-to-treat rheumatoid arthritis, defined by inadequate response to at least two b/tsDMARDs, were implanted and randomised 1:1 to active (STIM ON) or sham (STIM OFF) stimulation following five sentinel patients. Participants received 12 weeks of stimulation in the monotherapy phase, defined as stimulation added to stable conventional synthetic DMARDS (csDMARDs) and/or low dose corticosteroids. Participants with Disease Activity Score using Erythrocyte Sedimentation Rate (DAS28-ESR) improvement of ≥1.2 continued stimulation for an additional 12 weeks; non-responders discontinued stimulation and received baricitinib for 12 weeks. Thereafter, participants who were not in remission could receive adjunct tsDMARDs (baricitinib) together with SpNS in an open-label phase for six months, after which the participants moved to long-term-follow-up. The primary objective was to assess safety and tolerability of the Galvani System, its implantation and subsequent stimulation; key secondary outcomes included assessment of disease activity Disease Activity Score using C-Reactive Protein (DAS28-CRP) and patient-reported outcomes. This study is registered with ClinicalTrials.gov (NCT05003310).

Findings

Between January 2022 and November 2024, eleven participants (Five Open Label STIM ON, six randomised to STIM ON (n = 3) and STIM OFF (n = 3)) were successfully implanted with the Galvani System. No device-related serious adverse events, or unanticipated adverse device effects occurred. During the 12-week monotherapy phase, mean DAS28-CRP decreased by 1.02 ± 0.45 (mean ± Standard Error, [SE]) in the STIM ON group compared with −0.34 ± 0.14 in the STIM OFF group. Five of eight (62.5%) STIM ON participants achieved a clinically meaningful improvement (≥1.0-point reduction). Nine participants entered the adjunct phase, with mean (±SE) DAS28-CRP decreasing with −2.95 ± 0.56 points at week 12 and -2.28 ± 0.69 points at week 24; 55.6% (5/9) achieved low disease activity (LDA) and 33.3% (3/9) remission. Two participants remained on monotherapy. In exploratory analysis, biomarkers were assessed in the monotherapy phase at the end of 12 weeks and trends for systemic proinflammatory cytokine reductions and in TH17 (T-helper 17) cell populations were observed in responders as defined by DAS28-CRP changes.

Interpretation

SpNS demonstrated a favourable safety profile and was well tolerated in this small cohort. Clinical improvements were observed for monotherapy, and enhanced outcomes were observed for adjunct treatment resulting in multiple participants reaching LDA or remission. By engaging neuro-immune pathways rather than targeting single cytokines, this approach represents a novel therapeutic option that warrants further evaluation as stand-alone therapy or to complement existing pharmacological treatments for difficult-to-treat rheumatoid arthritis patients.

Funding

Galvani Bioelectronics.

Research in context.

Evidence before this study

We searched PubMed on 17 November 2025 using the terms (“rheumatoid arthritis”) AND (“neuromodulation”) AND (“splenic”) AND (“clinical trial”), from database inception with no language or date restrictions. This search returned no completed clinical trials of splenic nerve stimulation in rheumatoid arthritis (RA). Because splenic nerve stimulation is an emerging field, we performed an additional search using (“splenic nerve”) AND (“stimulation” OR “neuromodulation”) AND (“human”), which identified one relevant study: an acute intraoperative proof-of-concept demonstrating that electrical stimulation of the splenic neurovascular bundle was technically feasible in humans undergoing esophagectomy and was accompanied by a reduction in post-operative C-Reactive Protein (CRP) levels. We then broadened the search to neuromodulation approaches that activate the inflammatory reflex in RA and identified five publications related to vagus nerve stimulation (VNS) as the only neuromodulation modality with published clinical data in RA. These trials demonstrated that auricular and cervical VNS with implantable and non-implantable devices could reduce disease activity in some patients with RA, supporting autonomic-immune modulation as a therapeutic mechanism, but they do not directly target the spleen.

To assess biological plausibility, we reviewed preclinical studies identified through backward citation searches. These studies show that direct splenic nerve activation suppresses Tumour Necrosis Factor (TNFα) and other cytokines in rodent and porcine models of acute inflammation, arthritis, and colitis, and that ex vivo human splenic tissue exhibits dose-dependent β2-adrenergic immune modulation. No published trials were identified evaluating chronic, implantable splenic neuromodulation in patients with RA.

Added value of this study

The ConsideRAte study is, to our knowledge, the first randomised, sham-controlled trial to evaluate chronic splenic neuromodulation in patients with moderate-to-severe, difficult-to-treat rheumatoid arthritis who had an inadequate response to at least two biologic or targeted synthetic DMARDs (b/tsDMARDs). These findings demonstrate that surgical implantation is feasible and safe and that chronic delivery of therapy is safe and well tolerated. Patients with difficult-to-treat rheumatoid arthritis who received six daily 5-min sessions of low-frequency electrical stimulation of the splenic neurovascular bundle experienced reduced disease activity scores, improved patient-reported outcomes, and, in exploratory analyses, decreased inflammatory biomarkers. Moreover, greater clinical improvements were observed during the adjunct phase, when stimulation was combined with a b/tsDMARD, resulting in several participants achieving low disease activity or remission.

Implications of all the available evidence

Evidence from two cervical vagus nerve stimulation with an implantable device in RA has established that autonomic neuromodulation can meaningfully modulate inflammatory pathways in humans resulting in reduced disease activity. The present findings extend this concept by targeting the splenic nerve, which is the downstream, organ specific nerve that exerts the modulatory effect on the circulating immune cells, thereby allowing selective engagement of the immunomodulatory neural circuit without being limited by off-target side effects inherent to cervical vagus nerve stimulation.

Together with preclinical and acute human studies showing that splenic nerve activation suppresses pro-inflammatory cytokine production, these data provide the first clinical evidence that chronic splenic neuromodulation is technically feasible, well-tolerated, and associated with improvements in disease activity and patient-reported outcomes. If confirmed in larger, controlled trials, splenic neuromodulation, alone or combined with a b/tsDMARD, could represent a new treatment modality for patients with difficult-to-treat rheumatoid arthritis who remain without adequate treatment despite current advanced therapies.

Introduction

Rheumatoid arthritis (RA) is a systemic autoimmune disease associated with a chronic inflammatory process, which causes damage to joint together with multiple coexisting conditions, and extra-articular manifestations.¹ The introduction of biologic and targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs) have improved outcomes, however up to 30–40% of patients with RA respond inadequately to sequential b/tsDMARD therapy, and 15–20% remain persistent refractory.² Novel therapeutic approaches for this difficult-to-treat rheumatoid arthritis subgroup represents a major unmet need.

