Application of kilohertz-frequency block to mitigate off-target motor effects of vagus nerve stimulation in swine

Cheng, Kevin P.; Deshmukh, Ashlesha; Gholston, Aaron K.; Settell, Megan L.; Knudsen, Bruce E.; LaLuzerne, Maria; Trevathan, James K.; Suminski, Aaron J.; Pelot, Nicole A.; Grill, Warren M.; Ludwig, Kip A.

doi:10.1038/s41467-025-67823-0

Download PDF

Article
Open access
Published: 30 December 2025

Application of kilohertz-frequency block to mitigate off-target motor effects of vagus nerve stimulation in swine

Nature Communications volume 17, Article number: 1066 (2026) Cite this article

2331 Accesses
19 Altmetric
Metrics details

Subjects

Abstract

Selective activation of small versus large diameter nerve fibers represents a broad challenge in the electrical modulation of neural activity for clinical applications. For example, off-target activation of motor nerve fibers was implicated in several failed vagus nerve stimulation (VNS) clinical trials by effectively limiting maximum applicable current amplitudes, thereby preventing robust activation of smaller diameter fibers responsible for therapeutic effects. Kilohertz high-frequency (HF) block is one of several strategies proposed to overcome this gap, but has yet to be thoroughly investigated in large, multifascicular, mixed nerves comparable in scale to humans. Here, we evaluated effects of HF block on the neural, muscular, and cardiac responses evoked by traditional low-frequency (LF) VNS in a large animal (swine) model. We found that HF selectively blocks neural fibers, and their associated physiological responses, in a size dependent manner with a high degree of specificity and establish initial thresholds for achieving block in large multifascicular nerves. Critically, we demonstrate that appropriately titrating HF amplitude can eliminate off-target LF-evoked motor responses while preserving activation of smaller diameter fibers targeted by clinical VNS. Beyond VNS, these data establish HF as a potentially viable strategy in any neuromodulation therapies where off-target activation of large diameter fibers limits therapeutic efficacy.

Non-monotonic kilohertz frequency neural block thresholds arise from amplitude- and frequency-dependent charge imbalance

Article Open access 03 March 2021

Control of spatiotemporal activation of organ-specific fibers in the swine vagus nerve by intermittent interferential current stimulation

Article Open access 13 May 2025

Vagus nerve stimulation using a miniaturized wirelessly powered stimulator in pigs

Article Open access 17 May 2022

Introduction

Electrical modulation of neural activity, also known as bioelectronic medicine or electroceuticals, is a rapidly growing medical device sector with applications across a broad spectrum of conditions^1,2,3. For example, vagus nerve stimulation (VNS), which is FDA-approved for the treatment of depression, epilepsy, obesity, and stroke after-effects, is also being investigated to treat heart failure, diabetes, and chronic inflammatory diseases^4,5,6,7,8,9. The range of clinical applications of VNS reflects the breadth and complexity of the anatomy of the vagus nerve (VN) with afferent pathways conveying sensory information from the viscera and efferent pathways exerting autonomic influence on many end organs^10,11,12. However, as a mixed nerve, the VN also contains large diameter myelinated motor A-alpha (m-Aα) fibers innervating muscles of the neck^10,13,14,15. Due to their large diameter, these m-Aα fibers are activated at lower stimulation amplitudes than the smaller diameter A-delta (Aδ) sensory afferents and parasympathetic B-fibers that are therapeutically targeted for most VNS indications^{14,15,16,17,18}. The relationship between fiber diameter, function, and activation threshold highlights a broader challenge across bioelectronic medicines: activation of off-target large-diameter m-Aα fibers frequently limits the stimulation amplitude that can be applied due to unwanted, sometimes painful, side effects^14,19,20,21.

Experimental and computational modeling studies—spanning rodents, dogs, pigs, and humans—show that thresholds for activation of vagal m-Aα fibers are ~1/5^th of the amplitudes needed to activate small-diameter therapeutic fibers^{14,15,16,17,22}. Thus, current VNS clinical practice for epilepsy requires a months-long process to habituate patients to the m-Aα activation by incrementally increasing the applied VNS current, which typically plateaus at a ‘therapeutic dose’ of 1.5–2.25 mA¹⁹. Unfortunately, these amplitudes are still insufficient to activate all but the most superficial Aδ/B fibers, and may not activate any therapeutic target fibers in some patients depending on subject-to-subject variations in vagal anatomy^{14,15,16,17,23}. As a result, therapy-limiting side effects were implicated in a host of recent failed VNS clinical trials, despite promising results in animal models²⁴. For example, in the NECTAR pivotal trial of VNS to treat heart failure, it was reported that only 12% of patients exhibited the stimulation-induced changes in heart rate (HR) mediated by activation of parasympathetic B-fibers²¹. Off-target neck muscle activation was the most commonly cited reason for limiting the VNS current, resulting in amplitudes at or beneath the threshold for activating B-fibers in large animal models^{14,17,19,20,21}.

Recognizing the challenges of selective activation in translational neuromodulation, multiple strategies have been proposed to reverse the conventional order of stimulation recruitment and achieve activation of therapeutically-targeted small diameter fibers without activation of large diameter fibers that mediate side effects^{15,18,20,21,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40}. Spatially selective cuff electrode designs aim to exploit somatotopy of the VN by positioning small electrode contacts close to the therapeutic target fibers and far from fibers that cause side-effects. However, individual anatomical variability may limit the effectiveness of this strategy^{13,23,28,38,39}. Novel stimulation waveforms, such as anodal block, depolarizing pre-pulses, slowly rising ramp pulses, and very low-frequency (LF) sinusoids have also been developed and investigated to try and achieve selective activation of small diameter fibers^{18,25,30,33,41,42,43,44}. While some approaches have narrowed the difference in activation thresholds between m-Aα and smaller diameter therapeutic fibers, none has shown a sufficiently large effect to overcome the ~5X threshold difference needed to activate smaller fibers before m-Aα’s.

High-frequency (HF) waveforms (>5 kHz) have recently received increased attention for therapeutic applications. With appropriate waveform parameters, HF signals can block propagating action potentials (APs)^{45,46,47,48,49,50,51,52,53,54,55}, and larger diameter fibers are blocked at lower amplitudes than smaller diameter fibers^49,50,53,56. The feasibility of using solely HF signals to selectively activate small diameter fibers while blocking large diameter fibers was recently demonstrated in a rodent model⁴⁰. However, it is unclear whether similar results are possible in a more clinically relevant large animal model, with a thicker epineurium, multi-fascicular nerve morphology, a greater range of electrode-fiber distances, more numerous nerve fibers, and larger electrodes. Moreover, at sub-block amplitudes, HF evokes asynchronous activity as opposed to the well-characterized and synchronized activity generated by traditional LF stimulation^{46,50,51,52,53,56}. This creates a large unknown variable in appropriate selection of HF parameters for activation to achieve desired clinical effects when compared to LF stimulation where a large body of preclinical and clinical data establishing stimulation parameters, safety, and efficacy already exist^24,57.

We evaluated whether HF signals layered onto standard LF VNS (LF-VNS) could selectively block m-Aα fibers to achieve net selective activation of small diameter fibers in a large animal model (swine, n = 5). We measured the evoked neural, muscular, and cardiac responses to traditional LF-VNS while applying a 10 kHz HF sinewave at varying amplitudes. We demonstrate that increasing HF amplitudes can selectively block components of the LF-VNS evoked compound action potential (eCAP), following the order of recruitment from larger to smaller diameter fibers. After establishing selectivity, we then evaluate the HF dosing that blocks LF-VNS-evoked small fiber activity and associated HR responses. Importantly, we establish that the HF amplitude can be titrated to eliminate LF-VNS-evoked motor side-effects while preserving activation of smaller diameter fibers and their consequent effects on HR in a mixed nerve with a size and fascicular organization comparable to humans^13,23.

Results

10 kHz HF blocks VNS-evoked motor side effects

We applied cervical LF-VNS while measuring the evoked electromyogram from the cricoarytenoid neck muscles (EMG_CA) to assess the effects of a 10 kHz sinusoidal HF signal on LF-VNS motor side effects (Fig. 1A). The HF signal was applied through a separate cuff located distally to the LF-VNS cuff. Trials were broadly sub-divided into EMG and Cardiac categories which differed in both LF-VNS amplitude and frequency applied (Fig. 1B, see Methods). Each trial was further sub-divided into three periods: pre-HF to assess the physiologic effects of LF-VNS alone, LF-VNS + 10 kHz HF to assess the effects of HF on the LF-VNS-evoked responses, and post-HF to assess how the LF-VNS-evoked responses recovered following cessation of HF (Fig. 1B). Fig. 1C shows example EMG_CA responses to LF-VNS with increasing HF current amplitude.

**Fig. 1: 10 kHz HF blocks LF-VNS-evoked muscle activity.**

HF blocked the LF-VNS-evoked muscle activity in an amplitude-dependent manner, and increasing HF caused partial and then complete block of the evoked EMG_CA (Fig. 1C–E). To generate HF dose-response curves (DRCs), we calculated the root-mean-squared of the EMG_CA (V_rms) for the pre-HF and during-HF periods; to account for surgical and subject differences, EMG_CA values were normalized (Ṽ_rms) within trial by the pre-HF EMG_CA V_rms. The difference (ΔṼ_rms) between the pre-HF and during-HF Ṽ_rms was calculated to quantify the change upon LF-VNS evoked EMG_CA elicited by HF stimulation. Low HF amplitudes (0.5 mA) generated little to no block of the LF-VNS-evoked muscle activity (HF = 0.5 mA, ΔṼ_rms = 1.6 ± 0.86%, mean ± 1 standard error of mean (s.e.m.)). Intermediate HF amplitudes (HF = 1-3 mA) elicited partial block of EMG_CA (ΔṼ_rms = −32.6 ± 9.3%, −61.2 ± 11.7%, −89.1 ± 4.5 @ 1, 2, 3 mA respectively, Supplementary Fig. 1), which tended to recover in amplitude over time, but did not return to the full amplitude EMG_CA within the 30-second HF period (Fig. 1D, Supplementary Fig. 8). As the HF amplitude was increased >4 mA, we observed near total block of the evoked EMG_CA (HF = 4 mA, ΔṼ_rms = −95.6 ± 1.6%). The dramatic reduction in EMG and small variance across subjects at 4 mA HF and beyond suggest that this amplitude was sufficient to block nearly all m-Aα’s within the cervical VN regardless of anatomical differences. We also confirmed that the HF signal generated local conduction block of fibers within the HF cuff as the HF signal did not block EMG_CA when LF-VNS was applied caudally to the HF cuff (Supplementary Fig. 3). Further, the HF signal could not block the EMG evoked by the superior laryngeal nerve (SL) pathway because the SL nerve fibers do not pass through the HF cuff. The SL branch point is cranial to the LF-VNS cuff (and HF cuff) (Supplementary Fig. 4), and the SL pathway is activated by current escape from the LF-VNS cuff.

Titrating HF amplitude blocks individual eCAP components

For mitigating off-target muscle activation in clinical applications the goal of HF stimulation is to block activity of m-Aα’s without blocking smaller therapeutic target fibers and, ideally, with minimal impact on natural physiological traffic. After establishing that HF could block EMG_CA responses, we next sought to determine whether the progressive increase in thresholds to block smaller diameter nerve fibers in rodents^{49,50,53,54,56} could also be achieved in the swine VN, a mixed nerve of comparable size to the human vagus.

