Cutaneous Sensation of Electrical Stimulation Waveforms

Abstract

Background:

Skin sensation is the primary factor limiting the intensity of transcranial electrical stimulation (tES). It is well established that different waveforms generate different sensations, yet transcranial stimulation has been limited to a relatively small number of prototypical waveforms.

Objective:

We explore whether alternative stimulation waveforms could substantially reduce skin sensation and thus allow for stronger intensities in tES.

Methods:

We systematically tested a range of waveforms in a series of 6 exploratory experiments stimulating human adults on the forearm and in one instance on the head. Subjects were asked to rate skin sensation level on a numerical scale from “none” to “extreme”.

Results:

High frequency (>1 kHz) monophasic square wave stimulation was found to decrease in sensation with increasing duty cycle, baseline, and frequency, but the sensation was never lower than for constant current stimulation. For the purpose of injecting a net direct current (DC), a constant current is optimal. For stimulation with alternating current (AC), sensation decreased with increasing frequency, consistent with previous reports. Amplitude modulation did not reduce sensation below stimulation with constant AC amplitude, and biphasic square waveforms produced higher sensation levels than biphasic sinusoidal waveforms. Furthermore, for DC stimulation, sensation levels on the arm were similar to those reported on the head.

Conclusion:

Our comparisons of various waveforms for monophasic and biphasic stimulation indicate that conventional DC and AC waveforms may provide the lowest skin sensations levels for transcutaneous electrical stimulation. These results are likely generalizable to tES applications.

Introduction

Researchers have long explored the cognitive and behavioral effects of transcranial direct current stimulation (tDCS) on human subjects, typically passing weak currents of up to 2 mA through the scalp (1). tDCS studies have shown neurophysiological and cognitive changes in the stimulated areas of the brain and tested the effects on memory and learning (2,3). However, the reproducibility of tDCS results has been called into question (4,5), and is subject of persistent debate (6,7). One of several possible explanations is that field intensities of typical human experiments are too small to cause reliable effects. Computational models of electric fields, as well as intracranial recordings in humans, suggest that 2mA generates fields below 1V/m (8,9). Transcranial alternating currents stimulation (tACS) amplitudes have also been typically limited to 1 mA (10,11), yet in vitro and invasive animal work indicate that an electric field of less than 1 V/m may not be sufficient to produce measurable neural effects (12). Nonetheless, a number of in vitro and in vivo animal studies show that electric fields at these threshold values suffice to entrain neuronal spiking (11,13).

It may be possible to improve reliability of tES results by increasing stimulation field intensity, which depends both on current intensity as well as electrode size. Close to the surface, smaller electrodes typically cause stronger electric fields (14), whereas larger electrodes cause stronger sensation (15,16). Here we will focus on current intensity while keeping electrode size constant. Increasing stimulation intensity would be a straightforward solution, but it is severely limited by undesirable skin sensations. Sensation thresholds of electrical stimulation on the skin are around 0.4 mA (17) and stimulation becomes painful at 3 mA with current methods (18). Some studies reduced skin sensation using local anesthetics (19), and more recent efforts have used adaptive protocols to reach up to 4mA (20,21). Besides intensity and duration, other parameters of the stimulation protocol are known to influence skin sensation, such as electrode composition, electrode geometry, and the conductive medium used at the electrodes (15). Another important factor that affects skin sensation in transcranial stimulation is the stimulation waveform. There is therefore an interest in exploring alternative waveforms to reduce sensation and allow stronger stimulation intensities.

The waveform options for brain stimulation have largely remained limited to conventional constant-current, sinusoidal, and random-noise waveforms (tDCS, tACS and tRNS respectively). tRNS and low-frequency tACS have lower perceived sensation levels than tDCS (17,22). Sensations with tACS are non-monotonic with frequency in the range of 5 Hz to 10 kHz with a maximum at 50 Hz (23), similar to the pattern from transcutaneous stimulation studies (discussed above). Recent experimental work argued that DC or AC stimulation, when applied with a pulsed carrier at high frequencies (>1 kHz) may reduce sensation (24,25). The use of low-frequency (0.8 Hz) monophasic sinusoidal tDCS has also been proposed, with no difference in comfort from constant current (26). Another novel technique is to use temporal interference of sinusoidal waves to modulate deep targets (27,28), delivered through high carrier frequencies (>1 kHz) amplitude-modulated in low frequency (5–40 Hz) envelopes. Therefore, it is possible that other monophasic waveforms for passing a net current through the head may yield lower levels of sensation for better comfort during tDCS. tACS may also benefit from the use of waveforms other than sinusoids. Of course, different stimulation waveforms are likely to have different physiological effects and therefore differentially affect skin sensations.