The nervous system is recognised as a novel and accessible target in the regulation of inflammation.³ Neuromodulation uses electrical impulses to modify nerve activity and influence physiological processes, with downstream effects on inflammatory pathways and immune responses. Early clinical studies of vagus nerve stimulation (VNS) demonstrated that activation of the parasympathetic nervous system reduced systemic inflammation and improved clinical outcomes in RA and Crohn's disease.4, 5, 6, 7 Recent FDA approval of vagus nerve stimulation for the treatment of RA based on a pivotal trial executed in patients with moderate to severe RA, has further demonstrated the feasibility of a neuromodulation approach for the treatment of RA.⁸ However, stimulation of the cervical vagus nerve, which contains both afferent and efferent fibres, is known to produce off-target effects such as bradycardia, cough, and voice alteration which may limit nerve activation.7, 8, 9, 10

These off-targets effects have led to the exploration of splenic nerve stimulation (SpNS) as a targeted approach, closer to the end organ, that modulates sympathetic efferents within the splenic neuro-vascular bundle (NVB), directly modulating immune cell activity, while avoiding parasympathetic side effects. Preclinical studies showed that SpNS suppresses systemic inflammation across multiple disease models. In rodents, SpNS reduced circulating cytokines and arthritis severity.^11,12 In pigs, a decrease in pro-inflammatory cytokines and monocytes was observed while specialised pro-resolving mediators such as resolvins and protectins were increased.¹³ Importantly, these effects occur without broad immunosuppression, preserving immune competence.¹⁴ In humans, acute intraoperative stimulation of the splenic nerve during minimally invasive esophagectomy was shown to be feasible, safe, and associated with reduced postoperative C-Reactive Protein (CRP) concentrations.¹⁵ Collectively, these findings established SpNS as a viable and mechanistically distinct immunomodulatory approach compared to currently available options.

Building on this foundation, Galvani Bioelectronics developed a laparoscopically implantable neurostimulation system designed to deliver controlled electrical pulses to the splenic NVB. The device aims to achieve sustained modulation of immune activity via local noradrenaline release in the spleen with down-stream immunomodulatory effects, extending beyond inflammatory cytokine reduction and without direct immunosuppressive effects, priming the immune system toward a less inflammatory and more pro-resolving phenotype.

The ConsideRAte trial (NCT05003310) was therefore designed as a randomised, sham-controlled clinical trial to evaluate the safety, tolerability, and preliminary effectiveness of chronic SpNS in adults with moderate-to-severe difficult-to-treat rheumatoid arthritis. The study evaluated both the surgical safety and feasibility of implantation and the clinical effects of stimulation over 12- and 24- weeks periods as mono- and adjunct-therapy.

Methods

Ethics statement

The study followed the Declaration of Helsinki, ISO 14155:2020 Good Clinical Practice, and relevant United States (US) and European Union (EU) regulatory standards. Independent ethics committees approved the protocol. Advarra Institutional Review Board, US (MOD01078633); New York Langone Health (i20-00370_MOD04), Medical Research Ethics Committee (MREC) Amsterdam University Medical Center (UMC), Netherlands; NL-007746. Written informed consent was obtained from all participants before any study procedure. The trial was monitored by an independent data safety monitoring committee (DSMB) and was prospectively registered with ClinicalTrials.gov (NCT05003310).

Study design and participants

ConsideRAte was an early-feasibility, multicentre, randomised, sham-controlled clinical study evaluating the safety, tolerability, and biological activity of chronic splenic neuromodulation in adults with difficult-to-treat rheumatoid arthritis. The study included a screening phase, device implantation, a surgical recovery period, and four predefined treatment periods (Periods 1–4). Enrolment followed a staggered design. The first five participants comprised an open-label sentinel cohort and received active stimulation to allow procedural evaluation prior to the randomised phase. The subsequent six participants were randomised using stratified randomization.

Period 1 (Weeks 0–12; day 1–day 84)

Participants who had received the implantable system were randomly assigned 1:1 to active stimulation (STIM ON) or sham stimulation (STIM OFF) for 12 weeks. This part of the study is called the “monotherapy phase”. During this phase participants could receive a stable dose of conventional synthetic DMARDs (csDMARDs) and/or low-dose corticosteroids (≤10 mg/day).

Period 2 (Weeks 12–24)

After completing Period 1, all participants entered an open-label phase. Participants who had responded to active stimulation continued with stimulation alone. Participants who had received sham stimulation or who were classified as non-responders to stimulation initiated baricitinib monotherapy without stimulation for 12 weeks, according to the protocol.

Period 3 (Treat-to-target; up to 24 weeks)

Participants who demonstrated a response to stimulation in Period 2 but continued to exhibit signs or symptoms of RA entered Period 3, during which they received combination (adjunct) therapy consisting of stimulation combined with baricitinib.

Period 4 (Long-term follow-up; up to 5 years)

Participants who completed Period 3, or who escaped early from Period 1, 2, or 3 progressed to Period 4, the long-term safety follow-up period. Period 4 also functioned as the early-escape pathway from Periods 1–3 for participants requiring additional treatment. During this phase, participants could receive stimulation alone or in combination with other approved b/tsDMARDs at the discretion of the treating rheumatologist.

The full list of participating rheumatology and surgical sites and enrolment numbers is provided in Supplementary Tables S1 and S2. The complete study design, assessments, and detailed definitions for each period are included in Supplementary S2 (Protocol).

Eligibility

Adults aged 22–75 years with moderate-to-severe difficult-to-treat rheumatoid arthritis (defined as DAS28-CRP>5.1 or CRP≥ 5.0 mg/L) and inadequate response or intolerance to at least two b/tsDMARDs with different modes of action,² including one tumor necrosis factor (TNF) inhibitor were eligible. Prior use or not eligible for baricitinib use was an exclusion criteria due to protocol design including use of baricitinib as predefined rescue medication in Period 2 and 3. Other exclusion criteria included serious infection within three months, abdominal wall thickness>2.5 cm, oral corticosteroid dose >10 mg/day (prednisone or equivalent), or significant cardiovascular or surgical risk.

Study procedures and periods

Screening, implantation and activation

Informed consent was obtained from participants prior to initiation of any study-related procedures or study-specific data collection. All inclusion and exclusion criteria were verified during visits with the rheumatologist and implanting surgeon. If all eligibility criteria were met, patients proceeded to implantation of the investigational Galvani System which was performed laparoscopically under general anaesthesia on Day −28 (Galvani Bioelectronics, Stevenage, UK; Supplementary Figure S1). The system was designed to allow for implantation entirely within the abdominal cavity. The system was introduced into the abdominal cavity via laparoscopic trocars with the Implantable Pulse Generator (IPG) secured to the anterior abdominal wall within the pre-peritoneal fat pad of the falciform ligament (Supplementary Figure S2), and connected with a lead to the neural interface containing the active electrodes placed around the splenic arterial loop to encompass the splenic NVB.^15,16 Stimulation was initiated approximately four weeks post-implant (Day 1) (see Supplementary Figure S3 for timelines). Stimulation parameters were individualised and configured up to 20 mA amplitude, 2 ms pulse width, 10 Hz frequency (0.5 s ON, 4.5 s OFF), as derived from computational and translational studies.^14,15,17 Stimulation comprised six 5-min sessions daily (at 2-h intervals, during waking hours), automatically scheduled by the IPG, with usage and performance logs stored internally. Device function was assessed during follow-up visits via review of IPG log files. The IPG is rechargeable and requires approximately 30 min of charging per week over multiple sessions.

Staggered enrolment

To ensure participant safety, a sentinel cohort of five non-randomized participants (cohort 1) were implanted and followed until 3-months post-surgery at which point safety data (including the participants’ 3-month post-surgery CT angiograms) as well as preliminary efficacy data were submitted to FDA for approval to continue enrolment and implantation of the next cohort of patients up to the 15th implant (cohort 2).

Following a pre-specified interim analysis of safety and effectiveness data from cohorts 1 and 2 the results were submitted to the FDA. This was anticipated to occur once ten to twelve subjects had been implanted, but the timing would be determined based on safety signals observed from ongoing safety reviews by the DSMB. Sufficient data was obtained after patient 11 reached the end of Period 1. Based on these data FDA granted approval to expand enrolment to approximately 28 participants. However, the study was terminated prematurely specifically due to financial constraints faced by the sponsor. Full enrolment would have resulted in a sample size of 28 participants approximately.