Longer latency components of LF-VNS eCAPs (i.e., from smaller diameter fibers) were blocked at higher HF amplitudes than shorter latency components. We consistently observed at least three eCAP components, depending on the amplitude of LF-VNS (Fig. 2), and confirmed their neural origin using a muscle paralytic (Supplementary Fig. 4). At 1 mA LF-VNS, two eCAP components were present accompanied by an EMG artifact that occurred ~8 ms after the LF-VNS pulse (Fig. 2A) whereas the 3rd eCAP component was only present at higher LF-VNS amplitudes (Fig. 2B). These components can be broadly categorized into m-Aα/Aβ (Component C1, CV ≈ 43 m/s, red-shaded region), Aγ/Aδ (Component C2, CV ≈ 25 m/s, blue-shaded region), and slower Aδ-delta/B-fibers (Component C3, CV ≈ 10 m/s, green-shaded region) based on the Erlanger-Gasser scale^{14,15,58,59,60,61}.

**Fig. 2: Components of the LF-VNS-evoked nerve response (eCAP) are selectively blocked by increasing amplitudes of 10 kHz HF.**

The presence of EMG contamination in eCAP recordings can make interpretation of longer-latency eCAP components difficult as the EMG contamination may be mistaken for (e.g. Fig. 2A dashed box) or distort the long-latency, slower (small fiber), neural eCAP, (e.g. Fig. 2B, eCAP C3 at HF = 0 mA)^14,15. When the LF-VNS amplitude was increased to activate smaller diameter fibers and HF amplitude was above the complete EMG_CA block threshold, elimination of the EMG artifact provided clear identification of eCAP C3 (Fig. 2A, B dashed box). eCAP C3 had a conduction velocity consistent with activation of smaller Aδ/B fibers, i.e., the intended target for most VNS therapies. In addition, we took two approaches to confirm that the analyzed eCAP signals were neural in origin rather than being confounded by EMG artifacts. First, we repeated a subset of eCAP recordings with a paralytic to confirm the neural origin of eCAP C3 and its attenuation by high amplitude HF (n = 3/5 subjects, Supplementary Fig. 5). Second, as previously mentioned, the distally applied HF signal did not block EMG activated by current escape from the LF-VNS cuff cranially to the SL-nerve (Supplementary Fig. 4). Thus, we excluded subjects (P01, P04) and LF-VNS amplitudes that evoked a secondary short-latency EMG in the analysis of eCAP C3.

Each of the eCAP components was affected by the HF signal at different amplitudes in a latency-dependent—and therefore fiber size-dependent—manner from larger to smaller diameter fibers (Fig. 2, Supplementary Fig. 5 for individual responses). eCAP C1 (Fig. 2A, red-shaded region) was significantly attenuated at HF amplitudes ≥2 mA (ΔṼ_rms = −56.6 ± 6.0%, p = 0.007), eCAP C2 (Fig. 2A, blue-shaded region) was significantly attenuated at HF amplitudes ≥4 mA (ΔṼ_rms = −62.23 ± 9.3%, p = 0.006), and eCAP C3 (Fig. 2B, green-shaded region) was significantly attenuated at HF amplitudes ≥15 mA (ΔṼ_rms = −73.4 ± 5.4%, p = 0.026) (see Supplementary Fig. 1 for all HF amplitudes). Thus, HF can be titrated to block specific fiber groupings based on diameter while leaving the next largest fiber grouping largely unaffected until increased amplitudes of HF are applied.

The coincident changes in the EMG_CA response and eCAP C1 as HF amplitude was increased suggested that this component of the eCAP reflected activation (and block) of m-Aα nerve fibers responsible for therapy-limiting side-effects in clinical VNS (Fig. 2C, eCAP C1 vs EMG_CA DRC). This was supported by the strong correlation between the eCAP C1 and the LF-VNS evoked EMG_CA response (Spearman’s R = 0.95, Supplementary Fig. 6) and is consistent with a neural conduction block mechanism for HF. Conversely, eCAP C2 correlated poorly with the EMG_CA response (Spearman’s R = 0.46, Supplementary Fig. 6D) and may represent intermediate diameter vagal afferents.

Onset EMG effects scale with HF amplitude

The effect of HF on an individual nerve fiber is dependent on the strength of the HF electric field; at lower amplitudes, HF induces a transient onset response and can evoke sustained (albeit asynchronous) activation at amplitudes less than required to produce block^{47,49,52,53,62}. Therefore, within a given nerve with different fiber diameters and electrode-fiber distances, some fibers may be blocked, while others may be in an activated (sub-block) state. We analyzed this asynchronous activity by subdividing the EMG_CA signal following each LF-VNS pulse. We calculated V_rms of the LF-VNS-evoked EMG_CA based on the latencies of the responses during the pre-HF, LF-VNS-only phase; we calculated V_rms of the HF-VNS-evoked EMG_CA during the remaining time points, excluding the period of the LF-VNS stimulation artifact (Fig. 3A). To visualize the asynchronous activity we also generated 2D and 3D color-coded pulse-by-pulse raster plots of the EMG_CA over the course of a stimulation trial depicting the combined LF-VNS-evoked and HF-evoked EMG_CA response (Fig. 3B; see Supplementary Fig. 9 for explanation of raster plot generation).

**Fig. 3: Onset features of the HF response scaled with HF amplitude.**

Consistent with previous rodent studies, the EMG_CA showed an immediate onset response (Fig. 3B, onset spikes) at the initiation of HF^47,63,64. Depending on the HF amplitude, this onset response was accompanied by asynchronous muscle activity (Fig. 3B, LF-VNS + 10 kHz HF), indicative of sustained, asynchronous, m-Aα activation. We quantified the degree of HF-induced activity by the number of twitch responses evoked during the HF period (minus the LF-VNS-evoked EMG_CA; Fig. 3C) and their relative magnitude in comparison to the LF-VNS-evoked EMG_CA (Fig. 3D). This asynchronous activity generally diminished over the duration of HF at sub-block amplitudes (Fig. 3A, 1–3 mA HF, Supplementary Figs. 8, 10) and was less present at HF amplitudes that completely blocked the LF-VNS-evoked EMG_CA (Fig. 3A, 4 mA HF). The number of muscle twitches decreased as the HF amplitude was increased (Fig. 3B, C) whereas their relative magnitude was consistently ~20% of the LF-VNS (1 mA) evoked EMG_CA response regardless of HF amplitude (Fig. 3D, Kruskal-Wallis H = 2.59, p = 0.62). The consistent magnitude of the asynchronous twitches may be explained by asynchronous activation of motor units by HF, in contrast with the LF-VNS-evoked synchronous activation that results in temporal summation.

At HF amplitudes which evoked partial block, the LF-VNS-evoked EMG_CA appeared to be inversely correlated with the HF-evoked asynchronous activity (Fig. 3A, Supplementary Fig. 8). This suggests a potential alternative, or complementary, mechanism for partial block than noted above (i.e., variable electrode-fiber distances causing variable HF block thresholds): reduction of the LF-VNS EMG_CA response could be due to collision block between orthodromic LF-VNS evoked APs and antidromic HF evoked APs. In other words, as the HF-evoked asynchronous activity gradually diminishes over the course of the trial, fewer LF-VNS-evoked APs undergo collision block, resulting in a larger synchronized EMG_CA response. This decaying time course of HF-evoked activity is consistent with prior studies^62,65,66.

Lastly, consistent with rodent studies, in a subset of subjects, we observed carryover effects of HF where the LF-VNS-evoked EMG_CA remained blocked or partially blocked beyond the HF period, albeit only at the highest HF amplitudes tested (Supplementary Fig. 8 & 10)^47,67. However, as carryover effects were not consistent across subjects, and we did not perform trials extending the ‘post-HF, LF-VNS’ period to assess recovery, we did not perform cohort level analysis on this effect.

10 kHz HF effects on the LF-VNS cardiac response

Recent failed pivotal trials for VNS to treat heart failure demonstrate the necessity of developing methods to activate small diameter fibers selectively while blocking large diameter fibers responsible for limiting patients to sub-therapeutic doses^{21,24,68,69,70,71}. We sought to determine whether m-Aαs can be blocked while preserving LF-VNS cardiac effects by first characterizing the effect of varying levels of HF on the LF-VNS cardiac response (Fig. 4A).

**Fig. 4: HF blocking signal effects on LF-VNS evoked cardiac response.**

Low amplitudes of LF-VNS evoked tachycardia that transitioned to bradycardia at higher LF-VNS amplitudes (Fig. 4A, top row)⁷². When lower amplitude HF (but still above EMG_CA block threshold, Fig. 1) was also applied, no changes in cardiac effects were observed at any LF-VNS dose (Fig. 4A, green) indicating that HF levels that blocked EMG_CA (i.e., m-Aαs) were insufficient to cause activation or block of the smaller diameter vagal cardiac B-fibers responsible for LF-VNS bradycardic effects. At 10 mA HF, a small onset bradycardia was observed at the initiation of HF in some trials as evidenced by the sharp change in HR during the first several seconds of HF (Fig. 4A, yellow) which was stronger at 15 mA HF. This transient response is likely due to brief HF-induced onset activation of B-fibers in a manner similar to the onset EMG_CA activity (Fig. 3A, B). When the LF-VNS current amplitude was high enough to evoke strong bradycardia (Fig. 4A, right column), block of LF-VNS-evoked bradycardia only occurred at high levels of HF (Fig. 4A, gray) with higher levels of HF evoking more complete block of LF-VNS bradycardic effects (Fig. 4B).

In contrast, the effects of HF on LF-VNS-evoked tachycardia (Fig. 4A, left and middle columns) were more complex with some trials exhibiting apparent block whereas others only exhibited a transient bradycardic HF onset response. LF-VNS tachycardic effects are believed to be centrally mediated responses to vagal afferents with multiple effector pathways besides the VN^15,72. Thus, application of HF caudal to the LF-VNS cuff, as in our experiment, would not be expected to block these effects. Instead, trials in which HF induced a sustained bradycardic effect during an LF-VNS tachycardic response (Fig. 4A, 3 mA LF-VNS + 15 mA HF, Supplementary Fig. 11A) may have been the result of HF-induced activation of parasympathetic cardiac B-fibers overriding the centrally mediated tachycardic effects.

We compared eCAP C3, whose conduction velocity was consistent with small Aδ/B-fibers, with the HR response to determine whether the HF effect on LF-VNS cardiac responses was reflected in the eCAP. However, the long inter-trial duration for washout, the high LF-VNS amplitudes required to evoke B-fiber mediated bradycardia, and the complexity of the HR responses precluded generation of dose response curves with sufficient resolution to correlate bradycardia to the amplitude of the Aδ/B-fiber eCAP components. Nonetheless, the amplitude of eCAP C3 tracked well with the effects of HF on LF-VNS-evoked bradycardia indicating that it may represent these small diameter vagal efferents (Supplementary Fig. 12).

HF can selectively block LF-VNS-evoked EMG without influencing LF-VNS cardiac responses

Our results support the feasibility of titrating LF-VNS and HF amplitudes to eliminate EMG_CA responses (i.e., side effects of clinical VNS) while preserving LF-VNS-mediated changes in HR, including tachycardia (Fig. 5A, Supplementary Fig. 11A) and bradycardia (Fig. 5B, Supplementary Fig. 11B). Paired EMG_CA and HR responses demonstrate that 4 mA HF completely blocked EMG_CA and produced no effects on the LF-VNS cardiac response (Fig. 5A, B). In practice, this would enable higher amplitudes of LF-VNS to be applied clinically, as required for greater therapeutic effects, than would be possible without HF, where neck muscle activation limits the LF-VNS current level^{14,17,19,20,21}. Only when HF amplitude was increased sufficiently (>10 mA) to interact with smaller diameter fibers were effects evident on LF-VNS evoked cardiac responses (Fig. 5A, B, Supplementary Fig. 11).