The purpose of the present work is to quantify skin sensation as a function of various waveform parameters that have been recently proposed for transcranial stimulation (17,25,27,28). We aim to reduce skin sensations of electrical stimulation by using high-frequency carrier waveforms based on the hypothesis that peripheral neurons do not respond effectively to high-frequency stimulation above 100 Hz (23,28). We test this in a series of 5 exploratory experiments using various waveforms including pulsed, DC, sinusoidal AC and modulated AC, with parameters motivated in each case by prior literature. Stimulation was tested on the forearms of human subjects for convenience and accessibility to determine relative differences in skin sensation for reference.

Materials and Methods

Participants

In total, 22 healthy adults (8 female, 14 male, age range = 18–52 years, mean ± SD = 29.0 ±7.23 years) participated in the study after providing written informed consent. In Experiment 6 described below, an additional 59 subjects participated in context of an ongoing parallel study on tDCS (27 female, 32 male, age range = 18–50 years, mean ± SD = 23.7 ± 5.82 years).Participants were screened for skin disorders, injury, allergies, and metallic implants, and asked to report any pre-existing skin sensations prior to the experiment (none were excluded in this screening). Not all were naive to electrical stimulation, and to establish a baseline familiarity with the general sensation of electrical stimulation, every participant underwent direct current stimulation prior to full testing. Participants were given the option to withdraw from the study at any point should the stimulation sensation become intolerable, but all assigned experiments were completed. Some participants volunteered for multiple experiments and were compensated accordingly.

Experimental Design

Multiple experiments were carried out to compare the within-subject sensation effects between waveforms of different shapes and parameters. Experiments 1–5 involving a single electrode pair on the arm are described here. Experiment 6 involving multiple electrodes comparing sensation on the arm and head is described below. Each participant in each experiment underwent the same single-blind, multiple measures procedure involving stimulation with a pseudo-random sequence of waveforms. A given waveform was repeated multiple times in the pseudo-random sequence so as to present it in different contexts of preceding and succeeding waveforms, but the same waveform was not repeated consecutively. Participants were aware of the presence of different waveforms, but were blinded to the number of unique waveforms in the experiment. Sensation comparisons were done within-subject because the perception of sensation is inherently subjective and variable between individuals. The same procedures were followed for all experiments, although total stimulation time varied (from approximately 20 to 40 minutes) according to the number of waveforms being compared, and different stimulators were used for DC and AC waveforms. Electrical current was delivered using a custom voltage-controlled current source with input from a digital-to-analog voltage output module (NI-9263, National Instruments, Austin, TX) through a data acquisition device (DAQ) (cDAQ-9171, National Instruments, Austin, TX) interfaced with a computer via USB connection. The experiment was operated using a MATLAB (MathWorks, Natick, MA) graphical user interface (GUI). Figure 1A shows the hardware connections. For the purpose of this study, stimulation was administered on the forearm to isolate peripheral sensation from all other nervous effects. Silver/silver chloride sintered ring high-definition electrodes (Soterix Medical, New York, NY) were attached to the nondominant forearm, above the dorsal side of the wrist and on the proximal end of the forearm, above the dorsal side of the ulna, as the anode and cathode, respectively (Figure 1A). At each stimulation site, the electrode was attached using an electrode holder on an elastic fastener (Soterix Medical, New York, NY), with conductive gel (SignaGel, Parker Laboratories, Fairfield, NJ) between the electrode surface and the exposed skin. To minimize discomfort in light of the high current intensity, the electrode placement was selected at locations as far as possible from major sensory nerves. Because the anterior side of the forearm is innervated by the medial and ulnar nerves, the sensation is likely less intense on the posterior side if the electrodes are placed away from the radial nerve (29), as shown in Figure 1A. The cathode was placed approximately in the midpoint between the radial and ulnar nerves, but direct stimulation of these nerves cannot necessarily be excluded.

Figure 1.

(A) Experimental set up for sensation testing. Participants were treated with a randomized sequence of electrical stimulation waveforms delivered through the forearm. The stimulation was self-administered by the subject through a computer graphical user interface (GUI). Following each trial, the GUI asked the participants to rate the sensation on a visual analog scale. The GUI was interfaced with a data acquisition device (DAQ) to a voltage-controlled current source stimulator. The anode and cathode were placed above the posterior side of the wrist and near the elbow, respectively, to avoid the radial nerve. (B) Distribution of sensation ratings over all 3260 trials compiled over 5 different experiments.