Randomisation and masking

During the randomised phase, treatment assignment was masked to participants, investigators, joint assessors, and site staff. Unblinded tasks, such as device programming, were limited to designated personnel recorded in the delegation log. Randomisation (1:1) and treatment allocation were managed through the electronic data-capture system. Randomization was stratified on the basis of whether or not participants experienced a flare during the post-operative period requiring rescue treatment with otherwise prohibited therapies. Within each strata a 1:1 randomisation ratio was applied. An unblinded Field Clinical Engineer (FCE) programmed the devices according to allocation, activating or deactivating stimulation while maintaining identical interfaces across groups while participants were formally blinded to their allocation. To preserve the blind, the programming interface screen was strictly hidden from the participant and site staff during all adjustments. Furthermore, both STIM ON and STIM OFF participants were informed that active therapeutic parameters could frequently operate below the sensory threshold, producing no perceptible physical sensation. Participant-evoked sensations and qualitative comfort logs were captured during follow-up visits to continuously monitor safety and retrospectively evaluate blinding integrity. Unblinding was performed for each participant after completion of all their Week 12 assessments, and emergency unblinding was permitted if clinically required.

Follow-up and dose adjustment

Participants attended post-operative reviews on Days −18 and −7, and follow-up visits in Period 1 were on Days 1, 4, 7, 14, 28, 42, and 56 for potential dose escalation to their personal maximum tolerated dose (not exceeding 20 mA). Thereafter participants were followed up until Week 12. Supplementary Figure S3 shows schematic presentations of visits in Period 1. Sham participants underwent identical programming without current delivery. In Period 2, 3, and 4, participants on active stimulation (STIM ON) were monitored via in-clinic visits every three months and those with the device in STIM OFF every six months. For participants (re-) starting stimulation in Period 3 or 4, additional visits in line with the ramp-up schedule in Period 1 were scheduled. After the first two years in Period 4 participants were monitored every six months up to five-year follow-up. In Period 2, 3 and 4 participants could also undergo a ramp-up or -down dose adaptation at scheduled or unscheduled visits as required. The asymmetrical follow-up schedule between STIM ON and STIM OFF was designed to accommodate stimulation parameter optimisation, log-file reviews, and device safety and compliance for active neurostimulation, while minimizing the protocol burden on participants receiving standard of care without active stimulation.

Concomitant therapy

Concomitant b/tsDMARDs were not permitted during the 12-week randomised monotherapy phase. Stable background conventional synthetic DMARDs and/or low-dose corticosteroids (≤10 mg/day) were allowed in Period 1 if unchanged for ≥4 weeks before enrolment. During the open-label long-term follow-up phase, adjunct b/tsDMARDs were permitted at the investigator's discretion and were combined with SpNS.

TruCulture lipopolysaccharide (LPS)-induced cytokine assay

Whole blood was collected into TruCulture tubes (Rules-Based Medicine, Austin, TX) prefilled with media and lipopolysaccharide, incubated at 37 °C for ≈22 h, filtered, and frozen on site. Cytokines were quantified centrally using the OptiMAP multiplex immunoassay (Rules-Based Medicine/IQVIA), measuring Interferon (IFN)-α, IFN-γ, Interleukin (IL)-10, IL-12, IL-17A, IL-1β, IL-6, IL-8, and TNFα.

Serum cytokine and chemokine assays

For serum analysis, samples were processed in separator tubes, centrifuged at 1500 g for 15 min, and stored at −70 °C. Cytokine and chemokine concentrations were measured using Meso Scale Discovery (MSD) 10-plex and chemokine panels (Q2 Solutions/IQVIA).

Flow cytometry of cell populations

Whole blood was collected into Cyto-Chex tubes and was completely and gently inverted 10 times. Tubes were maintained at room temperature and analysed within 48 h via the Tcell Subset panel (Q2 Solutions/IQVIA).

Outcomes

The primary outcome was the safety and tolerability of device implantation and chronic splenic neuromodulation, including the incidence of procedure- or device-related adverse events and unanticipated adverse device effects. Adverse events were collected throughout the study and coded using medical dictionary for regulatory activities (MedDRAversion 26.1). Severity and relationship to the study device were evaluated by site investigators, with independent oversight by the study Data and Safety Monitoring Board (DSMB). Safety was also assessed through the collection of abnormalities in safety laboratory results, and any potential treatment-induced changes in 12-lead electrocardiogram (ECG) assessed during in-clinic visits. The key secondary outcome was change from baseline in DAS28-CRP at Week 12 (end of the monotherapy phase, Week 12). Additional secondary outcomes included changes in Clinical Disease Activity Index (CDAI), Simplified Disease Activity Index (SDAI), Health Assessment Questionnaire Disability Index (HAQ-DI), 36-Item Short Form Health Survey (SF-36; scores range from 0 to 100, with higher scores indicating greater well-being), and other patient-reported outcomes (pain, fatigue, and global assessment). Exploratory outcomes included the American College of Rheumatology (ACR) 20/50/70 response rate and changes in LPS-induced whole blood- and circulating cytokines. The Minimal Clinically Important Difference (MCID) on an individual level for DAS28-CRP, SDAI, CDAI, HAQ-DI, SF-36 outcomes were defined as −1.0, −13, −12, −0.375 and 5,18, 19, 20 respectively. The MCID on the group level for DAS28-CRP and HAQ-DI were defined as −0.6 and >0.15,¹⁹ respectively.

Sample size, estimand and statistical analysis

Sample size

The study was an early feasibility study and not powered to assess any endpoints.

Sample size consideration was based on the precision for the treatment difference estimate of the key secondary efficacy endpoint as measured by DAS28 (CRP) at Period 1 of the proposed design. From approximately 28 implanted participants in total (including the first 5 implanted participants who receive Open Label active stimulation and about 23 randomized participants), approximately 10 randomized participants per arm were anticipated to complete Period 1 allowing for an approximate 15% drop-out rate.

Assuming an estimate of the between participant standard deviation of 1.2 points for DAS28 (CRP) and using the t-distribution, ten (10) participants per arm completing Period 1 provides a precision estimate of 1.12 for the half width of the 95% confidence interval (which is analogous to 95% credible intervals assuming non-informative priors for all modelling parameters) for the difference between treatment arms. This ensures that the lower and upper bounds of the 95% credible intervals for the difference between Randomized Stim ON v Randomized Stim OFF will be within ±1.12 points of the estimated difference. Hence, if the observed treatment effect is greater than 1.12 then the study will yield a credible interval that excludes 0.

Due to sponsor financial constraints the study was terminated prematurely and therefore the full planned enrolment was not achieved.

Estimand strategy

For efficacy endpoints, the hypothetical estimand strategy which estimates the treatment effect in the absence of intercurrent events (IE) was used. This is to account for the possibility that a high proportion of participants receiving sham stimulation may need escape therapy with prohibited medications which may improve their outcome and reduce the treatment effect when compared to active stimulation. Under this estimand strategy, data collected after occurrence of an IE until the participant either completes Period 1, escapes from Period 1 to another period or withdraws from the study before Period 1 completion will not be used in the analysis but set to missing. Data set to missing or unobserved due to escape from Period 1 is imputed as if participants remain on their assigned treatment. This relies on missing at random (MAR) assumption. No explicit imputation was performed instead the mixed model for repeated measures (MMRM) was used which accounts for the imputation of missing observations as part of the estimation process assuming MAR.