**Fig. 5: HF can completely block LF-VNS-evoked EMG_CA without influencing LF-VNS-evoked heart rate changes.**

An ‘HF-window’ was identified in all subjects (Fig. 5C, light blue shaded region) where the LF-VNS EMG_CA was reduced to less than 20% of its pre-HF magnitude, while maintaining LF-VNS-evoked changes in HR. HF amplitudes of 2.4 ± 0.24 mA blocked EMG_CA, whereas HF amplitudes of 10.2 ± 1.42 mA were required for HF-evoked changes in HR (see Supplementary Fig. 11 for full cohort HF-windows). This approximately five-fold difference in thresholds is due to the fiber diameter selectivity of HF and suggests there is a dynamic range over which HF can be titrated to block the LF-VNS evoked EMG_CA without interfering with therapeutic cardiac effects. Note however that due to experimental time constraints, not all HF amplitudes within this range were tested (Fig. 5C, striped region), and it is possible that the HF window is smaller than reported here (i.e., on average, the highest HF amplitude tested that did not evoke any cardiac responses was 6.4 ± 2.19 mA).

Discussion

This study demonstrates the potential clinical utility of HF to selectively block nerve fibers within large-diameter mixed nerves like the human vagus. Prior studies of HF were predominantly conducted in rodents, where the combination of smaller electrodes, nerve anatomy, and electrode-fiber distances results in larger differences in transmembrane voltage changes at a given applied current in comparison to pig or human nerves^{16,23,40,47,73,74,75,76,77}. Failure to account for these scaling factors is likely, at least in part, to be responsible for the repeated inability to recapitulate promising rodent results in clinical translation to humans and open questions remained on the applicability of HF to large scale, multifascicular nerves⁷⁷. The few HF studies performed in large animals largely focused solely upon LF-evoked motor responses^{78,79,80,81,82}, which result from activation and/or block of large diameter m-Aα fibers that have 5-10x lower thresholds than the smaller diameter therapeutic fibers targeted for VNS therapies^{14,15,16,40,58}. In this study, we assessed whether HF could be used in conjunction with LF-VNS to prevent conduction of neural activity in the side-effect producing m-Aα’s in the pig large animal model, while leaving propagation intact in therapeutic smaller diameter fibers which form the basis for some clinical applications of VNS. Our data demonstrates that this can be achieved by 10 kHz HF with a high degree of selectivity for the fiber type blocked based on HF amplitude. Moreover, our results indicate that appropriately titrated HF creates conduction block local to the HF cuff (Supplementary Fig. 3,4)—consistent with prior studies of HF block^83,84,85—and would therefore not interfere with VNS therapies targeting afferent, centrally-mediated, effects (e.g., epilepsy, depression) while preventing motor associated side-effects. This opens the potential for greatly increasing current amplitude limits in clinical VNS by eliminating the primary constraint of neck muscle activation cited by multiple failed clinical trials^9,21,69,70.

HF signals putatively block conduction by generating rapid transmembrane voltage oscillations that are rectified by non-linear membrane properties to cause net depolarization and inactivation of sodium channels (i.e., comparable to the refractory period)^46,49,50. The sustained average depolarization during the application of HF prevents repolarization and sodium channel de-inactivation, leaving them in the inactivated state^46,49,50,86. Conduction block is created when a sufficient number of sodium channels within an axon segment are held in the inactivated state by the HF signal⁴⁹. Thus, the effect of HF is dependent on the magnitude of transmembrane potential oscillations experienced by the axon which are, in turn, dependent on the fiber diameter, the distance from the HF electrode, the fascicular structure of the nerve (and associated tissue conductivities), and the HF signal frequency, waveform, and amplitude. When generating smaller transmembrane potential oscillations, HF evokes asynchronously occurring APs for the period it is applied as an insufficient number of sodium channels are inactivated to create block^{40,47,49,50,51,53,62}. In contrast, when larger transmembrane potential oscillations are generated, HF will drive most sodium channels into the inactivated state thereby inhibiting or fully blocking the propagation of APs^{40,47,49,50,51,53,87,88}.

The graded block of LF-VNS EMG_CA as HF amplitude increased (Fig. 2C, Supplementary Fig. 2,8) is consistent with lower HF block thresholds for m-Aα fibers closer to the HF electrode and higher block thresholds for m-Aα fibers further from the electrode. The larger variance in the partial EMG_CA block responses at intermediate HF amplitudes (Figs. 2C, 1–3 mA) may be due to differences in individual fascicular anatomy or fiber organization across subjects. This would change the relative distance of the EMG_CA evoking fibers from the HF cuff leading to differences in sensitivity to HF blocking effects until an HF amplitude is reached which blocks all m-Aα’s regardless of position within the nerve. Alternatively, at sub-conduction block amplitudes, HF also generates asynchronous EMG_CA activity arising from asynchronously generated APs⁸⁹, distinct from the onset response that occurs prior to local axonal conduction block. The inverse relationship between the LF-VNS and HF-evoked EMG_CA over the course of a given trial (Fig. 3A) indicates that the LF-VNS EMG_CA partial block observed could also be due to collision block between LF-VNS-evoked APs and asynchronous HF-evoked APs, given that APs generated by electrical stimulation travel bidirectionally from the site of activation. While the mechanisms of collision block and HF-evoked asynchronous activity are established and described independently^40,65,89,90, here we provide evidence of collision block caused by asynchronous HF-evoked activity, which has implications for clinical implementation of HF.

It is plausible that both mechanisms (local axonal block and collision block) contribute to block in large diameter nerves, such as the pig and human vagus nerves. In Fig. 3A, the LF-VNS-evoked EMG_CA response during 1 mA HF is initially attenuated, but almost completely recovers to the pre-HF amplitude by the end of the HF phase indicating that the attenuation may have been due entirely to collision block between LF-VNS and HF evoked APs. When HF was increased to 2 mA, we observed a similar (albeit larger) initial attenuation of the LF-VNS EMG_CA that only recovered to ~80% of the pre-HF period despite similarly diminishing HF-evoked asynchronous activity. Here, the higher HF amplitude may be causing true conduction block on a subset of fibers (likely m-Aα fibers closest to the HF electrode) which mitigated recovery of the LF-VNS EMG_CA as the HF asynchronous activity and subsequent collision block receded. As HF amplitude was further increased, true conduction block of all m-Aα fibers in the vagus manifested as the lack of recovery in the LF-VNS EMG_CA during the HF period.

The HF-evoked asynchronous response tended to diminish as the total EMG_CA block threshold was approached (Fig. 3B, C), although it was often not completely eliminated until HF current levels above the LF-VNS EMG_CA total block threshold. We also found that the magnitude of asynchronous EMG_CA responses during HF were considerably smaller in amplitude than the LF-VNS evoked response regardless of the HF amplitude (at 1 mA LF-VNS, Fig. 3D), which may have manifested by an absence of temporal summation of the asynchronous m-Aα activity. These smaller asynchronous responses may be somewhat inconsequential if they are within tolerable limits for patients undergoing the habituation process. However, as our subjects (anesthetized pigs) were unable to report on the off-target sensation, it remains unclear whether the HF-evoked asynchronous activation would be as intolerable as the LF-VNS-evoked synchronous neck-muscle activation that has limited therapeutic current levels in clinical VNS. Similarly, the partial block that we observed at lower amplitudes of HF may itself be sufficient to facilitate higher clinical LF-VNS current amplitudes if partial block can diminish neck muscle activation to tolerable levels. This approach could conserve battery life by reducing power consumption and limit potential risks by using a lower HF amplitude while still enabling a higher applied current for therapeutic LF-VNS. However, our data suggests that at partial block amplitudes the reduction in LF-VNS EMG_CA by HF was due, at least in part, to collision block. The data upon which this analysis was performed was collected during EMG trials where the LF-VNS stimulation frequency was only 8.1 Hz to prevent muscle fatigue over the length of the trial. Current clinical VNS typically stimulates at higher frequencies (e.g. 25 Hz, as in our Cardiac trials) which is closer to tetanic frequencies and therefore the contribution of collision block to reduction of the LF-VNS evoked EMG_CA may be less pronounced. We did not perform trials at lower HF amplitudes during the 25 Hz LF-VNS Cardiac trials as the goal was to investigate effects on smaller diameter fibers, requiring much higher HF amplitudes than those in which potential collision block was observed on the LF-VNS evoked EMG_CA.

We analyzed HF asynchronous activity from the EMG_CA but not eCAPs due to the inherently lower signal-to-noise ratio of eCAPs vs. EMG in comparison to the HF artifact. Further, classifying eCAP components based on latency would not be possible for asynchronous HF-induced eCAP signals. Nonetheless, our cardiac data indicate that smaller diameter fibers likely also exhibited some HF-generated onset activity as evidenced by the transient bradycardia observed at the initiation of HF (Fig. 4A black arrowheads, Supplementary Fig. 11). While the HF-induced EMG activity may be undesirable but ultimately innocuous muscle twitches, long-term asynchronous activity in other vagal fibers, for which there is no clear and measurable effector output, remain unknown and should be better understood before clinical implementation of HF in VNS therapy. This may be especially true in VNS applications where the intended goal is to modulate central activity, entrainment, and/or plasticity as the asynchronous HF-induced APs could interfere with those effects^{91,92,93,94,95}. Further, the potential collision block mechanism revealed by our LF-VNS vs. HF EMG_CA analysis warrants consideration because HF-evoked asynchronous APs could also, unknowingly, prevent LF-VNS activation of smaller diameter fibers from reaching their intended downstream targets if the HF amplitude is set too high. HF stimulation, via conduction or collision block, could also interfere with natural physiological traffic, such as large diameter sensory afferents from the viscera if the HF amplitude is not appropriately titrated to leave their activity intact. Lastly, the high HF amplitudes required to block smaller diameter fibers also carries implications for clinical applications in light of the asynchronous activity evoked by HF at sub-block amplitudes. As the strength of an electric field falls off with distance from the electrode, application of high HF amplitudes to block small diameter fibers local to the HF electrode could cause unwanted asynchronous activation of large diameter fibers in nearby nerves (e.g., CA activation via the SL pathway, Supplementary Fig. 4). Thus, a thorough understanding of the putative consequences of HF evoked asynchronous activity must be considered in the translatability of HF block to clinical implementation.

Likewise, it remains unclear if persistent conduction block of large diameter fibers within the cervical vagus could create unwanted side effects. In the case of m-Aα’s, this may impair voluntary control of these muscles during the HF period and potentially cause chronic issues, such as those observed in long-term denervation⁹⁶. The long-term effects of HF on axonal homeostasis are poorly understood. Changes in sodium channel kinetics may result in compensatory changes in other voltage-gated ion channels, potentially impacting intracellular ion concentrations, and affecting the functions of the endoplasmic reticulum, mitochondria, and other organelles^97,98,99,100. In this study, we did not collect the HF stimulated nerves for any post-mortem assessments of damage as it would be difficult to separate pathologic effects caused by the experimental preparation (e.g. surgical dissection, cuff instrumentation, open air surgical pocket) from those caused solely by HF stimulation. However, comparison of eCAP responses to LF-VNS pulses presented at differing timepoints throughout the day suggest that the repeated HF stimulations did not have a significant deleterious effect on the underlying nerve’s ability to generate and propagate APs (Supplementary Fig. 13). In addition, the HF block thresholds for larger diameter fibers in this study are within the amplitude ranges applied in currently FDA-approved HF stimulation therapies for obesity and pain^101,102. Nonetheless, more in-depth and chronic studies should be performed profiling the genomic, transcriptomic, and proteomic effects of HF on individual nerve fibers that better capture its long-term effects, especially if the intent is to block smaller diameter fibers with high HF amplitudes^42,44,50.