Procedure

The participant was seated in front of a desk with a computer monitor displaying the GUI used to conduct the experiment. The participant rested their forearm freely on the desk in a comfortable position. Following alcohol sterilization of the stimulation sites, the electrodes were attached and the participant was given the pre-stimulation sequence. Prior to the start of the stimulation protocol, skin impedance was monitored on the stimulation device, which visually indicated contact quality at the electrodes (similar to the SMARTScan feature by Soterix Medical, New York, NY). The experiment did not begin until the contact quality was optimal, i.e. below 35 kOhm. The participant was then asked to operate the GUI using a computer mouse and instructed to start the experiment from the GUI when they were ready. Once started, the program began stimulating by gradually increasing the current intensity over 2 to 2.5 seconds until the full intensity was reached and maintained for 5 to 6 seconds, after which the intensity was ramped back down to zero. A progress bar on the GUI indicated the stimulation status, changing in length according to the stimulation intensity. At the end of each stimulation trial, the GUI prompted the participant to rate the sensation level (including discomfort or pain) on a discrete rating scale from the options, “No Sensation”, “Mild”, “Moderate”, “Severe”, and “Extreme”, recorded as discrete values of 1, 2, 3, 4, and 5, respectively. The participant was allowed to repeat the trial if they were not sure of their initial response. After they rated the sensation level, they proceeded to the next trial of their own volition, repeating the process until the end of the waveform sequence (Figure 1A). Before the start of the evaluation session, each participant received DC stimulation with incrementally increasing current intensities starting at 0.5 mA up to 3 mA (7 seconds at each intensity with linear ramps of 2 seconds before and after). This pre-stimulation sequence also served to decrease skin impedance and optimize connectivity, which was recorded prior to the experiment (15).

Monophasic (DC) Stimulation

Various monophasic waveforms were compared against each other and against constant DC stimulation in Experiments 1, 2, and 3. Sensation dependence on monophasic square wave parameters was determined by changing the duty cycle, baseline current, and frequency while maintaining the same net current of 3 mA (Figure 2A). Here baseline current is defined as the positive offset of the “off” phase of the waveform from zero current. Duty cycle is defined as the percentage of time the waveform period is “on”, or active. The peak amplitude is determined by the duty cycle and baseline such that the mean amplitude of all waveforms equals 3 mA. Two experiments comparing monophasic square waves were conducted, both including the conventional constant current for a total of 10 waveforms each. Experiment 1 tested 5 kHz square waves with three duty cycle values: 10%, 20%, and 30%, against three baseline values: 0.5 mA, 1 mA, and 2 mA, as well as constant 3mA. Experiment 2 tested 20% duty cycle waveforms of three frequencies: 5 kHz, 10 kHz, and 20 kHz against the same baseline values as the previous, plus constant current.

Figure 2. Sensation of monophasic square waveforms.

(A) An example of a monophasic square wave. Parameters were selected to deliver net 3 mA DC, with independent duty cycle and baseline values. Peak amplitude was calculated from the independent values by setting the average of the waveform at 3 mA. (B) An overview of the waveforms used in Experiment 1 (excluding the 3mA constant current), showing how the waveform shape changes with different combinations of duty cycle and baseline values. All of these waveforms deliver an average current of 3mA over time. (C, D) Mean sensation level ratings for 3 mA monophasic square waveforms. The rating for 3 mA constant direct current is shown as a line. Error bars and shaded areas indicate 95% confidence intervals.

Monophasic sinusoidal waveforms of different frequencies were compared with each other in Experiment 3. Regular sine waves with 3 mA amplitude were offset by 3 mA to yield a net current of 3 mA. Frequencies of 1 kHz, 5 kHz, 10 kHz, 25 kHz, and 50 kHz were tested against a constant 3 mA current. Each stimulation condition was repeated 9, 9, and 5 times per session in Experiments 1, 2, and 3, respectively.

Biphasic (AC) Stimulation

Biphasic amplitude-modulated sinusoidal waveforms simulating interferential AC stimulation (27) were tested for sensation effects of different carrier frequencies and amplitudes (Figure 4) in Experiment 4. Modulation frequency was kept constant at 5 Hz for this experiment, while carrier frequency was tested at 100 Hz, 500 Hz, 1 kHz, and 2 kHz. The effects of carrier frequency were compared against the effects of peak-to-peak amplitudes at 0.5 mA, 1 mA, and 2 mA. Amplitudes above 2 mA were not tested due to discomfort at the lower carrier frequencies. In addition to sensation level ratings, the participants were also asked to specify after each trial any qualities of the sensation they felt, from the options of “Prick”, “Pressure”, “Tingle”, “Vibration”, and “Contraction” (Figure S1). They were allowed to select more than one option. “Prick” refers to the sensation of a static needle or other sharp object poking the skin; “Pressure” refers to the sensation of a concentrated force applied to the skin; “Tingle” refers to low-frequency repeated light pricking sensation; “Vibration” refers to the sensation of high-frequency light shaking; and “Contraction” refers to the sensation of the arm muscles flexing.

Figure 4. Sensation of AM sinusoidal stimulation.

(A) Biphasic amplitude-modulated sine wave. This waveform simulates the resulting electric field of two interfering AC waveforms as described by Grossman et al. (28). Carrier frequencies are in the kHz range, modulated at low frequencies in the 1–10 Hz range. (B) Mean sensation level ratings for biphasic sinusoidal waveforms amplitude-modulated at 5 Hz. Mean ratings are shown with 95% confidence intervals. Two-way repeated measures ANOVA found significant effects from carrier frequency and amplitude, as well as the interaction between the two factors.