Statistical analysis

All implanted participants with at least one post-baseline assessment were included in the efficacy analyses according to randomised assignment (intent-to-treat population). Least-squares adjusted mean, mean difference and corresponding 95% confidence intervals were estimated using a mixed model for repeated measures (MMRM) including endpoint baseline value, treatment, visit, and treatment-by-visit interaction as fixed effects. Due to limited data, in the sham arm, resulting in statistical model convergence issues, a modification to the planned analysis was conducted for the monotherapy results. Here, results of analyses used last observation carried forward (LOCF) instead of an MAR assumption to account for data affected by IE for the 12-week randomised monotherapy phase. The hypothetical estimand strategy results, with a MAR assumption for missing data, were used in the open-label adjunct therapy setting. The minimal clinically important differences (MCIDs) were prespecified as a ≥1.0-point reduction in DAS28-CRP, ≥13 in SDAI, ≥12 in CDAI, and a ≥0.35-point decrease in HAQ-DI.¹⁹ Biomarker data was summarised and listed as appropriate with no formal statistical analyses performed due to samples only being available for a proportion of patients. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Deviation from methods specified in protocol

The planned primary analysis and reporting strategy involved a Bayesian framework; however, due to a premature study termination resulting in a limited participant cohort and subsequent convergence issues, a frequentist analytical approach was ultimately employed. Consequently, the statistical models incorporated a restricted number of covariates, deviating from the originally intended comprehensive adjustment for factors such as treatment, baseline values, visit, treatment-by-visit interactions, and other pre-specified covariates, including the stratification variable. Furthermore, a notable deviation from the protocol-specified analysis plan was the adoption of Last Observation Carried Forward (LOCF) for handling missing data influenced by intercurrent events (IEs) during the 12-week randomized monotherapy phase, as opposed to the hypothetical estimand strategy predicated on a Missing At Random (MAR) assumption.

Role of the funding source

The study was sponsored by Galvani Bioelectronics, which designed the trial and oversaw data collection and analysis. FCE activities were performed by sponsor employees. Academic and sponsor-affiliated authors contributed to data interpretation, manuscript preparation, and submission.

Results

Results presented were from the interim analysis summarising data of 11 participants enrolled in cohort 1 and 2 prior to the premature termination of the study. Participants were enrolled and implanted between January 2022 and November 2024. Database lock for data analysis was performed on October 31st, 2025.

Participants

Patients were screened for eligibility, and 11 participants were enrolled across six centres in the US and Netherlands (Fig. 1; Supplementary Tables S1 and S2). The primary reasons for screening failures included abdominal wall thickness exceeding the protocol-specified limits on ultrasound and CT-angiogram due to limitation of the charger to charge at depths greater than 2.5 cm, elevated body mass index (BMI), failure to meet the predefined active disease criteria at the time of the screening, a history of relevant comorbidities and surgical exclusion criteria ((e.g., hiatal hernia leading to distortion of local anatomy). The first five participants comprised the open-label sentinel cohort, followed by six participants randomised 1:1 to active (STIM ON, n = 3) or sham (STIM OFF, n = 3) stimulation for 12 weeks. Overall, in 12 weeks of Period 1, eight participants received active stimulation and three received sham stimulation. In subsequent periods, participants either continued on stimulation, on baricitinib only or moved to adjunct therapy.

Fig. 1.

Flow chart ConsideRAte study.

Baseline characteristics

Baseline demographics and disease characteristics are summarised in Table 1 (Supplementary Table S3 for participants in adjunct). Participants had a mean age of 51.6 ± 10.6 years, and seven (63.6%) were female. The mean duration of RA since diagnosis was 18.2 ± 11.0 years. Per inclusion criteria, all participants had an inadequate response to at least two b/tsDMARDs before enrolment, with a mean of 3.3 ± 1.2 prior therapies. All participants (100%) failed or had an inadequate response to at least one bDMARD, and eight (72.7%) had also failed or had an inadequate response to a tsDMARD. All patients classified as difficult-to-treat rheumatoid arthritis.²

Table 1.

Baseline demographics and disease characteristics.

	Open-label sentinel Stim ON (N = 5)	Randomised Stim ON (N = 3)	Randomised Stim OFF (N = 3)	All Stim ON (N = 8)	ConsideRAte total (N = 11)
Age (years)	52.4 ± 9.13 (5)	48.7 ± 12.50 (3)	53.3 ± 14.74 (3)	51.0 ± 9.80 (8)	51.6 ± 10.58 (11)
Sex (Female)	2 (40.0%)	2 (66.7%)	3 (100.0%)	4 (50.0%)	7 (63.6%)
Body Mass Index (BMI) -Screening	25.9 ± 1.19 (5)	26.8 ± 3.80 (3)	22.5 ± 2.71 (3)	26.2 ± 2.27 (8)	25.2 ± 2.85 (11)
Ethnicity
Asian	0 (0.0%)	0 (0.0%)	1 (33.3%)	0 (0.0%)	1 (9.1%)
Black or African American	1 (20.0%)	0 (0.0%)	0 (0.0%)	1 (12.5%)	1 (9.1%)
White	4 (80.0%)	3 (100.0%)	2 (66.7%)	7 (87.5%)	9 (81.8%)
Rheumatoid Arthritis (RA) Duration since diagnosis (years)	13.8 ± 3.96 (5)	17.3 ± 11.59 (3)	26.3 ± 17.01 (3)	15.1 ± 7.12 (8)	18.2 ± 10.99 (11)
High sensitivity (hs) C-Reactive Protein (CRP),mg/mL	52.2 ± 39.40 (5)	18.3 ± 5.83 (3)	13.9 ± 13.94 (3)	39.5 ± 34.70 (8)	32.5 ± 32.02 (11)
Rheumatoid factor (RF)
Positive	4 (80.0%)	2 (66.7%)	3 (100.0%)	6 (75.0%)	9 (81.8%)
Negative	1 (20.0%)	1 (33.3%)	0 (0.0%)	2 (25.0%)	2 (18.2%)
Anti-cyclic citrullinated peptide (anti-CCP)
Positive	5 (100.0%)	3 (100.0%)	2 (66.7%)	8 (100.0%)	10 (90.9%)
Negative	0 (0.0%)	0 (0.0%)	1 (33.3%)	0 (0.0%)	1 (9.1%)
Number of prior biologic or targeted synthetic Disease modifying anti-rheumatic drugs (b/tsDMARDs)	3.2 ± 1.64 (5)	3.0 ± 0.00 (3)	3.7 ± 1.15 (3)	3.1 ± 1.25 (8)	3.3 ± 1.19 (11)
Number of prior b/tsDMARDs
2	2 (40.0%)	0 (0.0%)	0 (0.0%)	2 (25.0%)	2 (18.2%)
3	2 (40.0%)	3 (100.0%)	2 (66.7%)	5 (62.5%)	7 (63.6%)
>3	1 (20.0%)	0 (0.0%)	1 (33.3%)	1 (12.5%)	2 (18.2%)
Prior b/tsDMARDs
Anti-IL-6 agents	1 (20.0%)	0 (0.0%)	0 (0.0%)	1 (12.5%)	1 (9.1%)
Anti-TNF agents	5 (100.0%)	3 (100.0%)	3 (100.0%)	8 (100.0%)	11 (100.0%)
B-cell depleting agents	2 (40.0%)	0 (0.0%)	0 (0.0%)	2 (25.0%)	2 (18.2%)
T-cell modulating agents	2 (40.0%)	1 (33.3%)	1 (33.3%)	3 (37.5%)	4 (36.4%)
Janus kinase inhibitor (JAKi)	3 (60.0%)	3 (100.0%)	2 (66.7%)	6 (75.0%)	8 (72.7%)
Disease activity score using 28 joints (DAS28- CRP)a	6.2 ± 0.74 (5)	5.8 ± 0.50 (3)	5.5 ± 1.01 (3)	6.0 ± 0.65 (8)	5.9 ± 0.75 (11)
Clinical Disease Activity Index (CDAI)a	41.6 ± 15.91 (5)	39.0 ± 6.62 (3)	39.3 ± 23.84 (3)	40.6 ± 12.60 (8)	40.3 ± 15.01 (11)
Health Assessment Questionnaire- Disability Index (HAQ-DI)a	1.9 ± 0.58 (5)	1.6 ± 0.38 (3)	1.7 ± 0.63 (3)	1.8 ± 0.50 (8)	1.7 ± 0.50 (11)