Application of HF at the current levels needed to achieve block of large diameter fibers also has important instrumentation, electrochemical, and energy consumption issues to consider. In terms of instrumentation, inadequate control of direct current (DC) leakage can occur during HF signals, which may be responsible for inconsistent results between studies and could cause harmful electrochemical reactions^50,103. In this study, a specialized DC blocking circuit was used to eliminate this problem¹⁰³. Moreover, the existing paradigm to assess long-term electrochemical risks—the Shannon limit¹⁰⁴—established guidelines derived from phenomenological studies that are only applicable to stimulation under specific conditions; analogous guidelines specifically associated with HF signals should also be investigated and established^{57,104,105,106,107}. The higher frequencies used by HF block offers a distinct advantage in comparison to traditional LF stimulation in regard to electrochemical risks. Previous studies¹⁰⁸ and benchtop electrode impedance spectra (EIS) collected prior to each experiment (validated in vivo in a single subject) show that the 10 kHz electrode impedance falls within the regime dominated by solution resistance (~500 Ω, Supplementary Fig. 14A, B). In combination with the small estimated charge density per phase (Supplementary Fig. 14C) afforded by the larger electrodes (relative to rodent studies) this suggests that charge transfer was primarily achieved via capacitive double-layer charging; thereby reducing risks from the buildup of faradaic electrochemical reactions⁵⁷ and allowing electrode polarization to remain within water window limits for platinum electrodes (dependent on electrode geometry)¹⁰⁸. Additionally, power and energy consumption are also a concern as the continuous application of HF could quickly deplete implantable pulse generator (IPG) batteries, potentially necessitating more frequent surgical replacements, which have risks, or require use of a frequently rechargeable system¹⁰⁹. The combination of higher frequencies and larger electrodes for pigs and human vagus nerves mitigates this problem to some degree as the low impedance at high frequencies reduces the stimulator voltages and power necessary to generate the specified currents (see Supplementary Fig. 14 for power and battery life estimates utilizing HF amplitudes tested in this study). The tripolar configuration in which we applied HF is also more energy efficient in achieving block and mitigating onset responses as compared to bipolar or monopolar configurations⁵². Notably, the ‘HF window’ found to block large diameter fibers without affecting small diameter fibers is potentially within the capabilities of already FDA-approved HF stimulation IPGs (e.g. for peripheral^110,111 and spinal cord (SCS)^55,112 stimulation) which greatly facilitates the potential translation and clinical implementation of HF block to VNS and other modalities. However, while these devices may operate at the frequency and voltage requirements capable of achieving HF block of large diameter fibers, they typically apply rectangular, biphasic pulses (akin to the LF stimulus pulses) at HF frequencies rather than a continuous sinewave as in our study. Thus, further investigation is required to ascertain whether any of the existing FDA-approved HF IPG’s are capable of recapturing the same effects and selectivity seen in our results.

As noted, HF signals are FDA-approved for abdominal VNS, peripheral nerve, and spinal cord stimulation^{4,45,55,102,110,111,112,113,114,115,116}. These therapies were predicated on the concept of blocking smaller diameter fibers that are responsible for sensations of hunger or pain. Rodent studies using LF to activate smaller diameter fibers and then HF to block the LF-evoked responses were used as evidence that HF could block smaller diameter fibers to support subsequent clinical efficacy studies. However, recent modeling and functional data have called into question whether HF at the levels often used clinically^{45,102,115,117} are sufficient to block these smaller diameter fibers^46,47. Critical determinants of the efficacy of HF, such as the range of electrode-fiber distances, electrode size, and nerve anatomy are very different in rodents compared to humans or large animals^13,23,77. Although it is hard to compare current levels across different electrode designs in different anatomical systems (abdominal and cervical VNs, peripheral nerves, etc.), our data indicates that the HF threshold to begin activating smaller diameter fibers is >10 mA in the ~3 mm pig VN, and in turn, even higher amplitudes (≥20 mA) are required to block those fibers. A recent modeling study compared VNS activation thresholds of large and smaller diameter fibers across populations of pig and human nerves; to achieve equivalent neural response in humans, stimulation amplitudes had to be increased by ~1 to 3x for larger diameter fibers and ~1 to 5.5x for smaller diameter fibers⁷⁷. Given the larger scaling factors for small diameter fibers, the HF amplitudes that blocked small diameter fibers in this study are potentially lower than those needed in humans and further indicates that currently used clinical HF amplitudes are insufficient to block small diameter fibers^53,77. The estimated voltage and power requirements to achieve HF stimulation at these amplitudes (Supplementary Fig. 14) may be impractical with current clinical IPGs as opposed to the lower HF amplitudes sufficient to block large diameter fibers which are already within the operating characteristics of existing IPGs. Although these idealized estimates are based on the specific electrodes, impedances, and HF block thresholds of the present study and may not directly translate to different applications. Consequently, other mechanisms of action should be explored to explain the clinical effects, such as the consequence of HF-evoked asynchronous activation or influence of HF on pre- to post-synaptic fidelity at the synapse. These mechanisms are being investigated in another clinical application of HF, SCS for pain, where the current delivered is set at sub-perception levels¹¹⁸.

There are several additional limitations in this study that must be noted with an eye toward chronic application in human patients. First, experiments in this study were conducted in acute animals, under anesthesia, with an open surgical pocket with notably more exposure of the cervical vagus than would be present in a chronic implantation procedure. The average distance between the LF-VNS cuff and HF cuff was ~3.5 cm to prevent electric field interactions and enable disambiguation of their respective effects. In a human patient, this would require a large surgical pocket and placement of two separate cuffs; future work should be done to investigate how LF-VNS and HF effects interact when administered through a single cuff with relatively short distances between them. The anesthetized state also prevents the cuffs from moving relative to the anatomy as would normally occur from head and neck movements during daily activities. As previously mentioned, the distance between the HF electrode and fibers is critical in determining whether the effect of HF is blocking or asynchronously activating. Small movements between the cuff and underlying nerve could change the relationship between the HF electric field, the nerve, and any surrounding electrically excitable off-target structures. This would potentially lead to the transient phasing in and out of block effects or HF-induced asynchronous activation of off-target side-effects. The open surgical pocket can also alter the distribution of electric fields and thus the associated block thresholds, both by any edema that occurs due to dissection, as well as air acting as insulation on the ventral side of the preparation; in our experiments all efforts were made to remove any buildup of edema (fluids) between stimulation trials. Acute responses may also differ from sustained chronic stimulation due to fibrous scarring/encapsulation of the electrodes, ongoing physiological activity, and potential habituation to LF and HF stimulation at the axonal, muscle, and central/synaptic levels.

Overall, the data presented in this study demonstrate that 10 kHz HF signals can block LF-VNS-evoked EMG and eCAP components—while conserving LF-VNS-evoked changes in HR—with previously unreported selectivity in a large animal model. These data demonstrate the feasibility of combining HF and LF signals to mitigate off-target activation of large diameter fibers during VNS that has adversely impacted several recent VNS clinical trials. This approach of effectively augmenting existing clinical LF stimulation therapies with HF to eliminate side-effects could be rapidly translatable as our data demonstrates the HF amplitudes needed to block off-target motor effects (in VNS) are within the range of existing FDA-approved IPGs utilized by HF therapies targeting pain relief^55,110. Moreover, our results indicate that in the large animal model, HF activation and block exhibits the same order of stimulation recruitment as traditional LF stimulation, which has important implications for existing clinical HF therapies targeting small diameter fibers (e.g., SCS, abdominal VNS). In summary, HF is a promising avenue for selective activation of small diameter fibers in neuromodulation, and although much work remains to deploy HF as a reliable clinical solution, our data demonstrates HF’s potential to increase therapeutic current limits in neuromodulation applications limited by off-target large diameter fiber activation.

Methods

The experimental cohort consisted of n = 5 domestic (Yorkshire/Landrace) swine (3 female / 2 male, mean weight = 49.02 ± 4.52 kg, sourced from University of Wisconsin Swine Research & Teaching Center, Arlington, WI) housed with ad libitum access to water, fed twice daily, and under a 12-hour light/dark cycle. On the day of the experiment, subjects were induced with a combination of Telazol (6 mg/kg) and xylazine (2 mg/kg) before being intubated, mechanically ventilated, and moved to isoflurane (1.5–2%) and fentanyl (5–12 µg/kg/hr) anesthesia for the remainder of the experiment. HR, blood oxygenation (SpO₂), expiratory capnograph (CO₂), blood pressure (BP), and temperature were continuously monitored using a veterinary surgical monitor (BioNet BM5Vet Elite, Seoul, South Korea) in all subjects to monitor depth of anesthesia. Separately, electrocardiogram (ECG) and strain-gauge pulse transducer time series were also collected using an ADInstruments PowerLab 8/35 system (sampling rate 1 kHz, Sydney, Australia) and LabChart 8 software (v8.1.30) configured to provide real-time calculation of HR and synchronize with our electrophysiological stimulation and recording system. In a subset of subjects (n = 3/5), arterial blood pressure (aBP) was also collected using a Millar catheter (Millar Inc., Houston, TX, Model #SPR-350S) introduced via the femoral artery and recorded by ADInstruments/LabChart. All experiments were performed on the right cervical vagus nerve (VN) which was surgically isolated from surrounding tissue from the nodose ganglion to the sternal notch. In the first subject (subject P01), the sympathetic trunk was traveling with the VN and was not separately isolated, leading to strong stimulation-evoked tachycardia. Thus, for all remaining subjects, extra care was taken to identify and isolate the vagus and sympathetic trunk from each other when they were found conjoined.

In each experiment, nerve cuffs were placed along the length of the VN as follows (Supplementary Fig. 18A). First, a custom multi-contact circumferential cuff (inner diameter (I.D.) = 3 mm, contact spacing = 3 mm center-to-center, inner contacts = 2 mm square, surface area = 4 mm², outer contacts = 1 mm strip, surface area = 8 mm², platinum iridium (Pt-Ir) contacts) was placed ~ 0.7–1 cm caudal to the nodose ganglion for application of LF vagus nerve stimulation (LF-VNS). Next, a multi-channel longitudinal full-wrap cuff (5-channel, I.D. 3 mm, contact spacing 2 mm center-to-center, all contacts = 1 mm strip, total surface area = 8 mm², Pt-Ir contacts, Ardiem Medical, Indiana PA, USA) was placed on the vagus further along the VN at the most caudal end of the surgical window to collect evoked compound action potential (eCAP) recordings. Finally, a 3-channel longitudinal full-wrap cuff (same dimensions as eCAP cuff minus two channels, Ardiem Medical, Indiana PA, USA) for application of the HF signal was placed approximately equidistant between the LF-VNS stimulation and eCAP recording cuffs. As the size of the surgical window varied across subjects, this approach of relative cuff positioning maximizing the distance between cuffs was taken to achieve the greatest separation between cuffs and minimize interactions between them (Supplementary Fig. 18B). To measure evoked muscle activity, bipolar electromyography (EMG) leads (RLSND121-2.5; RhythmLink, Columbia SC, USA) were placed into both the cricoarytenoid (CA) and cricothyroid (CT) muscles.

Hardware and Instrumentation

The LF-VNS stimulation cuff, eCAP recording cuff, and EMG leads were connected to the respective stimulation and recording banks of a Tucker-Davis Technologies (TDT) 64-channel electrophysiology system (sampling rate, f_s = 24414 Hz, RZ2, IZV10, SI8, WS8; TDT, Alachua FL, USA), which was controlled using TDT’s Synapse Software (v96). The muti-channel circumferential LF-VNS stimulation cuff was set-up in a tripolar configuration (center channel cathode). eCAP recording channels were single-channel recordings referenced against a needle placed into the fat layer at the superficial aspects of the pocket at approximately the same cranial-caudal plane as the eCAP cuff. EMG was recorded in differential mode across the two leads of the bipolar EMG electrodes. Stainless steel pads were placed beneath the animal to serve as stimulation and recording grounds. Synapse/TDT was configured to output a TTL-pulse to a dedicated channel on the ADInstruments PowerLab 8/35 to synchronize stimulation events with physiologic measures.