To test the effects of different AC waveform shapes, in Experiment 5 we compared square waves, sinusoidal waves and amplitude-modulated sinusoidal waveforms (with modulation frequencies of 5 Hz and 10 Hz) at carrier frequencies of 500 Hz and 1 kHz, all at 3 mA maximum. In this experiment participants were also asked to rate sensation qualities. Each stimulation condition was repeated 5 and 7 times per session in Experiments 4 and 5, respectively.

Direct Comparisons with Transcranial Stimulation

In Experiment 6, participants were given stimulation on the forearm, following the protocol from our ongoing study using High Definition tDCS (HD-tDCS). The HD-tDCS experiment uses a 4+4 electrode montage targeted at the primary motor cortex, with 4 anode-cathode pairs each delivering 1 mA for a total intensity of 4 mA. Anodes are placed on fronto-central locations and cathodes on right-posterior locations of the 10–10 international system. In this recreation on the arm we placed 4 anode and 4 cathode electrodes spaced evenly around the circumference of the nondominant arm, with the anodes at the wrist and the cathodes toward the elbow, as shown in Figure 1A for the other forearm experiments. The electrodes were attached using the same elastic fasteners mentioned above, with all anodes on one fastener and likewise for all cathodes. Stimulation in both the head and the arm experiments was applied using a M×N-9 HD-tES System (Soterix Medical, New York, NY), at a constant total current of 4 mA for 12 minutes. While the participants received the stimulation, they performed a finger tapping task (FTT) as conceived by Karni et al. (30) and adapted by Bönstrup et al. (31). Afterward, they were asked to rate sensation levels using a visual analog scale for pain from 0 to 10 (Wong-Baker FACES pain scale; see Supplement), with 10 being the most severe, at three time points throughout the stimulation: at the beginning of the stimulation, at the middle of the stimulation, and right after the stimulation had ended. Participants were also asked separately for each time point to select all sensation qualities they felt from the following list: “No sensation”, “Tingling”, “Pricking/Stinging”, “Itching”, “Burning”, “Other”.

Statistical Analysis

Statistical analysis was performed using SPSS (IBM, Armonk, NY) and MATLAB (MathWorks, Natick, MA). Statistical analysis for sensation ratings in Experiments 1–5 was applied on mean ratings across repeated measures for each participant. All statistical analysis treats subjects as a random effect, thus removing any differences in mean ratings across subjects. One-way and two-way repeated measures analyses of variance (ANOVA) were performed to test for within-subjects main effects of waveform parameters, as well as interactions between factors. For Experiment 6 stimulation location is treated as a between-subjects effect, against a within-subjects effect of time. Mauchly’s Test of Sphericity was applied for all data, and degrees of freedom were Greenhouse-Geisser corrected where sphericity is violated.

The goal of this study was to explore a variety of waveforms to find substantial effects on sensation, rather than focusing on a single comparison to identify small effects. Thus, for all experiments we selected a priori a sample size of 10 participants which can only detect large effects (effects of 1.1 have a power of 80% at 5% significance in within-subject paired t-test). For Experiment 6 the sample size of 59 with stimulation on the head was opportunistic as we use all data available at the time of analysis from a parallel ongoing study.

Results

In search of a comfortable waveform for higher intensity electrical stimulation, we applied DC and AC stimulation to the forearm (Figure 1A) and asked participants to rate sensation level. A total of N = 50 people (19 unique) participated in five separate experiments. Each experiment tested different waveform parameters and compared them against the conventional DC and AC waveforms. Waveforms were presented in a randomized sequence of trials (7–10 second duration), with each waveform appearing several times throughout the experiment. Following each trial, participants rated sensation level on a categorical scale ranging from 1 to 5, corresponding to “No Sensation”, “Mild”, “Moderate”, “Severe” and “Extreme”. Overall the stimulation was well tolerated in all experiments, with only 45 of 3260 trials (<2%) rated as “Extreme”, and the majority of trials were rated “Mild” (39%) or “Moderate” (29%) (Figure 1B). Participants were allowed to interrupt the study at any time, but all participants completed their respective experiments.