Note: Continuous data displayed as mean ± Standard Deviation [SD] (n); Categorical data displayed as n (%), the number and percentage of participants with that particular response.

^aCollected on Day 1, prior to stimulation.

At study entry, participants had high disease activity, with a mean DAS28-CRP of 5.9 ± 0.7, a mean tender joint count (TJC28) of 13.1 ± 7.4, and a mean swollen joint count (SJC28) of 13.1 ± 7.8. The mean HAQ-DI score was 1.7 ± 0.5, consistent with moderate-to-severe functional impairment. The mean high-sensitivity C-reactive protein (hsCRP) concentration was 32.5 ± 32.0 mg/L.

Treatment exposure

All participants completed in-clinic stimulation ramp-up over a period of multiple visits; Supplementary Figure S3), gradually increasing to their individual maximum tolerated amplitude, which ranged from 3 mA to 9 mA. From Day 1, participants received six active (STIM ON) or six sham (STIM OFF) stimulation sessions daily. Stimulation-evoked sensations were recorded during ramp-up visits. Across both groups, 154 stimulations were delivered, most sensations were reported as internally localised (66.9%, 103), mainly in the upper (56%, 86) and lower left abdomen (22.9%, 36), while 27.1% (43) of sessions were associated with no perceptible sensation. Reported discomfort was rated as none (40.4%, 63), mild (36.1%, 54), moderate (16.9%, 27), or severe (1.8%, 3). If severe discomfort occurred, the stimulation amplitude was reduced to a tolerable level. No stimulation events were associated with clinically significant changes in vital signs, including blood pressure or ECG. Stimulation-evoked sensations by the STIM ON AND STIM OFF group are reported in Supplementary Table S4.

In the STIM ON group during Period 1, one participant escaped due to non-compliance with charging requirements and missing over 20% of scheduled therapy, and another due to intolerable RA symptoms requiring medication prohibited by the protocol. In the STIM OFF group, two of three participants escaped early due to worsening RA symptoms. To date, participants have received Galvani therapy for approximately 14 patient-years (Supplementary Table S5).

Out of 11 participants, nine participants entered Period 2: four in the STIM ON-only group and five in the STIM OFF baricitinib-only group. Two participants moved directly from Period 1 to Period 4 based on investigators’ assessment that timely relief was unlikely. Of the four participants in the STIM ON group, three participants completed the 12-week Period 2, while one transitioned early to Period 4 due to a disease flare. Among those who completed Period 2, two progressed to Period 3, and one advanced to Period 4 because of baricitinib intolerance. Of the five participants in the STIM OFF group, two participants exited Period 2 early due to adverse events or flares. Of the three remaining participants who completed Period 2, two showed a DAS28 improvement of less than 1.2 and were moved to Period 4, while one progressed to Period 3 to receive combination therapy with baricitinib.

Three participants entered Period 3; two escaped early to Period 4. One met the predefined criteria for missed therapy days triggering an early exit, and the other escaped per investigator decision as the participant did not notice any improvement on the combination”.

In Period 4, two patients continued monotherapy and nine patients on adjunct therapy (see Supplementary Table S6 for participant accountability in Period 1, 2 and 3).

Implant and treatment safety

All implantations were completed successfully, and postoperative recovery was uneventful without any complications, Serious Adverse Events (SAEs), or Unanticipated Adverse Device Effects (UADEs). The treatment was started approximately 28 days after implantation (Day 1) and remained functional throughout the study. Computed Tomography Angiogram (CTA) images taken approximately between three- and four-months post implant showed no evidence of device migration, with the system maintaining its original intended position, and no intra-abdominal pathology caused by the surgical implant procedure, the system or stimulation was observed. There were no infections or wound healing complications.

Across all study periods, 42 AEs were reported among 11 participants of those AEs, 91% were unrelated to device or stimulation (Table 2). Four participants experienced an AE possibly or probably related to the device or implant procedure. These AEs included implant-site pain, dyspnoea, procedural nausea, defecation urgency, hunger, abdominal discomfort, and mild hyperhidrosis. All device-related, or study procedure related AEs resolved spontaneously without intervention. No deaths, serious AEs, or serious device effects occurred, and no participant discontinued due to an AE.

Table 2.

Adverse events in all Periods.

	Considerate (N = 11)
	Total events	Total participantsa
Unrelated, Adverse Events (AEs)b	31	10 (90.9%)
Related, Adverse Events (AEs)c	11	4 (36.4%)
Study Device AEsd	6	3 (27.3%)
Abdominal pain lower	1	1 (9.1%)
Defecation urgency	1	1 (9.1%)
Discomfort	1	1 (9.1%)
Hunger	1	1 (9.1%)
Hyperhidrosis	1	1 (9.1%)
Procedural nausea	1	1 (9.1%)
Adverse Device Effects (ADEs)e	5	3 (27.3%)
Dyspnoea	1	1 (9.1%)
Implant site pain	1	1 (9.1%)
Medical device site pain	3	1 (9.1%)
Study Procedure, Related AEsc	2	2 (18.2%)
Abdominal pain lower	1	1 (9.1%)
Suture related complication	1	1 (9.1%)
Serious Adverse Events (SAEs)	0	0 (0.0%)
Serious Adverse Device Effects (SADEs)	0	0 (0.0%)
Death	0	0 (0.0%)
AEs Leading to Study Discontinuation	0	0 (0.0%)

Includes events occurring during the surgery/recovery period and throughout follow-up periods.

Event types listed are based on Medical Dictionary for Regulatory Activities (MedDRA) Version 24.0 preferred terms.

^aData presented as number and percentage of participants with at least one event of that type.

^bUnrelated AEs include the events categorized as not related or unlikely to be related to the device.

^cRelated AEs include the events categorized as possibly, probably, or causally related. Rheumatoid Arthritis and study procedure relatedness determined by investigator only.

^dIncludes adverse events categorized as possibly related to the device.

^eIncludes adverse events categorized as probably or causally related to the device.

Efficacy during monotherapy (Period 1 and 2)

During the 12-week monotherapy phase (Period 1), participants receiving active splenic nerve stimulation (STIM ON, n = 8) showed numerical trends towards improvement across disease activity measures and patient-reported outcomes (PROs) compared with sham stimulation (STIM OFF, n = 3). Outcomes included DAS28-CRP, CDAI, SDAI, HAQ-DI, and SF-36. The time course of DAS28-CRP, CDAI, HAQ-DI and SDAI changes in Period 1 by treatment group is shown in Fig. 2a–d (Supplementary Figures S4 and S5).