The HF cuff was also configured for tripolar stimulation (center channel cathode) to limit the amount of HF artifact present. The instrumentation for delivering and quantifying the HF waveform varied slightly across subjects (Supplementary Fig. 13) as we developed the experimental preparation. In summary, a Keithley current source (Keithley 6221, compliance voltage = 60 V; Tektronix, Beaverton OR, USA) controlled via network connection by a custom Python script was used to specify the parameters of the HF waveform. Based on previous studies, we were initially concerned that the slew rate of the Keithley 6221 would not be able to keep up with the combination of a large load (electrode impedance), high frequency (10 kHz), and high current amplitudes. As a result, in some experiments (n = 2/5), an AM-Systems 4100 linear stimulus isolator was used as a pass-through amplifier for the Keithley’s output as in Pelot and Grill 2020⁴⁷. However, upon further investigation, we found that our electrodes, which are larger in size than electrodes used in rodent studies, had sufficiently low impedance to avoid significant attenuation of the Keithley 6221 output at 10 kHz within the range of HF amplitudes tested (Supplementary Figs. 14 & 15). Further, we found that the AM-Systems 4100 could introduce significant noise in eCAP recordings; therefore, the rest of the experiments (n = 3/5) were performed without the AM-Systems 4100 in the circuit. Regardless of the current source, a DC-blocking circuit^46,63 comprised of 1 µF capacitors and 100 kΩ resistors was placed in series with the HF output to prevent build-up of DC offset during HF stimulation. In addition, a 100 Ω resistor was placed into the HF circuit to assess the actual HF current delivered (measured via TDT’s built-in ADC (P01,P02) or a Keithley 6510 (P03-P05) in each subject accounting for differences in electrode impedance across subjects (Supplementary Fig. 15).

Experimental Protocol

Experimental trials were broadly divided into ‘EMG’ and ‘Cardiac’ trial sets wherein the goal of each category was to characterize the effect of HF on the LF-VNS-evoked muscle and HR responses, respectively. In all cases, the frequency of HF was 10 kHz, and the LF-VNS consisted of 250 µs/phase symmetric biphasic pulses. An individual trial set consisted of multiple LF-VNS amplitudes (randomized order) with a specific HF amplitude. Within each individual trial the presentation of LF-VNS and HF were configured to begin with an initial LF-VNS only period to establish the effects of LF-VNS followed by a combined LF-VNS + HF period to assess the effect of HF on the LF-VNS response, and lastly a post-HF LF-VNS period to assess the recovery of the LF-VNS response after HF (Fig. 1B).

EMG and Cardiac trials differed in the following ways; first, the frequency of LF-VNS in the Cardiac trials was set at 25 Hz consistent with current clinical investigations of LF-VNS for heart failure while the frequency in the EMG trials was set to 8.1 Hz. The lower frequency of the EMG trials was to prevent diminishment of the EMG signal due to muscle fatigue being mistaken as block of motor fiber conduction. Second, the duration of trial phases (i.e., pre-HF, during-HF, post-HF, and inter-stimulation periods) were longer in the Cardiac than EMG trials to account for the slower HR responses (both during stimulation and recovery after stimulation (Fig. 1B). Lastly, the range of LF-VNS and HF amplitudes applied in the EMG and Cardiac trials differed with both LF-VNS and HF amplitudes being smaller in EMG trials (LF-VNS ≤ 1 mA, HF ≤ 6 mA) than those applied in Cardiac trials (LF-VNS up to 12 mA, HF up to 25 mA).

The LF-VNS amplitude was limited during EMG trials to avoid causing longer duration cardiac effects which would have required a longer inter-stimulation period to washout. Based on our previous studies in pigs, and an initial instrumentation hardware check performed at the beginning of each experiment an LF-VNS current amplitude of 1 mA is sufficient to reliably evoke large EMG responses while avoiding significant cardiac effects⁴. In contrast, due to intersubject variability, the LF-VNS amplitude range during Cardiac trials was empirically determined by incrementally increasing the LF-VNS amplitude until a sustained bradycardic response was evoked. Likewise, HF amplitudes were empirically determined ad hoc by incrementally increasing the HF amplitude in a separate thresholding process where the LF-VNS amplitude was maintained at a constant level which evoked either strong visual neck muscle contractions (EMG trials, LF-VNS = 1 mA) or sustained bradycardia (Cardiac trials, LF-VNS varied by subject). The HF range was then determined by incrementally increasing the HF amplitude applied until A) application of HF caused visual cessation of the LF-VNS evoked neck muscle activation (EMG trials) or B) application of HF caused a reversal of the LF-VNS evoked bradycardia (Cardiac trials). For experimental comparison, we maintained relatively constant HF amplitudes within the EMG trials; however, the number of HF amplitudes tested within the Cardiac trials varied across subjects due to the substantially longer Cardiac trial times and experimental time constraints of our protocol (12-hrs from induction).

We previously showed that the electrical field generated by evoked muscle activity can contaminate neural eCAP recordings from the VN—i.e., EMG artifact—in such a way as to cause potential misinterpretation of the eCAP^14,15. Within the context of cervical VNS, activation of the CA and CT muscles may appear as components in the eCAP signal with a latency of approximately 5–10 ms post-stimulus which, depending on the distance between cuffs, places their conduction velocity in the slow Aδ to B-fiber range⁶¹. In many of our trials, application of HF signal removed this contamination by preventing activation of the CA/CT muscles via conduction block (Fig. 2). However, to confirm all eCAP components analyzed were neural in origin, we also performed a set of experimental trials with a muscle paralytic (vecuronium bromide, 1 mg/kg/hr) in a subset (n = 3/5) of subjects (Supplementary Fig. 5).

Data Analysis

To remove the 10 kHz HF artifact from the eCAP and EMG recordings, we used a custom software package developed by our lab for handling electrophysiological datasets known as pyeCAP¹¹⁹. We used a multi-step filtering process consisting of a 1D Gaussian low-pass filter (sigma = 1.29) to remove the majority of the HF artifact followed by a notch filter centered at 10 kHz to further attenuate any residual 10 kHz signal, and lastly a median high-pass filter (kernel = 201) to address any baseline drift (Supplementary Fig. 16, 17).

Identification of HF Start and End

Although the HF output was controlled via the TDT system’s computer, there was no direct synchronization between the Synapse software controlling the LF-VNS stimulation train and the Python script controlling HF output. To identify the specific LF-VNS pulses corresponding to the HF start and end times, we developed a simple algorithm which was run on the raw, unfiltered eCAP recordings where the 10 kHz artifact was most readily apparent. First, the rolling standard deviation (window size = 250 samples, ~100 ms) was calculated from the raw signal following each LF-VNS pulse. The window size of 250 samples was selected to flatten the standard deviation calculated around the relatively brief LF-VNS stimulation artifact whereas the constant 10 kHz artifact after HF started was unaffected. Next, we compared the maximum value of the rolling standard deviation array for each pulse to a threshold calculated as 3.5X the mean of the standard deviation during the starting portion of the pre-HF period of LF-VNS stimulation which excluded pulses that could have possibly had an HF start(e.g. the first 15 s of the pre-HF period when the delay for HF starting was 20 s). Pulses without the HF artifact had a low maximum value as both transient artifacts and slow drift were minimized by the rolling window, whereas pulses with HF had a high maximum value corresponding to a 250-sample window containing the 10 kHz HF artifact.

EMG_CA Analysis

After identifying the start and end of the HF signal in each trial, the filtered EMG data for a given LF-VNS/HF amplitude combination was sub-divided into pre-HF LF-VNS, during HF (LF-VNS + 10 kHz HF), and post-HF LF-VNS periods (Fig. 1B). For quantifying the effect of HF on evoked muscle activity, the root-mean-square (V_rms) of the stimulus-triggered average EMG_CA signal from each period was calculated with integration windows set to encompass the evoked muscle activity. This window was determined for each pig separately by plotting the EMG_CA for each muscle at the maximum LF-VNS amplitude (1 mA) tested in the EMG trials and selecting a window which excluded the LF-VNS stimulation artifact while fully encompassing the muscle response. We selected this approach rather than using the full EMG_CA recording to isolate the evoked response from activity induced by the HF signal (see HF Onset Effects). The V_rms of the during-HF and post-HF periods were normalized (Ṽ_rms) within trial using the V_rms of the pre-HF period to account for subject-to-subject anatomical and experimental preparation differences using the Eq. (1):

$${\widetilde{V}}_{{rms}}=\frac{{during\; HF}\,{V}_{{rms}}}{{pre\; HF}\,{V}_{{rms}}}$$

(1)

eCAP Analysis

The eCAP analyses were performed in a similar manner to the EMG by separating data into pre-HF, during-HF, and post-HF periods. In some subjects (n = 3/5), a large capacitive stimulus artifact was present at high amplitudes that persisted into the eCAP signal. To remove this artifact, the eCAP recordings in these subjects were re-referenced against another channel on the recording array post-hoc. The V_rms of stimulus-triggered averages from each HF period were then used to quantify the effects of HF on neural conduction with integration windows selected based on individual eCAP components. Integration windows were manually selected based on the phases/lobes of the eCAP signal. The V_rms of the during-HF and post-HF periods were normalized within trial against the pre-HF period and used to generate dose-response curves as in the EMG analysis.

eCAP C3 was normalized against the lowest HF amplitude tested (which still completely blocked EMG) during the high LF-VNS amplitude cardiac trials rather than the pre-HF period of the same stimulation trial (as in eCAP C1/C2). To address potential EMG contamination of eCAP C3 two approaches were taken. First, the number and latencies of eCAP components were characterized using eCAP recordings taken under paralytic (Supplementary Fig. 5) allowing us to confirm that eCAP C3 was neural in origin by removal of EMG signal sources. This allowed clear identification of eCAP C3 during the ‘+10 kHz HF’ periods in trials without paralytic if the HF amplitude was above the EMG block threshold, as identified by the absence of evoked EMG_CA during HF. Secondly, in some subjects, high LF-VNS amplitudes activated the superior laryngeal (SL) nerve which is located cranial to the location of the LF-VNS cuff and innervates both the CA and CT muscles. Activation of this pathway as opposed to the standard conduction path (via the recurrent laryngeal nerve) is distinguishable by the presence of a shorter latency EMG component⁵. As this pathway is due to current escape cranial to the LF-VNS cuff and not propagation down the VN through the HF cuff, the application of HF would not be expected to block muscle activation from the SL and EMG contamination of the eCAP signal could not be ruled out. Thus, eCAP C3 data were excluded from analyses when this short-latency component was present in the eCAP (P01, P04).

HR (bpm) Analysis

A stimulus-triggered TTL-pulse provided synchronization of LF-VNS with the physiologic recordings from the ADI system. HR was calculated from the raw ECG, pulse, and blood pressure (in pigs where arterial BP was collected via femoral catheter) using the built-in ‘Rate’ functions of ADI’s LabChart software. The ECG-derived HR was used for analysis except in subject P02 where the HF signal introduced a significant artifact into the ECG which interfered with the ADI’s rate calculation. To account for differential HR responses across subjects to LF-VNS (i.e. in subjects exhibiting larger bradycardic responses) we assessed the effect of HF on the LF-VNS response by normalizing the change in HR of the during-HF period by the pre-HF period using the Eq. (2):

$${HF\; Normalized\; dHR}=\frac{({{HR}}_{{BL}}-{{HR}}_{{HF}})}{({{HR}}_{{BL}}-{{HR}}_{{pre}-{HF}})}=\frac{{{dHR}}_{{HF}}}{{{dHR}}_{{pre}-{HF}}}$$

(2)

Where dHR_pre-HF is the maximum deflection from baseline (HR_BL) over the last 10 s of the pre-HF, LF-VNS only period and dHR_HF is the maximum deflection from baseline of the last 10 s during-HF. Baseline was defined as the mean HR during the 10 s prior to initiation of the pre-HF, LF-VNS pulse-train to account for inherent baseline drift in HR over long durations of anesthesia. The last 10 s of each period was used to allow the respective LF-VNS and HF cardiac effects to fully develop given the slower response time of HR to LF-VNS/HF signals. Using this method, a value close to 1 indicates that the HF had little to no effect on the LF-VNS-evoked cardiac effect whereas a value closer to 0 means that the HF signal returned the HR to near baseline, and a negative value indicates the HR during HF was higher than baseline.