Alternative monophasic (DC) waveforms do not reduce sensation

Recent experimental work argued that DC or AC stimulation, when applied with a pulsed carrier at high frequencies (>1 kHz) may reduce sensation (24,25). The rationale is that nerve fibers respond less to fast pulses, and thus cutaneous sensation may be reduced. To test this we explored net-DC stimulation with pulsed stimulation in two experiments. In Experiment 1 (N = 10 participants) we used a monophasic square waveform (Figure 2A) with 3 baseline and 3 duty cycle values in addition to constant DC resulting in 10 different waveforms (Figure 2B). Participants rated each waveform 9 times, totaling 90 trials per participant, presented in random order (total n = 900 trials). With increasing duty cycle and increasing baseline, we observed a reduction in mean sensation levels (Figure 2C). Two-way repeated measures ANOVA shows a main effect of duty cycle (F(2) = 4.93, ηp2 = 0.354, p = 0.0196), a main effect of baseline (F(1.072) = 16.9, ηp2 = 0.394, p = 2.04×10–3) and an interaction of these parameters (F(4) = 4.19, ηp2 = 0.317, p = 6.95×10–3). Sphericity was violated for the baseline effect (χ2(2) = 16.1, p = 3.20×10–4, Greenhouse-Geisser ε = 0.536). Both main factors of duty cycle and baseline point toward constant DC as the waveform with least sensation, and in fact, numerically that is the one with the lowest observed sensation. The interaction indicates that as the peak amplitude diminishes, the duty cycle becomes less influential on sensation. In the limiting case where “peak” amplitude equals the baseline value, the waveform becomes a constant current (Figure 2A) and the duty cycle has no effect on the waveform. Consequently, any peak amplitude that deviates from the constant current appears to increase sensation.

In Experiment 2 we tested the prediction that sensation would be reduced with increasing pulse frequency. To this end, we tested different frequency and baseline values (Figure 2D) and performed two-way repeated measures ANOVA with baseline and frequency as main effects. Similar to Experiment 1 we find a reduction of sensation with increasing baseline (F(2) = 7.78, ηp2 = 0.464, p = 3.68×10–3). This suggests once again, that the smallest deviation from baseline would produce the smallest sensation, which is indeed the case for DC stimulation. Sensation also decreased with increasing frequency (F(2) = 5.38, ηp2 = 0.374, p = 0.0147), but there was no interaction between baseline and frequency (F(4) = 1.044, ηp2 = 0.104, p = 0.398).

We speculated that perhaps the sharp ramp in rectangular waveforms caused nerve fiber activation (32) and thus increased sensation over constant DC. Thus in Experiment 3 we compared monophasic sinusoidal waveforms of different frequencies against constant DC (represented as 0 Hz), each rated 30 times by 10 participants (Figure 3). One-way repeated measures ANOVA of the data found a significant effect from frequency (F(1.77) = 7.35, ηp2 = 0.450, p = 6.78×10–3). Sphericity was violated for the frequency effect (χ2(14) = 33.4, p = 3.77×10–3, Greenhouse-Geisser ε = 0.353). Post-hoc pairwise comparisons of the waveforms found a significant difference between 1 kHz and all other frequencies (two-way pairwise t-test, q<0.05 FDR corrected over N=5 comparisons using the Benjamini-Hochberg procedures (3033)). At 1 kHz the mean sensation level was above “Moderate” and anecdotally participants reported a vibrating and itching sensation unique to that waveform. Sensations of the other waveforms were anecdotally reported by participants as burning and pricking, with intensities just above “Mild”.

Figure 3. Sensation level ratings for monophasic sinusoidal waveforms.

Sine waves with 3 mA amplitude were offset by 3 mA for a net current of 3 mA DC. Mean sensation ratings are shown with 95% confidence intervals. The sensation level at 1 kHz was significantly higher than all other waveforms, including constant current, shown as 0 kHz. The frequency effect was significant according to one-way repeated measures ANOVA and no waveform yielded intensity significantly less sensation than constant current.

Alternative biphasic (AC) waveforms do not reduce sensation

A recent proposal on how to reach deep targets in the brain was to use interference of high-frequency AC sinusoidal currents (27,28) which cause AM waveforms. A corollary of this work is that high-frequency AC stimulation would have little or no sensation on the skin surface. In Experiment 4 we test the sensation level ratings of AM biphasic sinusoidal waveforms (Figure 4A) with different pairings of carrier frequencies and amplitudes. Participants (N=10) rated each of the 12 waveforms 50 times (n = 600 total trials; Figure 4B). Two-way repeated measures ANOVA found significant effects from both carrier frequency (F(1.28) = 89.3, ηp2 = 0.908, p = 4.13×10–7) and amplitude (F(1.11) = 108, ηp2 = 0.923, p = 8.61×10–7) with interaction (F(6) = 10.0, ηp2 = 0.527, p = 1.98×10–7). Sphericity was violated for the carrier frequency effect (χ2(5) = 20.7, p = 1.05×10–3, Greenhouse-Geisser ε = 0.554) as well as the amplitude effect (χ2(2) = 13.0, p = 1.48×10–3, Greenhouse-Geisser ε = 0.426). The relationship between sensation level and carrier frequency appears to transition from exponential to linear as amplitude increases. The sensation effect of the carrier frequency is especially clear at 2 kHz, where the stimulation was almost imperceptible at all amplitudes, with mean ratings below “Mild”. Sensation quality was also noticeably different between different waveforms. There did not seem to be differences in quality between different carrier frequencies. Participants reported mostly pricking and tingling sensations at 1 mA amplitude, mostly vibration at 1.5 mA, and mostly pressure and vibration at 2 mA (Figure S1 in Supplementary Materials). Nonetheless, no apparent relationships between different sensation quality and AM sinusoidal waveform parameters were found.