Fig. 2.

Secondary and exploratory efficacy outcome measures for monotherapy overtime for 12-weeks (Period 1). (a) Disease Activity Score using 28 joints using C-reactive protein (DAS28-CRP), (b) Health Assessment Questionnaire-Disability Index (HAQ-DI), (c) Simplified Disease Activity Index (SDAI), (d) Clinical Disease Activity Index (CDAI) (n = 8, STIM ON; n = 3 STIM OFF). Adjusted data: Last Observation Carried Forward (LOCF) was used for missing data imputation.

By week 12, DAS28-CRP improved more in the STIM ON group than in the STIM OFF group (mean ± SEM, −1.02 ± 0.45 vs −0.34 ± 0.14) resulting in a between-group difference of −0.68 ± 0.48 (Fig. 2a; Table 3). A clinically meaningful response in DAS28-CRP (MCID ≥1.0-point reduction) (Ward et al., 2015) was observed in 62.5% of STIM ON participants; no MCID responders were observed in the sham group. The between group difference change for the control and treatment group exceeded the MCID, with greater improvement in those receiving active stimulation.²¹ Consistent results were seen for CDAI and SDAI, with greater reductions in the STIM ON group compared with STIM OFF and both indices in the STIM ON group exceeded their MCIDs from baseline (Table 4).

Table 3.

Secondary and exploratory efficacy outcome measures for monotherapy at Week 12 (end Period 1)-Endpoints DAS28-CRP, CDAI, SDAI, HAQ-DI, SF-36 Mental Component Score (MCS), and Physical Component Score (PCS).

Analysis visit	P1 all Stim ON (N = 8)	P1 randomized Stim OFF (N = 3)	Difference: P1 all Stim ON – P1 randomised Stim OFF
DAS28CRP	−1.02 ± 0.45∗	−0.34 ± 0.14	−0.68 ± 0.48∗
SDAI	−13.66 ± 5.81∗	−5.83 ± 3.11	−7.83 ± 6.60
CDAI	−12.21 ± 5.64∗	−5.66 ± 3.27	−6.55 ± 6.53
HAQDI	−0.45 ± 0.21∗	−0.29 ± 0.18	−0.16 ± 0.28∗
SF36 MCS	10.08 ± 4.87∗	1.66 ± 0.42	8.42 ± 4.89
SF36 PCS	4.42 ± 2.79	3.11 ± 3.22	1.31 ± 4.35

Adjusted data: LOCF was used for missing data imputation.

Note 1: Data displayed as adjusted mean difference ± SEM; LOCF (Last Observation Carried Forward) was used for missing data imputation.

Note 2: Results based on mixed model repeated measures analysis adjusted for treatment, baseline value, visit, and treatment by visit, with a compound symmetry covariance structure.

Note 3: Values marked with ∗ exceed minimal clinically important difference (MCID).

Table 4.

Key Secondary and exploratory outcome measures in patients on mono- and adjunct therapy.

Change from baseline (Adjusted Mean ± SEM)				MCID (individual/group level)
Measure	12-week Mono (n = 9)	12-week adjunct (n = 9)	24-week adjunct (n = 9)
DAS28-CRP	−1.09 ± 0.54	−2.99 ± 0.51	−2.27 ± 0.57	−1.0/−0.6
SDAI	−15.11 ± 6.56	−33.51 ± 6.31	−24.84 ± 6.88	−13/NA
CDAI	−14.26 ± 6.38	−30.36 ± 6.12	−21.35 ± 6.72	−12/NA
HAQ-DI	−0.47 ± 0.21	−0.68 ± 0.19	−0.80 ± 0.22	−0.375/>0.15
MCS (SF-36)	9.86 ± 4.08	9.10 ± 3.86	13.73 ± 4.33	5/NA
PCS (SF-36)	4.19 ± 3.20	11.12 ± 3.09	12.12 ± 3.37	5/NA
Bodily Pain (SF-36)	9.23 ± 3.15	14.86 ± 3.43	14.71 ± 3.77	5/NA
Pain (VAS)	−23.89 ± 10.55	−47.42 ± 10.08	−33.99 ± 11.12	−20/NA

Adjunct data at 12- and 24-weeks compared to 12-week monotherapy data–Endpoints DAS28-CRP, CDAI, SDAI, HAQ-DI, SF-36 Mental Component Score (MCS), Physical Component Score (PCS), and Patient VAS (pain).

Note: Change from baseline data displayed as adjusted mean ± SEM. Results based on mixed model repeated measures analysis with a compound symmetry covariance structure, adjusted for baseline value and visit (Hypothetical Strategy Estimand-Based Analysis).

Changes in PROs at week 12 supported these findings, with mean (±SEM) improvements in HAQ-DI (−0.45 ± 0.21 vs −0.29 ± 0.18) and SF-36 scores (MCS: 10.08 ± 4.87 vs 1.66 ± 0.42; PCS: 4.42 ± 2.79 vs 3.11 ± 3.22) for STIM ON vs STIM OFF, respectively (Table 3). Among participants (n = 3) who continued monotherapy through week 24 in Period 2, the clinical benefits of splenic neuromodulation were sustained across all parameters (Supplementary Table S7).

Adjunct therapy (Periods 3–4)

Nine participants entered open-label adjunct therapy (two continued on monotherapy), during which SpNS was combined with a b/tsDMARD (adjunct therapy) prescribed at the investigator's discretion. A post-hoc analysis assessed efficacy at 12 and 24 weeks after adjunct initiation. Adjusted mean (±SEM) changes factoring in the occurrence of an intercurrent event of high dose corticosteroids use in DAS28-CRP from baseline were −2.82 ± 0.60 at Week 12 and –2.28 ± 0.70 at Week 24 (Table 4, Supplementary Table S8). For comparison, the adjusted mean change after 12 weeks of monotherapy was −1.09 ± 0.54. Pain scores were also improved with larger reductions at week 12 with an adjusted mean change of −47.42 ± 10.08 and −33.99 ± 11.12 at Week 24.

Consistent improvements were seen across other disease activity measures. For CDAI, adjusted mean changes were −30.53 ± 7.34 at Week 12 and −22.86 ± 8.03 at Week 24; for SDAI, changes were −33.23 ± 7.44 and −26.10 ± 8.17, respectively. Functional improvement was reflected by reductions in HAQ-DI of −0.68 ± 0.18 at Week 12 and −0.82 ± 0.21 at Week 24 exceeding the MCID of 0.35 (Ward et al., 2015) (Table 4).

Fig. 3a–d shows the time course of key efficacy measures for 12 weeks of monotherapy followed by 12 and 24 weeks of adjunct therapy in the same participants, illustrating greater and sustained improvements with adjunct treatment (Supplementary Figure S6).

Fig. 3.

Clinical response over time for patients on mono- and adjunct therapy. Change from baseline at each time point in Period 1 of monotherapy, and 12 and 24 weeks of adjunct therapy (N = 9), (a) DAS28-CRP, (b) HAQ-DI, (c) SDAI, (d) CDAI. Plots show adjusted mean ± SEM from Mixed Model Repeated Measures (MMRM) models including baseline value and visit as covariates.