HF Onset Effects

To quantify HF onset effects, we analyzed the entire EMG_CA signal during HF. We only analyzed onset responses using EMG data, not eCAP data, because (1) the magnitude of the HF artifact relative to the EMG signal was much smaller and easier to filter, and (2) HF evokes asynchronous activity which is not amenable to stimulus-triggered averaging for improvement of signal-to-noise ratio nor determination of conduction velocities (fiber-types) based on latency. For visualization, filtered EMG_CA data was plotted using a rasterized format where each row on the y-axis corresponds to a LF-VNS pulse, the x-axis represents time relative to the LF-VNS pulse, and the amplitude of the EMG was denoted by mapping the peak-to-peak EMG signal onto a color axis (Results Fig. 3B; Supplementary Fig. 8, 10). This approach enabled visualization of the pulse-by-pulse effect of HF and evoked asynchronous activity. We applied a simple peak finding algorithm (SciPy, ‘find-peaks’) with parameters set to detect peaks whose amplitude was >10% of the LF-VNS evoked EMG_CA signal. We then quantified the onset response as the total number of twitches during HF and their magnitude relative to the LF-VNS-evoked twitch (peak-to-peak voltage, not V_rms).

To compare the effects of HF on the LF-VNS evoked EMG_CA with the HF evoked asynchronous activity, a separate analysis was performed in which the EMG_CA signal following each LF-VNS pulse was sub-divided into LF-VNS and HF windows and V_rms calculated on a pulse-by-pulse basis (Results Fig. 3A, Supplementary Fig. 8). The LF-VNS window was the same as in the EMG_CA above with the remaining signal (again excluding the LF-VNS stimulation artifact) placed into the HF calculation window. Pulse numbers were converted to trial time with the calculated V_rms values averaged into ~0.5 second bins for plotting.

Statistical Tests

All statistical tests were performed in Python using the SciPy.stats library. For EMG and eCAP data, statistical tests were run on the normalized data from the HF period relative to the pre-HF LF-VNS evoked signal. To assess the effects of HF on the LF-VNS evoked EMG and eCAP, the non-parametric Kruskal-Wallis test was run on the normalized data (during-HF vs. pre-HF) from all subjects organized by HF amplitude. This test was followed by the non-parametric pairwise Dunn’s test where the lowest HF amplitude applied (0.5 mA for EMG_CA, eCAPs C1/C2, 4,6 mA for HR and eCAP C3) served as the control group against which every other HF amplitude was tested. HF amplitudes for cardiac and eCAP C3 control groups were based on a level sufficient to completely block the LF-VNS evoked EMG_CA (which would contaminate the eCAP C3 data) without affecting eCAP C3 (Fig. 2C, confirmed via vecuronium Supplementary Fig. 5). To account for multiple comparisons, adjusted p-values were calculated using the Holm-Sidak correction. Correlations between the eCAP and EMG, as well as V_rms of the LF-VNS and HF evoked EMG_CA was performed by calculating the non-parametric Spearman’s correlation coefficient using functions built into the SciPy.stats library.

Ethics

All experimental procedures were carried out following a protocol approved by the University of Wisconsin-Madison Institutional Animal Care and Use Committee (IACUC).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Raw data collected and analyzed in this study are publicly available via Dryad (https://doi.org/10.5061/dryad.1rn8pk17k). Specific source data for plots are provided with this paper in the Supplementary Information/Source Data file. Any additional requests for information can be directed to, and will be fulfilled by, the corresponding author. Source data are provided with this paper.

Code availability

Code for analysis, plotting, and statistical tests included in this paper are available through a GitHub repository (accessible via Zenodo: https://doi.org/10.5281/zenodo.17781109). This software is provided under the Creative Commons CC0 1.0 license.