Having established the effect of carrier frequency on sensation of AM sinusoids, we wanted to compare sensation between different modulation frequencies, and put that in context of unmodulated biphasic waveforms (Figure 5). In this Experiment 5 all waveforms had an amplitude of 3 mA and zero net current. Ratings were between “Mild” and “Moderate” for 1 kHz and between “Moderate” and “Severe” for 500 Hz. Reports of sensation qualities are shown in the Supplementary Materials (Figure S1). The effect of the carrier frequency was significant (F(1) = 118, ηp2 = 0.844, p = 6.51×10–5) according to 2-way repeated measures ANOVA. Waveform shape also had a significant effect (F(3) = 16.5, ηp2 = 0.647, p = 2.75×10–6). There is no interaction between waveform shape and carrier frequency (F(3) = 1.95, ηp2 = 0.178, p = 0.145). Focusing on the AM waveforms only, we see that higher modulation frequency produces lower sensation (paired t-test on the mean between 0.5kHz and 1kHz: t(18)=2.57, p=0.019). Overall it is clear that the AM modulation is in the same range of sensation as the unmodulated waveforms. In total, for AC stimulation we conclude from Experiment 4 and 5 that the main factor for sensation is carrier frequency, with the lowest sensation ratings for the conventional unmodulated sinusoids.

Figure 5. Comparison of AM waveforms with unmodulated waveforms.

Mean ratings are shown with 95% confidence intervals. Two-way repeated ANOVA found significant effects from carrier frequency and waveform shape, but no interaction between the factors. The difference between sinusoidal waves modulated at 5 Hz and 10 Hz was significant when averaged between both carrier frequencies.

Skin sensation is the same between head and forearm

Thus far all results have been obtained with sensation ratings on the arm. To determine whether sensations are similar to a typical transcranial stimulation experiment, we reproduced in Experiment 6 the design of an ongoing HD-tDCS experiment. In this experiment subjects are engaged in a sequence learning task (pressing keys in a specified order) with their right hand, while receiving tDCS at 4 mA over the contralateral motor cortex (with 1 mA through each of 4 electrode pairs). In a separate group performing the same task we stimulated the contra-lateral forearm with the same number of electrodes, current, and electrode spacing. Sensation ratings were obtained at the beginning, middle and immediately after the 12-minute stimulation session (Fig. 6A). Two-way repeated measures ANOVA on the data yields no main between-subjects effect from stimulation location (F(1) = 5.05, ηp2 = 0.0110, p = 0.390), a main within-subjects effect of time (F(1.48) = 14.5, ηp2 = 0.178, p = 2.69×10–5), and no interaction between stimulation location and time (F(1.48) = 0.348, ηp2 = 5.17×10–3, p = 0.643). Sphericity was violated for the time effect (χ2(2) = 28.1, p = 7.67×10–7, Greenhouse-Geisser ε = 0.742). The null hypothesis is favored with a Bayes factor of 4.14 over an alternative model assuming stimulation location effect with a Cauchy prior scale of square root(2)/2, following the computational methods of Rouder et al. (34). Consistent with the lack of interaction between stimulation condition and time, visually the sensation levels follow the same adaptation pattern over time, decreasing incrementally from a mild to moderate rating below 4 on the VAS at the beginning to a very mild rating between 1 and 2 by the time the stimulation has ended. Likewise, specific sensation qualities (“No sensation”, “Tingling”, “Pricking/Stinging”, “Itching”, “Burning”, “Other”) are reported equally often between transcranial and arm stimulation (Figure 6B). Not only is there no significant difference with stimulation location, but the change in sensation over time is also quite similar. In summary, these results suggest that DC stimulation causes comparable skin sensations on the scalp and arm.

Figure 6.

Comparison of skin sensation on the scalp and the arm. (A) Mean VAS sensation levels of 4mA stimulation on the head (n = 59) and the forearm (n = 10), with a rating of 10 representing maximal severity of sensation, reported at the beginning of stimulation, middle of stimulation, and after stimulation. (B) Sample proportions of reported sensation qualities from the same stimulations in (A). “B”, “M”, and “A” represent the three time points of “Beginning”, “Middle”, and “After”.