In addition, the change in SF-36 Physical Component Score (PCS), Mental Component Score (MCS), and Pain for 12-weeks mono- and 12 -and 24-weeks adjunct therapy are shown in Table 4. Mean PCS improvements with adjunct therapy were 11.12 ± 3.09 at Week 12 and 12.12 ± 3.37 at Week 24, and MCS improvements were 9.10 ± 3.86 and 13.73 ± 4.33 and was sustained or improved further after 24 weeks. The SF-36 Bodily Pain domain also showed improvements with adjunct therapy, consistent with the reductions observed in patient-reported pain scores.

Clinical response rates

In exploratory analyses, clinical response rates for ACR20/50/70, based on the 28 joint count, and DAS28-CRP thresholds are summarised in Fig. 4. During the 12-week monotherapy period, 44.4% (4/9) achieved ACR20, 22.2% (2/9) achieved ACR50, and 11.1% (1/9) achieved ACR70, with none reaching low disease activity (DAS28-CRP < 3.2) or remission (DAS28-CRP < 2.6).

Fig. 4.

Exploratory endpoint results for American College of Rheumatology (ACR), Low Disease Activity (LDA) and remission response rates. (a) ACR20/50/70 response rates (n = 9), (b) LDA and remission rates (n = 9). Data presented are percentages of patients at 12-weeks monotherapy vs 12- and 24-weeks adjunct therapy for each score indicated.

Following the addition of a bDMARD or tsDMARD, response rates improved markedly. After 12 weeks of adjunct therapy, 77.8% (7/9) achieved ACR20, 44.4% (4/9) achieved ACR50, and 22.2% (2/9) achieved ACR70, while 44.4% (4/9) reached LDA and 22.2% (2/9) achieved remission. These responses were sustained through 24 weeks, with ACR20, ACR50, and ACR70 scores remaining at 77.8%, 44.4%, and 22.2%, respectively, and the proportions achieving LDA and remission increasing to 55.6% (5/9) and 33.3% (3/9).

Biomarker analyses

In exploratory analyses, splenic nerve stimulation was associated with reductions in proinflammatory cytokines measured ex vivo following 24-h LPS stimulation of whole blood. Median LPS-stimulated levels of IL-6, IL-8 and serum CXCL10 were decreased at Week 12 when compared to Baseline (Supplementary Table S9).

Further analysis stratified participants by clinical response, defining responders as those achieving a ≥1.0 decrease in DAS28-CRP at Week 12. Responders demonstrated trends for reductions in LPS-stimulated TNFα, IL-6, IL-1β, and IL-23 from baseline to Week 12, whereas non-responders showed increases in these cytokines (Fig. 5a, Supplementary Table S10). Serum cytokine profiling revealed higher baseline TNFα, IFNγ, and CXCL10 levels in non-responders compared with responders (Fig. 5b, Supplementary Table S11). In addition, flow cytometry analysis of cell populations identified reductions in Th17 cells in responders at Week 12, in particular in the Th17.1 subset (Fig. 5c).

Fig. 5.

Cytokine/chemokine and T-helper 17 (Th17) cell analysis by responder status in patients with rheumatoid arthritis (RA) receiving splenic nerve stimulation as monotherapy. (a) Change in lipopolysaccharide (LPS)-stimulated cytokines from Baseline (BL) to Week 12 between responders and non-responders; responder patients defined as achieving a ≥1.0 decrease in DAS28-CRP at Week 12. Data expressed as Mean ± SEM, n = 3 per grouping. (b) Comparison of monotherapy Stim-On BL serum levels of pro-inflammatory cytokine/chemokines by responder status (responder n = 4, non-responder n = 3). (c) Percentage change from BL to Week 12 in Th17 and Th17.1 cell populations. Left panel: % change in Th17 population as % of CD4+ cells, middle panel: % change in absolute Th17.1 cells, and right panel: Th17.1 population as % of CD4+ cells (responders n = 4, non-responders n = 3). All data is presented as Mean ± SEM).

Discussion

In this study, we present for the first time that chronic SpNS delivered through a laparoscopically implanted neuromodulation device is technically feasible, demonstrates a favourable safety profile, and is well tolerated in patients with difficult-to-treat rheumatoid arthritis. This trial specifically enrolled patients with active disease despite prior treatment and failures with multiple b/tsDMARDs with multiple mechanisms of action, referred to as difficult-to-treat rheumatoid arthritis patients.²

A comprehensive review of safety data collected throughout the study did not identify any significant safety issues associated with the system, its implantation, or the stimulation procedures conducted. All implantations were successfully completed using a laparoscopic technique. There were no complications and postoperative recovery was uneventful. The CTA confirmed stable positioning of the system with no pathology attributed to the procedure, system or therapy being identified. These findings are consistent with the idea that by placing system components laparoscopically and avoiding longer incisions and subcutaneous pockets avoids many of the complications seen with the implanting of other systems using open surgical techniques and facilitates long-term device performance. Adverse events were mostly mild and transient in nature. Importantly, no SAEs or SADEs were reported. For the efficacy outcomes, during the monotherapy phase, participants who had failed multiple prior b/tsDMARDs experienced clinically meaningful improvements with active stimulation, whereas sham-treated participants showed minimal change. SpNS was well tolerated, and participants were ramped up to their individual maximum tolerated amplitude.

Adjunct therapy produced greater improvements than monotherapy with numerically greater clinical responses vs baseline across disease activity measures. Nearly half of participants achieved low disease activity and over a quarter achieved remission by week 12. Seven participants showed sustained responses between 12 and 24 weeks, these patients that had failed 2–3 b/tsDMARDs at screening. Two participants were losing response between Week 12 and 24, both of whom had failed more than 3 b/tsDMARD at the time of screening.

Importantly, patients with inadequate clinical response were managed according to investigator judgement and could transition to rescue therapy or subsequent study periods. During the study, IPG explantation was not required for lack of efficacy alone. In a clinical setting, IPG explanation can be reserved for device-related complications, intolerance, or participant preference”.

The magnitude of DAS28-CRP reductions observed with adjunct therapy compares favourably with responses typically reported following later-line b/tsDMARD switching,²² suggesting that the combination of SpNS and pharmacological therapy may have additive or synergistic effects in this refractory population. As such, SpNS might constitute the first non-immunosuppressive adjunct treatment option for difficult-to-treat rheumatoid arthritis patients on b/tsDMARDs to yield clinically meaningful additive benefits and improved outlook towards low disease or remission. Consequently, such treatment options may become a major contributor to future RA treat-to-target strategies, and the adjunct use warrants further clinical investigation across patient strata.

The improvement in quality-of-life and pain measures in the current study deserves particular attention. Pain is often poorly controlled in RA even when inflammation is adequately managed²³ and available analgesic options have limited effectiveness and may carry significant side-effects.²⁴ The reductions in patient-reported pain and improvements in SF-36 Bodily Pain, with MCS and PCS exceeding the MCID suggest that splenic nerve stimulation may influence symptom pathways beyond inflammation. It should be noted that MCID estimates for SF-36 measures vary across populations and methodologies.18, 19, 20 In addition, the MCS scale had uniformly low responsiveness in studies of anti-rheumatic treatment in RA, however, in the current study the response rate was relatively high. Therefore, it is important to confirm these promising findings in larger studies.