References

Beekwilder, J. P. & Beems, T. Overview of the clinical applications of vagus nerve stimulation. J. Clin. Neurophysiol. 27, 130 (2010).
Article CAS PubMed Google Scholar
Koutsouras, D. A., Malliaras, G. G. & Langereis, G. The rise of bioelectronic medicine. Bioelectron. Med. 10, 19 (2024).
Article PubMed PubMed Central Google Scholar
Groves, D. A. & Brown, V. J. Vagal nerve stimulation: a review of its applications and potential mechanisms that mediate its clinical effects. Neurosci. Biobehav. Rev. 29, 493–500 (2005).
Article PubMed Google Scholar
Food and Drug Administration. Premarket Approval (PMA) P130019 MAESTRO RECHARGEABLE SYSTEM. FDA Website https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpma/pma.cfm?id=P130019.
Food and Drug Administration. Premarket Approval (PMA) P210007 Vivistim® System. FDA Website https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpma/pma.cfm?id=P210007.
Food and Drug Administration. Premarket Approval (PMA) P970003 VNS THERAPY SYSTEM. FDA Website https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpma/pma.cfm?id=P970003S050.
SetPoint Medical Corporation. Long Term Extension Study of the Safety and Efficacy of Neurostimulation Using a Vagus Nerve Stimulation Device in Patients With Rheumatoid Arthritis. https://clinicaltrials.gov/study/NCT04862117 (NCT, 2021).
Drewes, A. M. Treatment of Complications to Diabetic Autonomic Neuropathy With Vagus Nerve Stimulation. https://clinicaltrials.gov/study/NCT04143269 (NCT, 2023).
Boston Scientific Corporation. Neural Cardiac Therapy for Heart Failure Study. https://clinicaltrials.gov/study/NCT01385176 (NCT, 2024).
Câmara, R. & Griessenauer, C. J. Chapter 27 - Anatomy of the Vagus Nerve. in Nerves and Nerve Injuries (eds et al.) 385–397 (Academic Press, 2015).https://doi.org/10.1016/B978-0-12-410390-0.00028-7.
Pavlov, V. A. & Tracey, K. J. The vagus nerve and the inflammatory reflex—linking immunity and metabolism. Nat. Rev. Endocrinol. 8, 743–754 (2012).
Article CAS PubMed PubMed Central Google Scholar
Hoffman, H. H. & Schnitzlein, H. N. The numbers of nerve fibers in the vagus nerve of man. Anat. Rec. 139, 429–435 (1961).
Article CAS PubMed Google Scholar
Settell, M. L. et al. Functional vagotopy in the cervical vagus nerve of the domestic pig: implications for the study of vagus nerve stimulation. J. Neural Eng. 17, 026022 (2020).
Article ADS PubMed PubMed Central Google Scholar
Nicolai, E. N. et al. Sources of off-target effects of vagus nerve stimulation using the helical clinical lead in domestic pigs. J. Neural Eng. 17, 046017 (2020).
Article ADS PubMed PubMed Central Google Scholar
Blanz, S. L. et al. Spatially selective stimulation of the pig vagus nerve to modulate target effect versus side effect. J. Neural Eng. 20, 016051 (2023).
Article ADS Google Scholar
Musselman, E. D., Pelot, N. A. & Grill, W. M. Validated computational models predict vagus nerve stimulation thresholds in preclinical animals and humans. J. Neural Eng. 20, 036032 (2023).
Article ADS Google Scholar
Yoo, P. B. et al. High-resolution measurement of electrically-evoked vagus nerve activity in the anesthetized dog. J. Neural Eng. 10, 026003 (2013).
Article ADS PubMed Google Scholar
Vuckovic, A., Tosato, M. & Struijk, J. J. A comparative study of three techniques for diameter selective fiber activation in the vagal nerve: anodal block, depolarizing prepulses and slowly rising pulses. J. Neural Eng. 5, 275 (2008).
Article ADS PubMed Google Scholar
Fisher, B., DesMarteau, J. A., Koontz, E. H., Wilks, S. J. & Melamed, S. E. Responsive vagus nerve stimulation for drug resistant epilepsy: a review of new features and practical guidance for advanced practice providers. Front. Neurol. 11, 610379 (2020).
Article PubMed Google Scholar
Handforth, A. et al. Vagus nerve stimulation therapy for partial-onset seizures. Neurology 51, 48–55 (1998).
Article CAS PubMed Google Scholar
De Ferrari, G. M. et al. Long-term vagal stimulation for heart failure: Eighteen month results from the neural cardiac therapy for heart failure (NECTAR-HF) trial. Int. J. Cardiol. 244, 229–234 (2017).
Article PubMed Google Scholar
Chang, Y.-C. et al. Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers. Brain Stimul. 13, 1617–1630 (2020).
Article PubMed Google Scholar
Pelot, N. A. et al. Quantified morphology of the cervical and subdiaphragmatic vagus nerves of human, pig, and rat. Front. Neurosci. 14, 1148 (2020).
Article Google Scholar
Musselman, E. D., Pelot, N. A. & Grill, W. M. Empirically based guidelines for selecting vagus nerve stimulation parameters in epilepsy and heart failure. Cold Spring Harb. Perspect. Med. 9, a034264 (2019).
Article CAS PubMed PubMed Central Google Scholar
Ciotti, F. et al. Towards enhanced functionality of vagus neuroprostheses through in silico optimized stimulation. Nat. Commun. 15, 6119 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Aristovich, K. et al. Model-based geometrical optimisation and in vivo validation of a spatially selective multielectrode cuff array for vagus nerve neuromodulation. J. Neurosci. Methods 352, 109079 (2021).
Article CAS PubMed Google Scholar
Fitchett, A., Mastitskaya, S. & Aristovich, K. Selective neuromodulation of the vagus nerve. Front. Neurosci. 15, 685872 (2021).
Jayaprakash, N. et al. Organ- and function-specific anatomical organization of vagal fibers supports fascicular vagus nerve stimulation. Brain Stimul. 16, 484–506 (2023).
Article PubMed PubMed Central Google Scholar
Ahmed, U. et al. Strategies for precision vagus neuromodulation. Bioelectron. Med. 8, 9 (2022).
Article PubMed PubMed Central Google Scholar
Ahmed, U. et al. Anodal block permits directional vagus nerve stimulation. Sci. Rep. 10, 9221 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Wernisch, L. et al. Online Bayesian optimization of vagus nerve stimulation. J. Neural Eng. 21, 026019 (2024).
Article ADS Google Scholar
Berthon, A. et al. Using neural biomarkers to personalize dosing of vagus nerve stimulation. Bioelectron. Med. 10, 15 (2024).
Article PubMed PubMed Central Google Scholar
Tosato, M., Yoshida, K., Toft, E. & Struijk, J. J. Quasi-trapezoidal pulses to selectively block the activation of intrinsic laryngeal muscles during vagal nerve stimulation. J. Neural Eng. 4, 205 (2007).
Article ADS CAS PubMed Google Scholar
Austelle, C. W. et al. A comprehensive review of vagus nerve stimulation for depression. Neuromodulation Technol. Neural Interface 25, 309–315 (2022).
Article Google Scholar
Pérez-Carbonell, L., Faulkner, H., Higgins, S., Koutroumanidis, M. & Leschziner, G. Vagus nerve stimulation for drug-resistant epilepsy. Pract. Neurol. 20, 189–198 (2020).
Article PubMed Google Scholar
Thompson, N. et al. Organotopic organization of the porcine mid-cervical vagus nerve. Front. Neurosci. 17, 963503 (2023).
Pečlin, P. & Rozman, J. Alternative paradigm of selective vagus nerve stimulation tested on an isolated porcine vagus nerve. Sci. World J. 2014, 310283 (2014).
Article Google Scholar
Thompson, N. et al. Fascicular organisation and neuroanatomy of the porcine and human vagus nerves: allowing for spatially selective vagus nerve stimulation. FASEB J. 36, 3366 (2022).
Thompson, N. et al. Towards spatially selective efferent neuromodulation: anatomical and functional organization of cardiac fibres in the porcine cervical vagus nerve. J. Physiol. 603, 1983–2004 (2025).
Chang, Y.-C. et al. kHz-frequency electrical stimulation selectively activates small, unmyelinated vagus afferents. Brain Stimul. 15, 1389–1404 (2022).
Article PubMed PubMed Central Google Scholar
Nielsen, T. N., Struijk, J. J. & Sevcencu, C. Stimulation Waveforms for the Selective Activation of Baroreceptor Nerve Fibers in the Cervical Vagus Nerve. in Converging Clinical and Engineering Research on Neurorehabilitation II (eds Ibáñez, J., González-Vargas, J., Azorín, J. M., Akay, M. & Pons, J. L.) 995–999 (Springer International Publishing, 2017) https://doi.org/10.1007/978-3-319-46669-9_161.
Muzquiz, M. I. et al. In-vivo application of low frequency alternating currents on porcine cervical vagus nerve evokes reversible nerve conduction block. Bioelectron. Med. 7, 9 (2021).
Article PubMed PubMed Central Google Scholar
Plachta, D. T. T. et al. Blood pressure control with selective vagal nerve stimulation and minimal side effects. J. Neural Eng. 11, 036011 (2014).
Article ADS PubMed Google Scholar
Alhawwash, A. et al. In vivo peripheral nerve activation using sinusoidal low-frequency alternating currents. Artif. Organs 46, 2055–2065 (2022).
Article PubMed PubMed Central Google Scholar
Camilleri, M. et al. Intra-abdominal vagal blocking (VBLOC therapy): Clinical results with a new implantable medical device. Surgery 143, 723–731 (2008).
Article CAS PubMed Google Scholar
Peña, E., Pelot, N. A. & Grill, W. M. Quantitative comparisons of block thresholds and onset responses for charge-balanced kilohertz frequency waveforms. J. Neural Eng. 17, 046048 (2020).
Article ADS PubMed PubMed Central Google Scholar
Pelot, N. A. & Grill, W. M. In vivo quantification of excitation and kilohertz frequency block of the rat vagus nerve. J. Neural Eng. 17, 026005 (2020).
Article ADS CAS PubMed Google Scholar
Gerges, M. et al. Frequency- and amplitude-transitioned waveforms mitigate the onset response in high-frequency nerve block. J. Neural Eng. 7, 066003 (2010).
Article ADS PubMed PubMed Central Google Scholar
Ackermann, D. M., Bhadra, N., Gerges, M. & Thomas, P. J. Dynamics and sensitivity analysis of high-frequency conduction block. J. Neural Eng. 8, 065007 (2011).
Article ADS PubMed PubMed Central Google Scholar
Kilgore, K. L. & Bhadra, N. Nerve conduction block utilising high-frequency alternating current. Med. Biol. Eng. Comput. 42, 394–406 (2004).
Article CAS PubMed Google Scholar
Peña, E., Pelot, N. A. & Grill, W. M. Non-monotonic kilohertz frequency neural block thresholds arise from amplitude- and frequency-dependent charge imbalance. Sci. Rep. 11, 5077 (2021).
Article ADS PubMed PubMed Central Google Scholar
Peña, E., Pelot, N. A. & Grill, W. M. Spatiotemporal parameters for energy efficient kilohertz-frequency nerve block with low onset response. J. Neuroeng. Rehabil. 20, 72 (2023).
Article PubMed PubMed Central Google Scholar
Pelot, N. A., Behrend, C. E. & Grill, W. M. Modeling the response of small myelinated axons in a compound nerve to kilohertz frequency signals. J. Neural Eng. 14, 046022 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Avendaño-Coy, J., Serrano-Muñoz, D., Taylor, J. & Goicoechea-García, C. & Gómez-Soriano, J. peripheral nerve conduction block by high-frequency alternating currents: a systematic review. IEEE Trans. Neural Syst. Rehabil. Eng. 26, 1131–1140 (2018).
Article PubMed Google Scholar
Kapural, L. et al. Comparison of 10-kHz high-frequency and traditional low-frequency spinal cord stimulation for the treatment of chronic back and leg pain: 24-month results from a multicenter, randomized, controlled pivotal trial. Neurosurgery 79, 667 (2016).
Article PubMed Google Scholar
Couto, J. & Grill, W. M. Kilohertz frequency deep brain stimulation is ineffective at regularizing the firing of model thalamic neurons. Front. Comput. Neurosci. 10, 22 (2016).
Cogan, S. F., Hara, S. & Ludwig, K. A. Chapter 7 - The Safe Delivery of Electrical Currents and Neuromodulation. in Neuromodulation (Second Edition) (eds Krames, E. S., Peckham, P. H. & Rezai, A. R.) 83–94 (Academic Press) https://doi.org/10.1016/B978-0-12-805353-9.00007-3 (2018).
Verma, N. et al. Microneurography as a minimally invasive method to assess target engagement during neuromodulation. J. Neural Eng. 20, 026036 (2023).
Article ADS Google Scholar
Deshmukh, A. et al. Epidural Spinal Cord Recordings (ESRs): sources of neural-appearing artifact in stimulation evoked compound action potentials. J. Neural Eng. https://doi.org/10.1088/1741-2552/ad7f8b (2024).
Nicolai, E. N. et al. Vagus nerve stimulation using an endovascular electrode array. J. Neural Eng. 20, 046008 (2023).
Article ADS Google Scholar
Gasser, H. S. & Erlanger, J. The role played by the sizes of the constituent fibers of a nerve trunk in determining the form of its action potential wave. Am. J. Physiol. -Leg. Content 80, 522–547 (1927).
Article Google Scholar
Woo, M. Y. & Campbell, B. Asynchronous firing and block of peripheral nerve conduction by 20 kc alternating current. Bull. Los Angel Neurol. Soc. 29, 87–94 (1964).
CAS Google Scholar
Franke, M. et al. Combined KHFAC + DC nerve block without onset or reduced nerve conductivity after block. J. Neural Eng. 11, 056012 (2014).
Article ADS PubMed PubMed Central Google Scholar
Patel, Y. A. & Butera, R. J. Challenges associated with nerve conduction block using kilohertz electrical stimulation. J. Neural Eng. 15, 031002 (2018).
Article ADS PubMed Google Scholar
Bowman, B. R. & McNeal, D. R. Response of single alpha motoneurons to high-frequency pulse trains: firing behavior and conduction block phenomenon. Appl. Neurophysiol. 49, 121–138 (1987).
Google Scholar
Crosby, N. D., Janik, J. J. & Grill, W. M. Modulation of activity and conduction in single dorsal column axons by kilohertz-frequency spinal cord stimulation. J. Neurophysiol. 117, 136–147 (2016).
Article PubMed PubMed Central Google Scholar
Waataja, J. J., Honda, C. N., Asp, A. J., Nihilani, R. K. & Farajidavar, A. The duration and intensity of high frequency alternating current influences the degree and recovery of nerve conduction block. IEEE Trans. Biomed. Eng. 71, 2170–2179 (2024).
Article PubMed Google Scholar
ANTHEM-HFrEF Fails to Show Benefit of Vagal Nerve Stimulation Over GDMT. TCTMD.com https://www.tctmd.com/news/anthem-hfref-fails-show-benefit-vagal-nerve-stimulation-over-gdmt (TCTMD, 2023).
School, M. M. Preliminary Findings from LivaNova’s RECOVER Clinical Study on VNS Therapy for Treatment-Resistant Depression. Louis A. Faillace, MD, Department of Psychiatry and Behavioral Sciences https://med.uth.edu/psychiatry/2024/06/17/preliminary-findings-from-livanovas-recover-clinical-study-on-vns-therapy-for-treatment-resistant-depression/ (DPBS, 2024).
Pahuja, M. et al. Neuromodulation therapies in heart failure: a state-of-the-art review. J. Soc. Cardiovasc. Angiogr. Interv. 2, 101199 (2023).
Anand, I. S. et al. Comparison of symptomatic and functional responses to vagus nerve stimulation in ANTHEM-HF, INOVATE-HF, and NECTAR-HF. ESC Heart Fail 7, 75–83 (2020).
PubMed PubMed Central Google Scholar
Ardell, J. L. et al. Defining the neural fulcrum for chronic vagus nerve stimulation: implications for integrated cardiac control. J. Physiol. 595, 6887–6903 (2017).
Article CAS PubMed PubMed Central Google Scholar
Song, Z., Viisanen, H., Meyerson, B. A., Pertovaara, A. & Linderoth, B. Efficacy of kilohertz-frequency and conventional spinal cord stimulation in rat models of different pain conditions. Neuromodulation Technol. Neural Interface 17, 226–235 (2014).
Article Google Scholar
Lee, K. Y. et al. Low-intensity, kilohertz frequency spinal cord stimulation differently affects excitatory and inhibitory neurons in the rodent superficial dorsal horn. Neuroscience 428, 132–139 (2020).
Article CAS PubMed Google Scholar
Dai, F., Yin, J. & Chen, J. D. Z. Effects and mechanisms of vagal nerve stimulation on body weight in diet-induced obese rats. Obes. Surg. 30, 948–956 (2020).
Article PubMed Google Scholar
Wang, D., Lee, K. Y., Lee, D., Kagan, Z. B. & Bradley, K. Low-intensity 10 kHz spinal cord stimulation reduces behavioral and neural hypersensitivity in a rat model of painful diabetic neuropathy. J. Pain. Res. 15, 1503–1513 (2022).
Article PubMed PubMed Central Google Scholar
Musselman, E. D., Raha, I., Pelot, N. A. & Grill, W. M. Scaling of vagus nerve stimulation parameters does not achieve equivalent nerve responses across species. Bioelectron. Med. 11, 11 (2025).
Article PubMed PubMed Central Google Scholar
Bender, S. A. et al. Closed-Loop Multimodal Neuromodulation of Vagus Nerve for Control of Heart Rate*. in 2024 American Control Conference (ACC) 4536–4541 https://doi.org/10.23919/ACC60939.2024.10644421. (ACC, 2024).
Ackermann, D. M. Jr. et al. Electrical conduction block in large nerves: high-frequency current delivery in the nonhuman primate. Muscle Nerve 43, 897–899 (2011).
Article PubMed PubMed Central Google Scholar
Knutson, J. S. et al. First-in-human demonstration of high-frequency electrical motor nerve block: case report. Am. J. Phys. Med. Rehabil. 104, e37 (2025).
Article PubMed Google Scholar
Fjordbakk, C. T. et al. Feasibility of kilohertz frequency alternating current neuromodulation of carotid sinus nerve activity in the pig. Sci. Rep. 9, 18136 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
Fisher, K. M., Jillani, N. E., Oluoch, G. O. & Baker, S. N. Blocking central pathways in the primate motor system using high-frequency sinusoidal current. J. Neurophysiol. 113, 1670–1680 (2015).
Article PubMed Google Scholar
Kilgore, K. L. & Bhadra, N. Reversible nerve conduction block using kilohertz frequency alternating current. Neuromodulation J. Int. Neuromodulation Soc. 17, 242–255 (2014).
Article Google Scholar
Bhadra, N., Foldes, E., Vrabec, T., Kilgore, K. & Bhadra, N. Temporary persistence of conduction block after prolonged kilohertz frequency alternating current on rat sciatic nerve. J. Neural Eng. 15, 016012 (2018).
Article ADS PubMed PubMed Central Google Scholar
Bhadra, N., Lahowetz, E. A., Foldes, S. T. & Kilgore, K. L. Simulation of high-frequency sinusoidal electrical block of mammalian myelinated axons. J. Comput. Neurosci. 22, 313–326 (2007).
Article MathSciNet PubMed Google Scholar
Hernandez, C. M. & Richards, J. R. Physiology, Sodium Channels. in StatPearls (StatPearls Publishing, 2024).
Miles, J. D., Kilgore, K. L., Bhadra, N. & Lahowetz, E. A. Effects of ramped amplitude waveforms on the onset response of high-frequency mammalian nerve block. J. Neural Eng. 4, 390 (2007).
Article ADS CAS PubMed Google Scholar
Rosenblueth et al. Of alternating currents on nerve. Am. J. Physiol. -Leg. Content 125, 251–264 (1939).
Article Google Scholar
Vargas, L., Musselman, E. D., Grill, W. M. & Hu, X. Asynchronous axonal firing patterns evoked via continuous subthreshold kilohertz stimulation. J. Neural Eng. 20, 026015 (2023).
Article ADS Google Scholar
van den Honert, C. & Mortimer, J. T. A technique for collision block of peripheral nerve: single stimulus analysis. IEEE Trans. Biomed. Eng. 28, 373–378 (1981).
Article PubMed Google Scholar
Okonogi, T. et al. Stress-induced vagal activity influences anxiety-relevant prefrontal and amygdala neuronal oscillations in male mice. Nat. Commun. 15, 183 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Conway, C. R. & Xiong, W. The mechanism of action of vagus nerve stimulation in treatment-resistant depression: current conceptualizations. Psychiatr. Clin. North Am. 41, 395–407 (2018).
Article PubMed Google Scholar
Lai, J. & David, S. V. Short-term effects of vagus nerve stimulation on learning and evoked activity in auditory cortex. eNeuro 8, 0522 (2021).
Loerwald, K. W., Borland, M. S., Rennaker, R. L., Hays, S. A. & Kilgard, M. P. The interaction of pulse width and current intensity on the extent of cortical plasticity evoked by vagus nerve stimulation. Brain Stimul. 11, 271–277 (2018).
Article PubMed Google Scholar
Borland, M. S. et al. Cortical map plasticity as a function of vagus nerve stimulation intensity. Brain Stimul. 9, 117–123 (2016).
Article CAS PubMed Google Scholar
The pathophysiology of spasticity - Sheean - European Journal of Neurology - Wiley Online Library. https://onlinelibrary.wiley.com/doi/full/10.1046/j.1468-1331.2002.0090s1003.x (2002).
Marom, S. & Marder, E. A biophysical perspective on the resilience of neuronal excitability across timescales. Nat. Rev. Neurosci. 24, 640–652 (2023).
Article CAS PubMed Google Scholar
Persson, A.-K., Hoeijmakers, J. G. J., Estacion, M., Black, J. A. & Waxman, S. G. Sodium channels, mitochondria, and axonal degeneration in peripheral neuropathy. Trends Mol. Med. 22, 377–390 (2016).
Article CAS PubMed Google Scholar
Taylor, A. L., Goaillard, J.-M. & Marder, E. How multiple conductances determine electrophysiological properties in a multicompartment model. J. Neurosci. 29, 5573–5586 (2009).
Article CAS PubMed PubMed Central Google Scholar
Rosario, J. S. D. et al. Sustained depolarization induces homeostatic plasticity in mouse and human sensory neurons. J. Pain. 23, 6 (2022).
Article Google Scholar
Kapural, L. et al. Novel 10-kHz high-frequency therapy (HF10 Therapy) is superior to traditional low-frequency spinal cord stimulation for the treatment of chronic back and leg pain: the SENZA-RCT randomized controlled trial. Anesthesiology 123, 851–860 (2015).
Article PubMed Google Scholar
Apovian, C. M. et al. Two-year outcomes of vagal nerve blocking (vBloc) for the treatment of obesity in the recharge trial. Obes. Surg. 27, 169–176 (2017).
Article PubMed Google Scholar
Franke, M., Bhadra, N., Bhadra, N. & Kilgore, K. Direct current contamination of kilohertz frequency alternating current waveforms. J. Neurosci. Methods 232, 74–83 (2014).
Article PubMed PubMed Central Google Scholar
Shannon, R. V. A model of safe levels for electrical stimulation. IEEE Trans. Biomed. Eng. 39, 424–426 (1992).
Article CAS PubMed Google Scholar
Cogan, S. F., Ludwig, K. A., Welle, C. G. & Takmakov, P. Tissue damage thresholds during therapeutic electrical stimulation. J. Neural Eng. 13, 021001 (2016).
Article ADS PubMed PubMed Central Google Scholar
Wilks, S. J. et al. Non-clinical and pre-clinical testing to demonstrate safety of the barostim neo electrode for activation of carotid baroreceptors in chronic human implants. Front. Neurosci. 11, 438 (2017).
Li, Y. et al. NeurostimML: A machine learning model for predicting neurostimulation-induced tissue damage. bioRxiv 2023.10.18.562980 https://doi.org/10.1101/2023.10.18.562980 (2023).
Ghazavi, A. & Cogan, S. F. Electrochemical characterization of high frequency stimulation electrodes: role of electrode material and stimulation parameters on electrode polarization. J. Neural Eng. 15, 036023 (2018).
Article ADS PubMed Google Scholar
Ikramuddin, S. et al. Effect of reversible intermittent intra-abdominal vagal nerve blockade on morbid obesity: the ReCharge randomized clinical trial. JAMA 312, 915–922 (2014).
Article CAS PubMed Google Scholar
Kapural, L., Syed Shah, N., Fang, Z.-P. & Mekhail, N. Multicenter, double-blinded, randomized, active-sham controlled clinical study design to assess the safety and effectiveness of a novel high frequency electric nerve block system in the treatment of post-amputation pain (The QUEST Study). J. Pain. Res. 15, 1623–1631 (2022).
Article PubMed PubMed Central Google Scholar
Premarket Approval (PMA). Altius® Direct Electrical Nerve Stimulation System. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpma/pma.cfm?id=P230020.
Premarket Approval (PMA) P130022 NEVRO SENZA SPINAL CORD STIMULATION (SCS) SYSTEM. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpma/pma.cfm?id=P130022.
Johannessen, H. et al. Vagal blocking for obesity control: a possible mechanism-of-action. Obes. Surg. 27, 177–185 (2017).
Article PubMed Google Scholar
Finch, P., Price, L. & Drummond, P. High-frequency (10kHz) electrical stimulation of peripheral nerves for treating chronic pain: a double-blind trial of presence vs absence of stimulation. Neuromodulation Technol. Neural Interface 22, 529–536 (2019).
Article Google Scholar
Shikora, S. A. et al. Sustained weight loss with vagal nerve blockade but not with sham: 18-month results of the recharge trial. J. Obes. 2015, 365604 (2015).
Article PubMed PubMed Central Google Scholar
Khalid, et al. Altius electrical nerve stimulation for post-amputation pain treatment: a comprehensive review. Pain Manag. 15, 183–189 (2025).
Waataja, J. J., Tweden, K. S. & Honda, C. N. Effects of high-frequency alternating current on axonal conduction through the vagus nerve. J. Neural Eng. 8, 056013 (2011).
Article ADS PubMed Google Scholar
Rogers, E. R., Mirzakhalili, E. & Lempka, S. F. Model-based analysis of subthreshold mechanisms of spinal cord stimulation for pain. J. Neural Eng. 20, 066003 (2023).
Article ADS PubMed Central Google Scholar
Cheng, K., Trevathan, J. & Ludwig, K. Ludwig_HFBlock_NComms_2025: HF Block NComms 2025 code v1.0 (v1.0). Zenodo. https://doi.org/10.5281/zenodo.17781109 (2025).