Qualitative side effects

Anecdotally, a majority of participants reported desensitization and slight numbness toward the later halves of the DC experiments (Experiments 1–3), but mean sensation level ratings across all subjects in those experiments do not show a significant decrease over time (Figure 7). The reduction in sensation may be the result of well-established neuronal adaptation, typical for skin sensations (35). The anecdotal reports of minor numbness requires further investigation. Mean ratings appear to be negatively correlated with time, and simple least-squares regressions of these data suggest that time weakly predicts a decrease in sensation level with DC stimulation. Nevertheless, sensation levels still varied with the different waveforms toward the ends of the experiments. Hypoalgesic effects were not reported in the AC experiments, and the sensation levels are not correlated with time for Experiments 4–5 (Figure 7). Throughout all experiments the participants were allowed to take short breaks between trials for any numbness to subside. Approximately half of those in Experiments 1 and 2, each comprising 90 trials, took one or two breaks of about 1 minute in duration. Most participants took about 10 seconds between trials to rate sensation levels, allowing some time for the sensations to fade.

Figure 7.

Mean sensation level ratings over all trials in Experiments 1–5, shown in (A)-(E), respectively. A least-squares linear regression model was fit for each experiment, with R-squared values shown in each plot. The coefficients of determination are higher for Experiments 1–3, which used only monophasic waveforms, than those for Experiments 4–5, which used only biphasic waveforms. Shaded areas represent 95% confidence intervals.

Pre- and post-stimulation questionnaires were used to rate the levels of any tingling, itchy, burning, or numb sensations they felt at those points, on the same 1–5 scale of corresponding levels ranging from “No Sensation” to “Extreme”. The changes in the specific sensations following each experiment are shown in Figure 8. There was a small increase in all types of sensation after stimulation, but there was no significant difference between experiments. Anecdotally, participants reported that any sensation they felt immediately after the experiments subsided in less than 30 minutes. Skin redness was observed across almost all participants following all experiments. Those who participated in Experiments 1–2 reported that skin coloration did not fully fade until as long as 2 to 3 days after stimulation. Skin coloration reportedly faded within 3 hours following Experiments 3–5. This disparity is likely due to the long stimulation duration in Experiments 1–2 compared to the other experiments. A combination of high current density due to a high net current intensity applied over a small area, with prolonged exposure, as well as use of alcohol during skin preparation, may have caused some persistent skin irritation.

Figure 8.

Mean changes in levels of specific sensations reported in pre- and post-stimulation questionnaires. Sensation levels of all types increased following all experiments. There were no differences between experiments in the amounts of change. Error bars represent 95% confidence intervals.

Discussion

Previous efforts to reduce sensation of transcranial electrical stimulation have not considered varying waveform parameters except frequency. To address this gap, we explored alternative waveforms that either inject a net current as in tDCS, or deliver charged balanced stimulation as in tACS. We compared these alternative waveforms to conventional tDCS and tACS and parse the effects of waveform parameters on sensation level. The goal was to find the most promising waveforms to deliver maximal current intensity with minimal sensation. Our results suggest that the conventional tDCS and sinusoidal tACS are already optimal in terms of reducing sensation at a given current intensity.

Most tDCS protocols use a constant current and our data suggest that any waveforms where stimulation does not remain constant over time cannot reduce sensation, except for the typical ramp-up and ramp-down which we used here as well. Compared to 15 unique high-frequency (> 1 kHz) monophasic square waveforms delivering net 3 mA, sensation of a constant 3 mA current was either significantly less intense or no more intense. Likewise, it was no more intense than 3 mA high-frequency monophasic sinusoidal waveforms from 1 to 50 kHz. In consideration of hardware design, constant tDCS is the most practical to implement because the required range for compliance voltage is the smallest and the required circuitry is simpler, as it precludes the need for a function generator or a software equivalent. Therefore, because there is no significant sensation reduction or any other advantage in adopting alternative DC waveforms, constant current remains the ideal waveform for tDCS.

Although no alternative DC waveforms lowered sensation level, several equaled 3 mA constant current in ratings. The use of periodic DC waveforms allowed us to observe the patterns in parameter effects on sensation level. Duty cycle, baseline offset, and frequency were found to have significant effects on sensation level, while peak amplitude was set as the dependent parameter. It is likely that the short pulse width of the “active” state of a monophasic square waveform does not effectively alleviate the intensity of a high peak amplitude. As both duty cycle and baseline increase, the square waveform approaches the linear shape of constant DC, correlating with a decrease in sensation level. The effects of duty cycle and baseline may therefore explain why constant current causes the least sensation. Conversely, existing works on transcutaneous stimulation showed that shorter monophasic pulses led to lower levels of sensation (36,37). In those cases, either the frequency or the net current was changed with the pulse width, whereas here the shorter pulses were associated with peak amplitude. Our comparisons of high-frequency monophasic sinusoidal stimulation with constant current did not show a difference at 2 kHz and above, similar to the negative result with 0.8 Hz (26). However, the sensation was significantly higher at 1 kHz (and even more so at lower frequencies not tested due to discomfort), which highlights that there is a range of frequencies to which peripheral nerves are more sensitive (38).