Although biomarkers were analysed in exploratory analyses in a small proportion of the patients there was a clear trend in dampened inflammatory cytokine levels (TNFα, IL-1β, IL-6 and CXCL10) among responders during the monotherapy phase of the study. These findings were consistent with preclinical evidence showing that SpNS modulates systemic inflammation, via adrenergic signalling pathways, resulting in reduction of pro-inflammatory mediators.^13,15,25 The contrasting serum and LPS stimulated cytokine profile in biomarker data observed between responders and non-responders suggests a potential difference in immunological states between these groups. Responders appeared more quiescent or anergic at baseline in a “resting” systemic state, whereas they maintain a primed/hyper-responsive cellular phenotype capable of mounting an enhanced cytokine release under inflammatory conditions.²⁶ Furthermore, there was a reduction in pathogenic Th17 and Th17.1 cells. In responders following SpNS. Overall, these findings suggest that direct SpNS can selectively and effectively achieve neuroimmunomodulatory control of inflammatory responses in humans.

This pilot study also has limitations. As an early feasibility trial, (a) it was not powered for statistical hypothesis testing in monotherapy or the adjunct setting and the small sham group limits comparative inference including the impact of demographic characteristics, (b) while stimulation parameters were safely individualized to each participant's maximum tolerated threshold, this feasibility study was not statistically powered to perform an evaluation of how variations in stimulation parameters impact overall therapeutic efficacy, (c) DAS28-CRP, while a key efficacy endpoint, may be more sensitive to short-term inflammatory changes and less stringent than CDAI or SDAI; however, concordant improvements across multiple composite and patient-reported outcomes support the robustness of the clinical signal, (d) open adjunct-therapy analyses were post-hoc with the longest current follow-up at 24-weeks and without assessment of long-term disease modification, (e) cytokine analyses were exploratory and not corrected for multiplicity (f) in the current study no imaging assessments (e.g., x-ray), were utilised to monitor the effect on structural joint disease or synovitis. Larger, controlled studies are needed to identify predictors of response, determine optimal parameter settings, assess durability of the therapeutic effects, determine integration of stimulation into standard-of-care RA treatments, and monitor stimulation effects on additional immunological parameters and structural joint changes.

Clinical studies of cervical vagus nerve stimulation in RA have shown that neuromodulation can attenuate inflammatory activity and reduce disease activity.4, 5, 6, 7^,27 Since the vagus nerve innervates multiple organs, multiple off-target effects cap the maximal tolerable dose for cervical vagus nerve stimulation and likely also limit effective therapy.²⁸ Splenic neuromodulation, in contrast, targets the splenic nerve which is the downstream organ-specific nerve that exerts the modulatory effects on the circulating immune cells present in the spleen, thereby allowing selective engagement of the immunomodulatory neural circuit without being limited by the off-target side effects inherent to cervical vagus nerve stimulation. This mechanistic selectivity and the promising initial effectiveness data in difficult-to-treat rheumatoid arthritis patients presented here provides a strong rationale for further clinical evaluation of this approach.

In conclusion, chronic SpNS with the Galvani System demonstrated a favourable safety profile, and tendency for improvements in disease activity, pain and quality of life in patients with difficult-to-treat rheumatoid arthritis. More importantly, adjunct treatment for patients with difficult-to-treat rheumatoid arthritis demonstrated potential for higher and sustained responses in DAS28-CRP scores, improvements across multiple other disease and patient reported outcome measures, and, in exploratory analyses, a twofold increase in ACR20/50/70 responses compared to monotherapy. The observed reduction in pro-inflammatory cytokines and Th17 cells in biomarker assessments in exploratory analyses further supports the immunomodulatory effects of the therapy. The preliminary potential indication of positive efficacy results and favourable risk–benefit profile from this study warrants a larger, randomized trial to evaluate the long-term efficacy and role of splenic neuromodulation as monotherapy and in adjunct use with b/tsDMARDs.

Contributors

S.W.T., P.P., D.H.G., A.K., C.B., K.F., E.I., P.H.M., O.R., M.J.V., I.G., and R.M.A. conceptualised the study, designed the methodology, conducted the formal analysis and wrote the original draft of the paper. J.C., M.J.H.B., M.P.-B., C.-Q.C., E.L., P.W., V.V.A., T.L.F., M.J., M.D.P.L, J.W., F.A.K, K.K., L.L.S., K.R., D.M.S., H.W., A.M., I.G., and J.C.A.B. performed the investigation. All authors interpreted the data and contributed to reviewing and editing the paper. I.G., and O.R. accessed and verified the data. The corresponding author had full access to all data and final responsibility for the decision to submit for publication. P.P., and M.P.-B. could not be reached to complete Author statement forms, but all other authors vouch for their contribution.

Data sharing statement

The datasets generated and analysed during this study are not publicly available. Galvani Bioelectronics is committed to enabling qualified external researchers to request access to anonymised patient-level data and supporting clinical documents from eligible studies. Data-sharing requests are evaluated and approved based on scientific merit and feasibility. All shared data are anonymised in accordance with applicable laws and regulations to protect participant privacy. For further information on the process, or to submit a request, please contact Galvani Bioelectronics.

Declaration of interests

S.W.T received institutional grants from GSK, Pfizer, Galvani bioelectronics, Citryll, Roche, AstraZeneca, Eli Lilly; consulting fees from Galvani bioelectronics; honoraria for lectures from NovoNordisk, Pfizer, UCB. J.C. declared that his institution received payments for the support of the present manuscript and in support for attending meetings and/or travel. M.J.H.B. received institutional funding from Galvani Bioelectronics as the co-investigator on the study in support of the present manuscript and she also participated in advisory board for Galvani Bioelectronics to discuss outcomes and further avenues of research coming forth from this study. D.H.G., P.W., T.L.F., V.V.A. received an institutional grant from Galvani bioelectronics. A.K. received an institutional grant from Galvani Bioelectronics, consulting fees from AbbVie, Allinbio, Gilead, Genzyme, GSK, Novartis, Pfizer, Sanofi/Regeneron, Synact, UCB, payment or honoraria for lectures, presentations from AbbVie, Excel continuing Education, Lilly, GSK, Pfizer, Prime, Sanofi-Regeneron, UCB, participation on a DSMB or Advisory Board from UCB. M.P.-B., have no conflicts of interest to declare. C.-Q.C. received consulting fees from GeneScience Pharmaceuticals. E.L. declared that Galvani (Sponsor) and Inland Rheumatology Clinical Trials, Inc. (Institution) have executed a clinical trial agreement that includes administrative start-up funding. P.W. received payment for honoraria/speaker bureaus from Abbvie, Lilly, Novartis, Amgen and AstraZeneca; payment for expert witness consulting on patent issues from Abbvie and research support on clinical trials from Abbvie, Lilly, Setpoint, Sunpharma, Viela Bio, Galvani and Gilead. J.W. received an institutional grant from Galvani bioelectronics and consulting fee from Boston Scientific Corp. M.D.P.L received an institutional grant and consulting fee from Galvani bioelectronics. C.B. and O.R. are employees of GSK supporting Galvani for statistical aspects relating to the study; GSK was reimbursed for their work. K.F., E.I., K.R., K.K., P.H.M., H.W. L.L.S., M.J.V., D.M.S., I.G., and R.M.A. are employees of Galvani Bioelectronics/GSK and declared that Galvani provided reimbursement for meeting/congress fees, travel, employee compensation including long term incentives or patents related to treatment with a neuromodulation device for acute and chronic conditions. K.F. declared that GSK is an investor in Galvani Bioelectronics. J.C.A.B., received an institutional grant from Galvani bioelectronics, Medtronic, Hanarth fund, and consulting fee from Galvani Bioelectronics. M.J., and A.M. have no conflict of interest to declare. F.A.K., and P.P. could not be reached to complete ICMJE forms so their declarations are not known.