Download references

Acknowledgements

This work was supported through NIH R01 NS126376-01 (KPC, KAL, NAP, WMG). We would like to thank the staff of the University of Wisconsin’s Center for Biomedical Swine Research & Innovation (CSBRI) and Research Animal Resources and Compliance (RARC) anesthesiology unit, especially Liz Clem and Devon Klipsic, for their operational and veterinary support during these experiments.

Author information

Authors and Affiliations

University of Wisconsin – Madison, Madison, WI, USA
Kevin P. Cheng, Ashlesha Deshmukh, Aaron K. Gholston, Megan L. Settell, Bruce E. Knudsen, Maria LaLuzerne, James K. Trevathan, Aaron J. Suminski & Kip A. Ludwig
Duke University, Durham, NC, USA
Nicole A. Pelot & Warren M. Grill

Authors

Kevin P. Cheng
View author publications
Search author on:PubMed Google Scholar
Ashlesha Deshmukh
View author publications
Search author on:PubMed Google Scholar
Aaron K. Gholston
View author publications
Search author on:PubMed Google Scholar
Megan L. Settell
View author publications
Search author on:PubMed Google Scholar
Bruce E. Knudsen
View author publications
Search author on:PubMed Google Scholar
Maria LaLuzerne
View author publications
Search author on:PubMed Google Scholar
James K. Trevathan
View author publications
Search author on:PubMed Google Scholar
Aaron J. Suminski
View author publications
Search author on:PubMed Google Scholar
Nicole A. Pelot
View author publications
Search author on:PubMed Google Scholar
Warren M. Grill
View author publications
Search author on:PubMed Google Scholar
Kip A. Ludwig
View author publications
Search author on:PubMed Google Scholar

Contributions

N.A.P., W.M.G., and K.A.L. conceived the study and provided guidance on experimental design. K.P.C. developed the experimental protocol, analyzed data, developed data analysis software, and prepared the manuscript. K.P.C., A.D., A.K.G., M.L.S., B.E.K., and M.L. carried out experimental procedures and raw data collection. A.K.G. and J.K.T. assisted with testing, programming, and validation of HF stimulation hardware. K.P.C., A.D., A.K.G., A.J.S., N.A.P., W.M.G., and K.A.L. reviewed and edited the manuscript and addressed peer reviewer comments.

Corresponding author

Correspondence to Kip A. Ludwig.

Ethics declarations

Competing interests

KAL is on the Scientific Advisory Board of LivaNova PLC, which markets clinical vagus nerve implantable devices. WMG also serves on the Scientific Advisory Board of LivaNova PLC, has received grant support from iVEAcare and LivaNova, and is an inventor on pending patent applications on systems and methods for nerve block. KPC, JKT, and KAL are co-inventors on patent application 19/382419: ELECTRODE FOR CLOSED-LOOP ACTIVATION OF NERVE FIBERS which incorporates high-frequency waveforms into the stimulation and feedback control methods in a closed-loop system. The remaining authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Kirill Aristovich, who co-reviewed with Nicole Thompson and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. [A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Reporting Summary

Transparent Peer Review file

Source data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Cheng, K.P., Deshmukh, A., Gholston, A.K. et al. Application of kilohertz-frequency block to mitigate off-target motor effects of vagus nerve stimulation in swine. Nat Commun 17, 1066 (2026). https://doi.org/10.1038/s41467-025-67823-0

Download citation

Received: 04 April 2025
Accepted: 09 December 2025
Published: 30 December 2025
Version of record: 28 January 2026
DOI: https://doi.org/10.1038/s41467-025-67823-0

Subjects

Abstract

Similar content being viewed by others

Non-monotonic kilohertz frequency neural block thresholds arise from amplitude- and frequency-dependent charge imbalance

Control of spatiotemporal activation of organ-specific fibers in the swine vagus nerve by intermittent interferential current stimulation

Vagus nerve stimulation using a miniaturized wirelessly powered stimulator in pigs

Introduction

Results

10 kHz HF blocks VNS-evoked motor side effects

Titrating HF amplitude blocks individual eCAP components

Onset EMG effects scale with HF amplitude

10 kHz HF effects on the LF-VNS cardiac response

HF can selectively block LF-VNS-evoked EMG without influencing LF-VNS cardiac responses

Discussion

Methods

Hardware and Instrumentation

Experimental Protocol

Data Analysis

Identification of HF Start and End

EMGCA Analysis

eCAP Analysis

HR (bpm) Analysis

HF Onset Effects

Statistical Tests

Ethics

Reporting summary

Data availability

Code availability

References

Acknowledgements

Author information

Authors and Affiliations

Contributions

Corresponding author

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Supplementary information

Supplementary Information

Reporting Summary

Transparent Peer Review file

Source data

Source data

Rights and permissions

About this article

Cite this article

Share this article

Search

Quick links

EMG_CA Analysis