Like tDCS, tACS treatments are limited in waveform options, typically to the sinusoid. We found that at 1 kHz carrier frequency, the regular sinusoid is significantly less intense than square waves and amplitude-modulated sinusoids. Lower frequencies in the range of endogenous neural oscillations may be preferable for entrainment, and using temporal interference (TI) stimulation proposes pairing electric fields at kilohertz frequencies to produce an amplitude-modulated field envelope inside the brain that neurons can respond to (27,28). The same strategy has also been adopted earlier for transcutaneous stimulation using 4 kHz carrier frequency, although no differences in sensation were found from unmodulated square pulses (39,40). Similar AM modulation in deep brain areas might be achieved by directly applying AM modulation of the scalp surface with multiple electrodes (41). Inspired by these methods, we implemented amplitude modulation at the electrode level to test how the envelope would feel on the skin. Despite the differences from unmodulated sine, the sensations for AM sine are still well tolerated below a “Moderate” level when using a carrier frequency of 1 kHz or higher. These ratings indicate that TI stimulation and AM tACS may be viable options.

The findings of changes in sensation with carrier frequency in AC stimulation are consistent with past ones. They generally agree that skin sensation of sinusoidal AC currents decreases with frequency in the range of 100 to 10,000 Hz (23,42,43). Our measurements of sensation level show a significant decrease from a 500 Hz sinusoid to a 1 kHz sinusoid. Furthermore, this difference holds true for square waves and AM sinusoids. As seen in the differences within the 100 to 2,000 Hz carrier frequency range, sensation intensities of modulated sinusoids follow the trend of unmodulated sinusoids, contradicting a previous report of increasing sensation levels up to 10 kHz (44). The significant differences in sensation between 5 Hz and 10 Hz modulation frequencies at both 500 Hz and 1 kHz carrier frequencies contrast with the absence of an effect of modulation frequency between 0 and 100 Hz when using 4 kHz carrier frequency (40). Both results are consistent with the significant decrease in sensation level at high carrier frequencies we found. AC waveforms with carrier frequencies above 2 kHz were not tested because of the already low sensation intensities at 2 kHz.

A wide variety of alternative waveforms for tDCS and tACS were compared. Unexpectedly, we find that conventional waveforms remain the optimal options for minimizing peripheral nerve stimulation. The effects of periodic waveform parameters on skin sensation were characterized to a limited extent, as a range of other waveform factors remain to be explored. For example, others have demonstrated with charge-balanced biphasic square waves that the symmetry of the waveform shape does not affect sensation (45) and neither does the polarity (anodal vs. cathodal; 37). tRNS was not included in our comparisons. The dependence of tRNS waveforms on different frequency components could prove to be a challenge to investigate, but the difference in sensation from tDCS and tACS or lack thereof (17,22) could be examined further. Meanwhile, we could expand the ranges of the parameters we tested, such as the duty cycles of square waves, which were limited to a maximum of 30%. Given the effect of decreasing sensation with increasing duty cycle, duty cycles above 50% may be worth considering. In addition, only high-frequency waveforms were tested here. Based on existing studies, the relationship between the carrier frequency and sensation level is more complex for AC stimulation below kilohertz frequencies (23,42,43,46).

Lastly, the waveforms outlined in this study were not applied transcranially, so our results do not take into account sensory effects other than skin sensation that can occur at for AC stimulation at 6 mA when applied to the head (17). A more comprehensive assessment of waveform viability would include reports of visual and auditory responses to different waveforms as done by Turi et al. (46) as well as by Zeng et al. (23). Nonetheless, our direct comparisons of sensations in high intensity 4mA stimulation on the head and the arm show that sensation levels or DC stimulation are the same, including the desensitization patterns over time. Moreover, specific sensation qualities were reported at similar rates for the head and the arm. If there were in fact any difference, it appears that the sensation levels in the arm may be slightly higher. Notably, DC stimulation on the head at 4 mA elicited only skin sensation. Thus, our findings on skin sensation on the arm are likely to carry over to sensation on the scalp, at least for DC stimulation, which consistently elicited the weakest sensation of all waveforms tested here. This study is limited by the lack of a distinct option to report pain in the qualitative assessments of the waveforms, since many stimulation conditions were uncomfortable but not necessarily painful. It would be worthwhile in the future to more precisely distinguish waveforms that elicit more or less nociception.

One caveat of this exploratory study is that it was not properly powered to detect small effects. Rather, our goal was to explore a variety of waveforms identity the most promising contrasts. Therefore, as in any other study, any null results reported here are not evidence for a lack of an effect. The only exception is the comparison between arm and head sensation, where our statistical analysis favors the interpretation that there is no difference between the two.

Highlights.

• Substitute waveforms for monophasic stimulation do not reduce skin sensation

• Substitute waveforms for biphasic stimulation do not reduce skin sensation

• Conventional waveforms elicit minimal skin sensation

• Skin sensation of direct current stimulation is the same on the head and the arm